Allocating the Costs of Motor Vehicle Crashes Between Vehicle Types

TRANSPORTATION RESEARCH RECORD 1635

Paper No. 98-0067

81

Allocating the Costs of Motor Vehicle Crashes Between Vehicle Types TED MILLER, DAVID T. LEVY, REBECCA S. SPICER, AND DIANE C. LESTINA When a truck and an automobile are involved in a crash, the harm to occupants tends to vary with the weight of the vehicles involved. In determining the appropriate level of government expenditures for traffic safety, costs in multivehicle crashes involving different vehicle types must be allocated between occupants and nonoccupants of a particular vehicle type. Four methods for allocating costs among different vehicle types are considered, corresponding to different perspectives, including that of occupants of a vehicle and that of society under different property right assignments. Costs based on the four allocation methods for the United States as a whole and per vehicle mile are also estimated. The allocation method was found to have large effects on the relative magnitude of costs.

Efficient allocation of public dollars to passenger car, truck, and bus traffic safety should depend upon the social costs associated with crashes involving each of these vehicle types. In addition, it may be desirable that fees (e.g., registration and user fees) and taxes on vehicles reflect the costs to society of crashes. While these charges are generally differentiated by vehicle type, they currently tend to reflect expected wear and tear on roads, not public safety concerns. In determining crash costs, practical difficulties arise in allocating costs in multivehicle crashes involving different vehicle types. A reasonable approach might be to allocate costs to the vehicle that caused a crash, but accurate information on the responsibility of each vehicle in a multivehicle crash is generally not available; often responsibility is shared. Even if responsibility for each crash is known, lighter at-fault vehicles may sustain greater damage due to the presence of heavier vehicles on the road. When a passenger car underrides or crashes head-on into a heavy truck, it may be argued that at least part of the excess cost to the car and its passengers should be allocated to the truck. Economists have examined how to allocate crash costs in multivehicle type crashes in the context of optimal road use decisions. The analysis is in terms of the average costs that can be expected to be imposed on others by a particular vehicle type, that is, ante accident externalities. Extending earlier work by Vickrey (1) and Newberry (2), Janson (3) derives these charges when crashes involve a vehicle (the more protected party) and a cyclist (the unprotected party). The author shows that the charges depend upon each party’s percentage change in crashes per 1 percent change in miles driven. When crashes increase proportionately to the increase in road use by both cars and bicycles, the optimal charge for cars varies between zero (when the elasticity of crashes with respect to car use is zero) and the full cost of the crash (when the elasticity is one). Recently, Persson and Odegaard (4) estimated the accident externality charge per vehicle mile for different vehicle types in 14 European countries under the assumption that crashes increase proportionately with vehicle use. T. Miller, R. S. Spicer, and D. C. Lestina, National Public Services Research Institute, 8201 Corporate Drive, Suite 220, Landover, MD 20785. D. T. Levy, Department of Economics, University of Baltimore, Baltimore, MD 21201.

This paper considers costs by vehicle type. Rather than focusing on externality charges per se, we examine the more general question of how to allocate costs between different vehicle types. We suggest four different methods corresponding to different perspectives. The different perspectives include that of occupants of a vehicle and that of society under different property right assignments that may be relevant in allotting government expenditures for traffic safety or assessing fees. We then estimate costs based on the four allocation methods. We present estimates for the United States as a whole, and per vehicle kilometer and per vehicle. Unlike many earlier estimates of U.S. crash costs—for example, Rice et al., National Safety Council, NHTSA, Blincoe, and Hartunian et al. (5–9)—we present estimates that include the costs of personal losses. ALLOCATION OF COSTS IN MULTIVEHICLE CRASHES INVOLVING DIFFERENT VEHICLE TYPES Cost in multivehicle crashes involving different vehicle types must be allocated between occupants and nonoccupants of a particular vehicle type. In developing methods for allocating costs, emphasis is placed on the costs associated with injuries and fatalities rather than property damage, because they make up most of the costs. Harm to occupants in multivehicle type crashes varies with the weight and class of the vehicles involved: occupants of the heavier vehicle tend, on average, to fare better (10). Body types were grouped into eight different vehicle categories, using General Estimates System (GES) and Fatal Accident Reporting System (FARS) classifications of body types. The categories were chosen in terms of traditional classifications and median body weight. The categories were placed in order from light to heavy: • Motorcycles—Motorcycles and mopeds. • Passenger Car/Van—Passenger cars and vans, including minivans, passenger vans, and sport utility vehicles. • Light Trucks—Light trucks, including larger vans, pickups, light truck–based vehicles; gross vehicle weight less than 4 540 kg (10,000 lb). • Unknown/Other—Vehicles of unknown type and other miscellaneous vehicles such as street cleaners and snowmobiles. • Unknown Truck—Unknown medium and heavy trucks, including farm and construction equipment other than trucks, and motor homes with gross vehicle weight exceeding 4 540 kg (10,000 lb). • Other Single Truck—Other single medium and heavy trucks, including step vans, single unit straight trucks; gross vehicle weight exceeding 4 540 kg (10,000 lb). • Buses—All buses. • Combination Truck—Tractor trailers, triples, truck tractor without trailer.

