Rating the Aggressivity of Australian Passenger Vehicles towards Oth Er Vehicle Occu Pants and U N Protected Road Users

Since 1 990, Monash University Accident Research Centre has conducted a series of studies to provide consumer advice on the crashworthiness of individual makes and models of Australian passenger cars. Crashworthiness has been defined as the relative safety of a vehicle in preventing severe injury to its driver when involved in a crash. A concept complementary to crashworth iness is vehicle aggressivity. Aggressivity can be defined as the risk of injury which a vehicle poses to occupants of other vehicles which it impacts, and to unprotected road users such as pedestrians, bicyclists and motorcyclists. This paper describes the development of aggressivity ratings for Australian passenger vehicles. Two measures have been considered: 1. Aggressivity to occupants of other cars : This type of aggressivity rating is based on two-car crashes between passenger vehicles and measures the injury risk each make/model in the collisions poses to the drivers of the other vehicles. 2. Aggressivity to unprotected road users : The aggressivity ratings reflect the threat of severe injury to pedestrians, bicyclists and motorcyclists by the make/model of vehicle coll iding with them. The analysis was based on nearly 1 02,000 drivers involved in tow-away crashes with the makes/models which were the focus . of the study, and on nearly 22,000 injured pedestrians, bicyclists and motorcyclists. The aggressivity ratings are presented and discussed, along with their relationship to crashworth iness ratings for the same makes and models of vehicles. The results suggest that crashworthiness and aggressivity are two different aspects of a vehicle's safety performance, with good performance on one dimension not necessarily being associated with good performance on the other. Paper presented to the 1998 IRCOBI Conference on the Biomechanics of Impacts, September 16-18, 1998 Göteborg, Sweden IRCOBJ Co11fere11ce Göteborg, September 1998 27 SINGE 1 990, Monash University Accident Research Centre (MUARC) has conducted a series of studies to provide consumer advice on the crashworthiness of individual makes and models of Australian passenger cars. Crashworthiness was defined as the relative safety of vehicles in preventing severe injury to their occupants in crashes. Ratings of crashworthiness, measured by the rate of serious driver injury in tow-away crashes, were produced for ind ividual models (Cameron et al 1 992; Cameron et al 1 995; Cameron, Newstead and Skalova 1 996). This paper extends MUARC's previous work in this area to add measures of the "aggressivity" of individual car models when they crash. Aggressivity ratings measure the risk of injury which a vehicle poses to occupants of other vehicles which it impacts, and to other, unprotected road users such as pedestrians, bicyclists and motorcyclists. The addition of aggressivity ratings represents further consumer advice which purchasers of cars could take into account when choosing a specific model . DEVELOPMENTS IN EUROPE Broughton ( 1 994, 1 996) has defined an aggressivity index which is calculated for each car model (eg. M) from mass data on two-car crashes in which at least one driver is injured. The index was based on the same type of data used by Folksam lnsurance, Sweden (Gustafsson et al 1 989) and the U . K. Department of Transport ( 1 995) to calculate their respective crashworthiness indices (Table 1 ) . In both Sweden and the U.K. , non-injury crashes are not (fully) reported, so the number of two-car crashes in which neither driver is injured (ie. n4) is not known and hence cannot be used in either a crashworthiness or aggressivity index. Table 1 : Number of two-car crashes between specific make/model (M) and other makes/models (0) Drivers of other Drivers of make/model M makes/models 0 INJURED NOT INJURED INJURED n2 n3 NOT INJURED n1 n4 1 1 ---' Broughton's aggressivity index is: Broughton has also pointed out that the U . K. DoT crashworthiness index, and the Folksam crashworthiness index, 28 IRCOBI Conference Gii tehorg, September 1 998 are related through the relationsh ip D / A = R. Broughton shows that both D and R are influenced by the aggressivity of the specific model (ie. independent of their occupant protection capabilities, the models which inflict more injuries on "other'' drivers will have lower crashworthiness indices), but also that D comes closer than R to the ideal of being independent of the casualties in the "other'' car. Thus it appears that the Folksam index, R, is particularly likely to be a measure of not only the crashworth iness of a car model , but also its aggressiveness (the above relationship suggests that a more aggressive model will appear to be more crashworthy, if R is used as the measure of crashworthiness). Turning to Broughton's aggressivity index, A, and the DoT crashworthiness index, D, both Broughton and DoT have reported a strong inverse relationship between them when each is calculated for a range of car models included in the U.K. ratings. Relatively few models departed substantially from the curvi l inear relationship fitted to the data by Broughton. This suggests that either U.K. cars are truly characterised by an inverse relationship between aggressivity and crashworthiness, or that the apparent relationship is at least in part an artefact of the constraints on the data (ie. crashes with at least one injured driver) used to produce the indices for each model . lt is hoped to avoid these constraints by the use of Australian data on crashes based on the tow-away criterion for collection. More recently, an international working group on criteria for the safety assessment of cars has had searching discussions at two meetings in Germany during 1 995 and 1 996 (Langwieder and Fildes, 1 997). Whi le not specifically addressing the question of aggressivity ratings to date, they resolved at their second meeting that "rating criteria for „ . compatibility and partner protection should be subject to further consideration in future". DEVELOPMENTS IN THE UNITED STATES Hol lowell and Gabler ( 1 996) describe an NHTSA research program aimed at measu ring vehicle aggressivity and compatibil ity and then linking these to vehicle design characteristics. They calculated an "aggressivity metric", defined as Deaths in Other Vehicle Aggressivity Metric = (Total Registrations in Subject Vehicle) / 1 ,000,000 For each make and model of cars, light trucks and vans under 1 0,000 pounds, the metric was calculated from data on two-car coll isions involving a fatality, recorded in the Fatal Accident Reporting System (FARS) database for 1 99 1 -93. This metric showed that the most aggressive vehicles were light trucks (pickups and