Validating analytical judgments: the case of the airbagʹs lifesaving effectiveness

Based primarily on experimental testing and engineering judgment, early evaluations of airbags made variable predictions of their lifesaving benefits. The official estimates of the US government from 1977 to 1987 were that between 6000 and 9000 lives could be saved each year if all passenger cars were equipped with full-front airbags. Now, over a decade later, extensive real-world crash experience in the US has been analyzed to validate early predictions. Lifesaving estimates have been revised downward and the official US government position in 1997 was that, overall, approximately 3000 lives will be saved each year when the vehicle fleet is fully-equipped with frontal airbags. This article explores the reasons for poor validity of early lifesaving forecasts for airbags. We found that airbag effectiveness for unbelted adult occupants was overestimated, the number of adult motorists who would wear safety belts was underestimated, and that early analyses generally did not adequately address heterogeneity in the vehicle fleet or people. For each source of error, we discuss why technical specialists may have erred and what lessons can be gleaned from this case study in validation.