82

Paper No. 98-0067

This ordering is roughly based on averages. Within category, weights may vary from vehicle to vehicle, especially at the higher end, depending on the weight of the load being carried. Costs of motor vehicle crash victims were computed using the following four methods. Method 1 The cost of each victim is assigned to the vehicle type that he or she occupied. For example, if a passenger car hits a light truck, costs of passenger car occupants are attributed to the passenger car, and costs of the light truck occupants are attributed to the light truck. This method makes no assumptions about responsibility for the crash, injury severity, or how crash costs are affected by vehicle types. This method reflects the view that the occupants of each vehicle type bear their own risk. Measures derived using this method (e.g., cost per vehicle mile) are useful in gauging the safety of occupants in a particular vehicle type. Method 2 All costs are assigned to every vehicle in the crash. For example, if a passenger car hit a light truck, injuries for the passenger car and the light truck would all be attributed to both the passenger car and the light truck. Due to multiple counting of all multivehicle crashes, the total cost of injuries over all categories is the actual cost of the injuries multiplied by the number of vehicles in the crash. To compare Method 2 results with other methods, we normalized Method 2 costs: total costs for each vehicle type are multiplied by the ratio of total crash costs for all vehicles and Method 2 multiple-counted “total” costs for all vehicles. This method is consistent with the approaches suggested in the optimal accident externality price literature by Vickrey, Newberry, and Janson (1–3). From that viewpoint, optimal road use dictates that each driver in a multivehicle crash take into account his or her own costs as well as those to others. No judgment is made about which party is responsible for the crash, but the method incorporates the higher costs from crashes involving heavier vehicle types. Method 3 Method 3 uses the ranking of vehicle types from light to heavy given above. In multivehicle crashes, the heavier vehicle is allocated the excess cost of injury in the lighter vehicle when these injuries are compared to injuries in a multivehicle crash that involved only the lighter vehicle type. By this method, the heavier vehicle is assigned responsibility for any increased injury severity in the lighter vehicle. For two-vehicle crashes involving vehicles of the same type, costs in each vehicle are assigned to its own vehicle type. For injury costs in two-vehicle crashes involving different vehicle types, the lighter vehicle type is assigned the average cost per occupant in a crash involving two vehicles of the same type. The heavier vehicle type is assigned the residual cost of the injuries in the lighter vehicle (i.e., their actual cost minus the cost assigned to the lighter vehicle) plus the cost of injuries in the heavier vehicle. When the cost of a same-type lighter vehicle crash is greater than the cost to the lighter vehicle in a light-heavy vehicle crash (which generally arises when sample sizes are small), the costs of injury in each vehicle are assigned to its own vehicle type.


