Death Rates Vary Among Cars by More Than 700 Percent

Occupant deaths occur far more often in some cars than in others. Insurance Institute for Highway Safety researchers computed death rates for 1985-87 cars and found that among the 103 biggestselling vehicle series, the one with the lowest death rate is Volvo s four-door 740 760 a midsize specialty car, with 06 deaths for each 10,000 of these cars registered. The series with the highest death rate (5.2) is the Chevrolet Corvette a smali sports car. The Corvette's death rate is more than eight times as high as the Volvo s (see table on pages 4-5).

For each vehicle series, rates were computed from the total number of occupant deaths and registrations that occurred in the three calendar years, 198688. A vehicle series is a family of cars within a make that has the same wheelbase and a degree of commonality in construction such as body, chassis, etc.) Results for individual 1985-87 series are presented only for those with at least 150,000 registrations during 1986-88 (1988 models arent included because registrations are counted midyear when new cars are still being sold). Results for all cars and for each car size body style group were computed from death and registration counts for all vehicle series regardless of the numbers registered

Car size is an important influencing
See 700 Vanation, p. 2)

Many Cars Performing Better (or Worse) Than Predicted

Actual death rates for individual car series tell only part of the story (see Death Rates Vary Among Cars by More

Than 700 Percent) because they re influenced by the age and sex of the people who drive them, as well as by the size of the cars. So Insurance Institute for Highway Safety researchers computed for each of 103 popular car series a predicted as well as an actual death rate with the predicted rate taking into account 'he three important factors th encing the actual death rate The researchers developed a regression model relating actual death rales to car size, driver age and driver sex Because vehicle registration information cannot be broken down by age and sex, the re searchers' model uses information about the drivers in the fatal crashes

Specifically, the researchers estimated a nonlinear relationship between the death rates of individual vehicle senes and their wheelbases, proportion of occu pant deaths that occurred with drivers under 30 years old, and proportion of male drivers (A detailed explanation of the research methods is available from the Institute on request)

From this model, the researchers computed a predicted death rate for each car series using its actual wheelbase proportion of young drivers and proportion of male drivers Then the researchers com(See Better or Worse p 3)