sports util ity vehicles) and vans and that, among cars, the heavier models were the most aggressive. However, the detailed results also . showed that weight is not always the over-iiding co"ntributor to aggressivity, as some heavy cars had relatively low scores on the metric. IRCOBJ Conference Göteborg, September 1998 29 Hollowell and Gabler recognised that a poor aggressivity rating may reflect characteristics of the d river and how the vehicle is driven (eg. speed) , as well as any structural or weight factor. At the very least, their metric is influenced by the crash rate per mil l ion registrations of each model, and variations in this rate should not be included in an aggressivity index aiming to measure the threat to other road users involved in a crash with the subject model. They also calcu lated two other metrics to at least partially overcome these problems: • the other vehicle fatalities divided by the subject vehicle fatalities, and • the other vehicle fatalities divided by the subject vehicle fatal accidents. These two metrics reduced the initial aggressivity ranking of a model which presumably had a relatively high crash rate. The first of these alternative metrics is similar to Broughton's aggressivity index, where "injury" is redefined as death. When describing future work, Hollowell and Gabler alluded to the lack of an accepted measure of aggressivity, even though they had presented three d ifferent metrics in their paper. They listed a number of proposed variations which will be evaluated in future, namely: 1 . Normalis ing by number of accidents instead of number of registrations 2. Normalis ing for the effect of restraint usage in either vehicle 3 . Normalising for accident severity 4. Examining the metric in prescribed accident modes, eg. frontal-side impacts or frontal-frontal impacts 5. Examining rollovers and full ejections from either vehicle 6. Limiting the other vehicle fatality count to cases where the subject vehicle was the striking vehicle. AGGRESSIVITY RATINGS FOR AUSTRALIAN PASSENGER VEHICLES This paper describes an investigation of the feasibility and methods of providing aggressivity ratings for Australian passenger vehicles in terms of the threat which each subject model represents to: 1 . Occupants of other cars col l iding with the subject model cars, and 2 . Pedestrians, bicyclists and motorcyclists impacted by the subject model cars. The aggressivity ratings were based on one of the data sets used to produce crashworthi ness ratings (Newstead, Cameron and Le 1 997) , namely Police reports of crashes in New South Wales (NSW) resulting in death or injury or a vehicle being towed away. Crashes involving pedestrians, bicyclists and motorcyclists are seldom reported to the Police in NSW unless someone is killed or injured (usually the unprotected road user). This means that an estimate of the risk of injury was not calculable for the unprotected road users for inclusion in the second type of aggressivity rating (a measure of inju ry severity was). This problE!m did not occur for drivers of other cars, for whom the · available data allowed estimates of both the risk of injury and of their injury severity. 30 IRCOBI Conference Göteborg. September 1998 AGGRESSIVITY TOWARDS OCCUPANTS OF OTHER CARS As in Europe and the United States, this type of aggressivity rating has been based on two-car crashes between light vehicles (ie. heavy vehicle collisions were excluded). The subject vehicles were the passenger cars, station wagons, four-wheel drive vehicles, passenger vans and light commercial vehicles manufactured during 1 982-95 whose makes and models have been identified in the NSW crash data. The NSW data on two-car crashes involving each model of the subject vehicle has been extracted in the same form as Table 1 . In this case, the number of crashes in which neither d river was injured (ie. n4) was available, at least so far as tow-away crashes are concerned. The measure of the risk of injury of the other drivers coll iding with the subject model, unadjusted for any other factors, was: lnjury risk of other drivers = RO = (n2 + n3) / (n1 + n2 + n3 + n4). The injury severity of other drivers could be measured in a number of ways from the information on injury recorded on NSW Police reports (viz. kil led; admitted to hospital; or injury requiring medical treatment) . The measure of injury severity, similar to that used in the crashworthiness ratings project, was: lnjury severity of other drivers = SO = proportion of injured drivers who were kil led or admitted to hospital. The aggressivity measure for each subject car model, assuming RO and SO are independent, was then calculated as: Aggressivity to other car drivers = AO = RO x SO. This measured the risk of a driver of other cars being kil led or admitted to hospital when involved in collisions with the subject model cars. Before this aggressivity measure was calculated, consideration was given to taking into account the d ifferences betjVeen the crash circumstances of th� subject car models which may result in a distorted view of its aggressivity only partly related to the characteristics of the subject cars. Factors available in the data to consider such d ifferences included: • speed l imit at the crash location • subject driver age (younger drivers may be driving at relatively fast speeds not fully represented by the speed l imit) • subject driver sex (male drivers may be driving at relatively fast speeds) • other car driver age (older drivers are more susceptible to injury) • other car driver sex (female drivers are more susceptible to injury, but males appear to be associated with relatively high injury severities). AGGRESSIVITY TOWARDS PEDESTRIANS, BICYCLISTS AND MOTOR CYCLISTS The aggressivity ratings reflecting the threat to pedestrians, bicyclists and motorcyclists have been based on NSW data on collisions between these road user types and the subject model vehicles with identified makes and JRCOBI Confere11ce Göteborg, September 1998 31 models . The aggressivity measure was based on injured pedestrians, bicyclists and motorcyclists and reflects the injury severity of their outcome. The measure of the aggressivity of the subject models towards unprotected road users was: Aggressivity to unprotected road users = AU = proportion of unprotected road users injured in collisions with subject car models who were killed or admitted to hospital. As with the measure of aggressivity to drivers of other cars, consideration was given to taking into account any major differences in the crash circumstances related to the following factors which may distort the results: • speed l imit at the crash location • subject driver age • subject driver sex • unprotected road user age • unprotected road user sex • type of unprotected road user.