For fatal crashes involving two vehicles, a similar methodology was utilized. Since the costs associated with a fatality are the same no matter what vehicle type, the excess fatalities associated with the difference in case-fatality rate in mixed-vehicle crashes—measured by the ratio of fatal to nonfatal injuries—was assigned to the heavier vehicle. The rules become ambiguous when a crash involves three or more vehicles, especially when there are more than two vehicle types. For crashes involving three or more passenger cars or light trucks, each occupant’s costs are assigned to their vehicle type. For any three or more nonfatal vehicle crash involving a heavy vehicle (other/ unknown, other single truck, unknown truck, bus, combination truck), the passenger car or light truck is assigned the average cost per occupant in a three or more vehicle crash involving only passenger cars and light trucks. The heaviest vehicle type is assigned the residual cost of injuries in the passenger cars or light trucks (i.e., their actual cost minus the cost assigned to them), plus the cost of injuries in the heavier vehicle. Due to sample size limitations for fatal crashes involving three or more vehicles, each vehicle was assigned the costs associated with its occupants (i.e., Method 1). Exceptions to the general allocation rules were also made for motorcycle and nonoccupant (e.g., pedestrians and pedalcyclists) costs. Costs per motorcyclist, as for nonoccupants, are exceptionally high regardless of the other vehicle involved. Indeed, cost per rider in a two-motorcycle crash was higher than the cost per motorcycle involved in a crash with any other vehicle type. Since these costs are not due primarily to the weight of the other vehicle (i.e., they reflect the lack of protection afforded to riders), all motorcycle occupant costs were assigned to the motorcycle itself (i.e., using Method 1). Nonoccupants also suffer greater injury due to the lack of protection, but are not subject to motor vehicle regulations such as registration fees, taxes, or design requirements. Furthermore, the injury cost when two pedestrians collide should be minimal. Consequently, they are not assigned separate costs under Method 3. For nonoccupants in fatal and single vehicle crashes, all costs were assigned to the striking vehicle. In a nonfatal nonoccupant crash involving more than one vehicle, the GES data did not distinguish the striking vehicle. Nonoccupant crash costs were then assigned to the heaviest vehicle in the crash. In the case of nonoccupant fatalities, FARS identified the striking vehicle. All nonoccupant crash costs are then assigned to that vehicle type. Method 3 has the advantage that it is directly based on differences in average crash costs between vehicle category. While neither party is designated as responsible for the crash, the heavier vehicle bears the excess costs of the lighter vehicle. This measure is useful for determining public expenditures on different vehicle types, since it implicitly takes into account weight and vehicle class in assigning costs. There is an arbitrary element to assigning costs based on average injury crash costs and likelihood of fatality, but it directly relates to the relative harm to occupants. Assignment of costs in the case of motorcyclists and nonoccupants is also somewhat judgmental and depends upon the purpose of the analysis.

Method 4 In Method 4, all costs are assigned to the heaviest vehicle. For example, in a crash between a light truck and a passenger car, injury costs from both the passenger car and light truck are assigned to the light truck, and no costs are assigned to the car. Method 4 uses the same ranking of vehicle types, except that the rankings of combination trucks and buses are reversed. As a passenger vehicle, costs of bus

Miller et al.

crashes with other heavy vehicles are high due to the injury and fatality costs of the many passengers. Since these are not due to weight, but rather to the nature of the “cargo,” their higher costs were assigned to their own vehicle type. We do not, however, take into account any such differences for other vehicle types. This method maintains additivity over categories, but makes the extreme assumption that the heavier vehicle assumes all responsibility for the crash. This method is widely used in tabulating crash counts and costs by vehicle type (11,12,13). Here it is presented largely for comparative purposes, but would be relevant from the perspective that lighter vehicles (e.g., passenger vehicles) have property rights to the roads. MEASUREMENT OF CRASH COSTS Motor vehicle crashes may involve three types of outcomes: fatalities, nonfatal injuries, and property damage. The analysis further categorizes injury survivors by degree of severity and type of injury. Following Miller (14), incidence is multiplied by the cost per victim for injured motor vehicle crash victims by body region and Abbreviated Injury Scale (AIS) threat-to-life severity (15) to compute annual crash costs. Incidence To estimate injury incidence, we followed procedures similar to those of Blincoe and Faigin, Miller and Blincoe, Miller, Galbraith et al., and Miller et al. (16–19). We apply the methods developed in these studies to NHTSA’s GES, Crashworthiness Data System (CDS), National Accident Sampling System (NASS) and FARS data to estimate the cost of crashes in the United States. (11,20–22). GES, CDS, and NASS are nationally representative samples of police-reported crashes, and FARS is a census of fatal crashes in the United States. GES provides a sample of nonfatal crash injuries in the United States by police-reported severity (KABCO) for all crash types. Miller et al. (12) and Blincoe and Faigin (16 ) pointed out the great diversity in KABCO coding across cases and developed weighting factor corrections, which we applied to compensate for a systematic undercount of police-reported cases in GES,CDS, and NASS. Police reporting is not an accurate medical description of the injury. It is made by a nonmedical professional who may not even examine the victim. To minimize the effects of variability in police coding, 1988–91 CDS (CDS changed its injury coding system in 1992 to a system we could not readily cost) and 1982–86 NASS were used to obtain a medical description of the injury by body region and AIS. CDS describes injuries to passenger vehicle occupants involved in towaway crashes. The 1982–86 NASS data provide the most recent medical description available of non-CDS crash victims. The GES data are used to weight the CDS and NASS data so they represent the 1992–93 GES injury victim mix in the CDS and NASS sample strata. In applying these weights, we controlled for police-reported injury severity; vehicle type; and restraint use (belted, unbelted, unknown, in a child seat). The NASS data were coded with the 1980 version of AIS, which differs slightly from the 1985 version; NHTSA, however, made some of the AIS-85 changes well before their formal adoption. At the completion of the weighting process, we had a hybrid CDS-NASS file with weights that summed to the estimated annual GES incidence by police-reported injury severity and other relevant factors. Fatality counts came from the 1993 FARS (22).

Paper No. 98-0067

83

The next step was to adjust the weights on nonfatal cases to correct for known GES undercapture of police-reported crashes (8,12,16,19) and for underreporting of crashes to the police. The undercapture correction is a multiplier of 1.12 (8,16 ). The percentage underreporting for crashes without injury is from Blincoe (8). The undercount correction for injury survivors, an average multiplier of 1.088 that varied by AIS, came from Miller et al. (19). Following Blincoe and Faigin (16 ), we distributed the unreported cases in proportion to reported nonhospitalized cases by AIS. Costs Costs are in September 1995 dollars. We include the following categories of costs: • Medically Related Costs—These include hospital, physician, rehabilitation, prescription, and related payments. Also included are coroner and burial costs for fatalities, and claims processing costs of medically related loss compensation through insurance and the courts (omitting time spent on the loss recovery process). Nonfatal medical care and coroner costs are from Miller (14). • Emergency Services—These include police, fire, ambulance, and helicopter services. • Property Damage—These are the costs to repair or replace damaged vehicles and property, including the costs of damage compensation. • Lost Productivity—This includes wages, fringe benefits, and household work lost by the injured, as well as the costs of processing productivity loss compensation claims. It also includes productivity loss by those stuck in crash-related traffic jams and by coworkers and supervisors recruiting and training replacements for disabled workers, and repairing damaged company vehicles. Excluded are earnings lost by family and friends caring for the injured and the value of schoolwork lost. • Quality of Life—This values the pain, suffering, and quality of life that the family loses because of a death or injury. For fatalities, this loss is computed from the amount people routinely spend (in dollars or time) to reduce their risk of death and injury. It is based on studies of the explicit or implicit expenditures on auto safety features and smoke detectors, and extra wages paid to workers to take risky jobs. To avoid double counting, we subtract lost productivity from quality of life lost estimates. For nonfatal injuries, the costs are developed from estimated quality-adjusted life years lost. This study uses a major upgrade (8,19,23) of the related highway crash cost estimates in Miller et al. (12), Blincoe and Faigin (16 ), and Miller (14). Lifetime productivity and quality of life losses are reestimated with a 2.5 percent discount rate, at the high (conservative) end of the rates typically used in liability compensation (24). A higher (lower) discount rate would lower (raise) future costs, especially those that occur farther into the future. The lifetime earnings and household production loss models in Miller et al. (12) are updated with 1990–91 demographic data, earnings profiles, and life tables. Employer productivity losses are recomputed using the assumptions in Miller (25). We assume that supervisor and coworker staff time lost to a permanently disabling injury equal the losses for a fatality. Finally, the cost estimates use improved insurance administrative and legal expense models. The models introduced a $100,000 average policy limit on liability claims and a $500,000 limit on average court awards for catastrophic injuries. Legal costs were reestimated with unit costs from Kakalik and Pace (26 ) and probabilities of lawsuit from Hensler

84

Paper No. 98-0067

et al. (27 ), as well as the updated medical care and productivity loss estimates. We present human capital and willingness to pay cost estimates. The human capital method, used in cost of illness studies, focuses on productivity losses. This method places low values on children and old people, values women less than men, and excludes pain, suffering, and lost quality of life [see e.g., Mishan, Jones-Lee, and Scodari and Fisher (28–30)], but is used in many cost of injury and cost of illness studies due to difficulties in measuring pain and suffering [see e.g., Rice et al. (5)]. The willingness to pay method includes personal losses from pain and suffering, and lost quality of life from injuries and fatalities (31,28). This is the theoretically more appropriate method, but places substantially greater emphasis on harm to individuals than does the human capital approach. Since they are often relevant from a public perspective, external costs—costs not borne by the crash-involved transport user—were also calculated. In developing external cost estimates, costs borne by vehicle occupants are excluded under the assumption that drivers accurately assess and internalize their own risks and those of their passengers. These costs include occupant quality of life losses and human capital payments except employer productivity losses (training, foregone taxes, and medical payments and property damage not covered by insurance). Nondriver occupants are often family members. Prior studies indicate that willingness to pay is often the same or higher for family members [see e.g., Jones-Lee (32)]. In an actuarially fair insurance system, policy holders pay their expected crash costs over time (e.g., through surcharges on automobile insurance and premiums paid in years with modest claims costs). Because it is unclear how much of the costs are actually passed on to the driver, we define three external cost measures. One measure, labeled immediate external costs, excludes costs covered by health and automobile insurance even though at least some of the insurance costs are passed on to the policy holders. A second measure, which we call ultimate external costs, treats all costs reimbursed by insurers as internal. This measure recognizes that in an actuarially fair system, policy holders ultimately pay their expected costs over time. This method is similar to the approach taken by Persson and Odegaard (4), who exclude property damage costs because they are “paid for as part of the compulsory insurance premium a motorist is required to pay.” Finally, for heavy trucks (and other commercial vehicles), the employer is an internal payer. From this vantage point, we computed heavy truck costs external to owners and operators. The external cost measures are computed by vehicle type using Method 3, which takes into account the excess burden to lighter vehicles from being in a crash with a heavier vehicle. Thus, excess light vehicle occupant costs resulting from crashing into a heavy rather than a light vehicle are counted as external costs attributed to the heavy vehicle. In computing ultimate external costs, we also employed Method 2 automobile insurance costs by subtracting them from Method 3 to compute excess lighter vehicle automobile insurance costs allocated to heavy trucks. These costs are treated as internal.

COST ESTIMATES BY VEHICLE TYPE AND BY ALLOCATION METHOD Table 1 summarizes estimated crash costs per victim by policereported victim injury severity and cost category (human capital costs and willingness to pay costs). The data should be treated with caution, since police-reported injury severity often is inconsistent with reality. CDS-NASS data show that many victims whom the police


TABLE 1 Cost per Victim by Police-Reported Injury Severity and Cost Category

code as O—“Property Damage Only”—(6 ) are actually injured; for clarity, we renamed this category “No Injury Observed.” Without the ability to examine the victim, police have been known to code bloodspattered crash victims as “A - Disabling” injury. Indeed, the majority of “A - disabling” injuries do not result in hospital admission (12). We recoded all fatalities as “K - Fatal,” regardless of police-coded injury severity. Tables 2 and 3 present the total cost estimates using, respectively, the human capital and willingness to pay approaches. The estimates are presented by vehicle type and by the four methods for allocating costs in multivehicle crashes. Total costs using the willingness to pay measure are $373 billion using each of the four methods. Human capital costs for most vehicle types are about 40 percent of willingness to pay cost estimates. For motorcycles, they are about 35 percent of willingness to pay estimates due to the high proportion of serious injury and death. The rankings of total costs by vehicle type are fairly consistent over the cost approaches and allocation methods. Generally, passenger cars make up the largest percentage of total costs, light trucks are second, followed by combination trucks, other unknown, motorcycles, unknown trucks, other single trucks, and finally buses. The lowest positions in the ranking depend upon types of costs and the allocation method. The rankings reflect the prevalence of each vehicle type on the roads. The percentage of costs by vehicle type is also presented in Tables 2 and 3. These measures provide a useful indicator of the relative importance of different vehicle types in the total composition of crash costs. Their rankings are, of course, consistent with those of total costs. Nonetheless, the percentages vary substantially with the allocation method used. For example, based on the willingness to pay approach, passenger cars comprise from about 59 percent of all costs allocated by Method 4 to 70 percent of all costs allocated by Method 2. As expected, trucks contribute more to total costs using Methods 3 and 4. In general, the percentage of total costs made up by a particular vehicle type is sensitive to the allocation method. In gauging vehicle safety, we control for differences in vehicle type presence on the roads using two exposure measures: vehicle kilometers traveled and number of registered vehicles. Vehicle miles traveled by vehicle type were computed from the National Personal Transportation Survey (33); registered vehicle counts are from NHTSA’s Traffic Safety Facts 1992 (11). Table 4 presents costs per vehicle mile traveled and per registered vehicle for each vehicle type using the willingness to pay approach and Method 3. Unknown truck costs were allocated proportionately between other single truck and combination truck; other/unknown vehicle costs were allocated proportionately among all vehicle types. Results obtained using the human capital approach and other allocation

Miller et al.

Paper No. 98-0067

TABLE 2

Human Capital Costs* of Highway Crashes by Method of Analysis, United States, 1993

methods were generally subject to the same variations as those observed above for total cost measures. Per vehicle mile traveled, motorcycle crashes result in nearly 13 times the costs of passenger car crashes. The discrepancy is even greater between motorcycles and light trucks. Per vehicle mile, costs resulting from crash-involved light trucks are about 70 percent that of passenger car costs. Combination truck costs per vehicle mile are 50 percent higher than passenger car costs, largely due to the harm to occupants in other vehicles resulting from the weight of combination trucks. Bus costs per vehicle mile are almost four times passenger car costs. This reflects the large number of passengers buses carry. The discrepancy between costs of motorcycles and other vehicles is less when costs are in terms of registered vehicles instead of vehicle miles. Nevertheless, the costs per motorcycle are still more than twice passenger car costs. The reduction in the variation in costs is largely due to the high mileage of other vehicles, particularly of trucks, compared to motorcycles. The relative cost per registered vehicle increases the most for combination trucks, due to the exceptionally high number of miles traveled by these commercial vehicles. Table 4 also distinguishes total costs into single- and multivehicle crashes. This distinction provides some insight into the types of crashes experienced by different vehicle types and the vehicle responsible for the crash. In single vehicle crashes, the vehicle responsible for the crash is not an issue, nor is cost allocation between vehicles. Nearly half (42 percent) of motorcycle crashes are single vehicle, while other vehicle types are involved in three to five times

TABLE 3

85

as many multivehicle as single vehicle crashes (11). The costs of single vehicle crashes relative to multivehicle crashes to some extent reflect these differences. Other single vehicle truck injuries in multivehicle crashes may be more severe. Motorcycles’ propensity for single vehicle crashes suggests that they are disproportionately at fault in their crashes. Table 5 presents external cost estimates using the different methods for calculating external costs. External costs for all vehicles are about 35 percent of total willingness to pay estimates and about 80 percent of total human capital estimates. Within each vehicle type, passenger car and, particularly, motorcycle costs comprise a smaller portion of external costs than total costs; heavier vehicle costs make up a larger portion of external costs. Ultimate external costs are lower than immediate external costs, as expected. The differences are relatively small, however, except for motorcycles. Table 5 also presents the excess external burden imposed by trucks and buses. Combination trucks make up the bulk of these costs (65 percent), followed by other single trucks (25 percent). The optimal road use literature (1,2,3) argues that taxes on cars should reflect their safety costs. Fees and taxes are often differentiated by vehicle type. Registration fees and taxes can vary by vehicle weight, and the amount of gasoline taxes collected depends on vehicle miles traveled. If these collections are to be related to traffic safety costs, it is useful to examine cost per vehicle mile traveled— which is directly related to gasoline consumption—and the costs per registered vehicle. The external costs per vehicle mile and per registered vehicle are measured using the ultimate external cost measure.

Willingness to Pay Costs* of Highway Crashes by Method of Analysis, United States, 1993

86

Paper No. 98-0067

TABLE 4


Breakdown of Willingness to Pay Costs of Highway Crashes Using Method 3

The results indicate that cars have the lowest external cost per vehicle mile, followed closely by light trucks. Motorcycles have nearly 10 times the cost. In terms of external costs per registered vehicle, cars still have the lowest costs; buses have the highest costs, followed by combination trucks. The high costs of buses and combination trucks reflect their exposure on the road.

CONCLUSIONS We have presented measures of crash costs by vehicle type for the United States using four different methods of allocating costs in crashes involving different vehicle types. The methods correspond to different perspectives toward attributing the costs of a crash to the heaviness of a vehicle. While the ranking of costs by vehicle type was not particularly sensitive to the allocation method, we found large differences in the percentage of total costs attributable to particular vehicle types. Consequently, differences in allocation methods can be important in determining expenditures for safety programs aimed at particular vehicle types. Comparisons between vehicle types were not particularly sensitive to whether costs were measured using a human capital approach or a willingness to pay approach. However, the magnitude of overall crash costs were generally 2.5 times greater using the theoretically more appropriate willingness to pay approach. We also measured external costs. Overall, external costs are much lower than total costs, largely due to the exclusion of occupant costs associated with injury and death. The role of motorcycles is reduced, and the role of heavier vehicles is increased, in relative terms. TABLE 5

We also examined costs of different vehicle types adjusted by two exposure methods—vehicle miles and number of registered vehicles. The high safety costs of traveling by motorcycle were apparent. In addition, the costs associated with travel by heavy trucks became more apparent. These results may be used to justify recovering safety costs to society through vehicle registration or other government fees. In particular, the expected external safety costs of different vehicle types may be internalized through these regulations. Current fees do not reflect these costs of traveling. Taxes based on actual road use would be more desirable, but the substantial difference in cost per vehicle mile between passenger cars, light trucks, and motorcycles makes it quite clear that a gasoline tax is a highly inequitable safety cost recovery mechanism. In light of how the allocation method affects the percentage of costs attributable to different vehicle types, it becomes important to choose carefully how costs are allocated in crashes involving different vehicle types. The methods that we studied varied from one that allocated costs strictly based on the costs directly incurred to the individual vehicle and its passengers to one where the costs were completely allocated to the heaviest vehicle. Since the lighter vehicle is likely to bear more of the injury and fatality costs in a crash with a heavier vehicle just by virtue of differential momentum, it seems reasonable to allocate at least some of those additional costs to the heavier vehicle in making decisions about public expenditures for crash avoidance programs. It is probably not reasonable for most purposes to allocate all costs to the heavier vehicle. Method 3 makes a reasonable compromise; costs are allocated to the heavier vehicle to the extent that the crash involved greater costs to the lighter vehicle than would occur in an average crash with the same vehicle type. There-

External Costs of Highway Crashes, United States, 1993

Miller et al.

fore, we suggest using this measure when the aim is to attribute crash costs based not only on the costs to the particular vehicle type, but also the costs to others. A reasonable and computationally simpler alternative is to allocate the costs of both vehicles to each crash (Method 2); however, this method tends to overweight multivehicle crashes and underassigns costs to heavy trucks and motorcycles. When the aim is to allocate costs based on the harm to the passengers (e.g., for crashworthiness research budget allocations), the method based on allocating costs to the individual vehicle of occurrence (Method 1) would be appropriate. In conclusion, the method of allocating costs in multivehicle crashes with different types of vehicles makes a large difference to the relative magnitude of costs for each vehicle type. Little research has been done on the topic. Further work might attempt to determine how costs are affected by allocating to the vehicle responsible for crashes.

ACKNOWLEDGMENT The research was supported in part by a grant from Nissan Research and Development and in part by a grant from the Maternal and Child Health Bureau, U.S. Department of Health and Human Services, and the National Highway Traffic Safety Administration.

REFERENCES 1. Vickrey, W. Accidents, Tort Law, Externalities, and Insurance: An Economist’s Critique. Journal of Law and Contemporary Problems, Summer, 1968, pp. 463–487. 2. Newberry, D. Road User Charges in Britain. Economic Journal, Vol. 98, 1988, pp. 161–176. 3. Janson, J. O. Accident Externality Charges. Journal of Transport Economics and Policy, Vol. 28, No. 1, 1994, pp. 31–43. 4. Perrson, U., and K. Odegaard. External Cost Estimates of Road Traffic Accidents: An International Comparison. Journal of Transport Economics and Policy, Vol. 29, No. 2, 1995, pp. 291–304. 5. Rice, D. P., E. J. MacKenzie, A. S. Jones, S. R. Kaufman, G. V. deLissovoy, W. Max, E. McLoughlin, T. R. Miller, L. S. Robertson, D. S. Salkever, and G. S. Smith. Cost of Injury in the U.S.: A Report to Congress. Institute for Health and Aging, University of California, San Francisco, and the Johns Hopkins University, Baltimore, 1989. 6. Accident Facts, 1990 Edition. National Safety Council, Chicago, 1991. 7. The Economic Cost to Society of Motor Vehicle Accidents. NHTSA, U.S. Department of Transportation, 1983. 8. Blincoe, L. J. The Economic Cost of Motor Vehicle Crashes, 1994. Report DOT HS 808 425. NHTSA, U.S. Department of Transportation, 1996. 9. Hartunian, N. S., C. N. Smart, and M. S. Thompson. The Incidence and Economic Costs of Major Health Impairments. Lexington Books, Lexington, Mass., 1981. 10. Evans, L. Traffic Safety and the Driver. Van Nostrand Reinhold, New York, 1991. 11. Traffic Safety Facts 1992. NHTSA, U.S. Department of Transportation, 1993.

Paper No. 98-0067

87

12. Miller, T. R., J. G. Viner, S. B. Rossman, N. M. Pindus, W. G. Gellert, J. B. Douglass, A. R. E. Dillingham, and G. C. Blomquist. The Costs of Highway Crashes. The Urban Institute, Washington, D.C. 1991. 13. Special Report 216: Zero Alcohol and Other Options. TRB, National Research Council, Washington, D.C., 1987. 14. Miller, T. R. Costs and Functional Consequences of U.S. Roadway Crashes. Accident Analysis and Prevention, Vol. 25, No. 5, 1993, pp. 593–607. 15. The Abbreviated Injury Scale 1985. Association for the Advancement of Automotive Medicine, Des Plaines, Ill., 1985. 16. Blincoe, L. J., and B. M. Faigin. The Economic Cost of Motor Vehicle Crashes, 1990. Report DOT HS 807 876. NHTSA, U.S. Department of Transportation, 1992. 17. Miller, T. R., and L. J. Blincoe. Incidence and Cost of Alcohol-Involved Crashes in the United States. Accident Analysis and Prevention, Vol. 26, 1994, pp. 583–592. 18. Miller, T. R., M. S. Galbraith, D. C. Lestina, T. Schlax, P. Mabery, R. Deering, D. Massie, and K. Campbell. Understanding the Harm From U.S. Motor Vehicle Crashes. Proc., 39th Annual Association for the Advancement of Automotive Medicine, Des Plaines, Ill. 1995. 19. Miller, T. R., L. C. Lestina, and R. S. Spicer. Highway Crash Costs in the United States by Victim Age, Driver Age, Restraint Use, and Blood Alcohol Level. Accident Analysis and Prevention, Vol. 30, No. 2, 1998, pp. 137–150. 20. National Accident Sampling System Crashworthiness Data System 1991–1993. NHTSA, U.S. Department of Transportation, 1995. 21. National Accident Sampling System 1986. NHTSA, U.S. Department of Transportation, 1987. 22. Fatal Accident Reporting System 1993. NHTSA, U.S. Department of Transportation, 1994. 23. Miller, T. R. Societal Costs of Transportation Crashes. In The Full Social Costs and Benefits of Transportation (D. Greene, D. Jones, and M. Delucchi, eds.). Springer-Verlag, Heidelberg, 1997, pp. 281–314. 24. Jones and Laughlin Steel Corp. v. Pfeifer. 103 Supreme Court Reporter. Washington, D.C., 1983, pp. 2541–2558. 25. Miller, T. R. Costs of Injuries to Employers: A NETS Compendium. Two volumes. NHTSA, U.S. Department of Transportation, 1993. 26. Kakalik, J. S., and N. M. Pace. Costs and Compensation Paid in Tort Litigation. Report R-33 91-ICJ. RAND, Santa Monica, Calif., 1986. 27. Hensler, D. R., M. S. Marquis, A. F. Abrahams, S. H. Berry, P. A. Ebener, E. G. Lewis, E. A. Ling, R. J. MacCoun, W. G. Manning, J. A. Rogowski, and M. E. Vaiana. Compensation for Accidental Injuries in the United States. Report R-3999-HHS/ICJ. RAND, Santa Monica, Calif., 1991. 28. Mishan, E. Cost Benefit Analysis, 4th ed. Unwin Hyman, Boston, 1988. 29. Jones-Lee, M. W. Editor’s Introduction. In The Value of Life and Safety. North-Holland Publishing, New York, 1982. 30. Scodari, P. F., and A. Fisher. How Uncle Sam Values Mortality Risk Reduction. In Highway Safety: At the Crossroads (Robert Stammer, ed.). American Society of Civil Engineers, New York, 1988, pp. 182–198. 31. Gramlich, E. M. A Guide to Cost-Benefit Analysis. Prentice Hall, Englewood Cliffs, N.J., 1990. 32. Jones-Lee, M. W. The Value of Transport Safety. Oxford Review of Economic Policy, Vol. 6, No. 2, 1990, pp. 39–58. 33. 1990 National Personal Transportation Survey. FHWA, U.S. Department of Transportation, 1993. All analyses, opinions, and conclusions are strictly those of the authors. Publication of this paper sponsored by Committee on Traffic Records and Accident Analysis.