Home » Status Report » 2006 » Article
Status Report, Vol. 41, No. 2 | February 25, 2006 Subscribe

Incomplete data distorts conclusions about frontal airbags

Are you safer in a vehicle with or without a frontal airbag? With the airbag, of course, despite a study by Mary Meyer and Tremika Finney of the University of Georgia. The main finding — that airbags cause more deaths than they prevent — is contradicted by years of published research establishing that airbags save lives.

It's true that deploying airbags have caused some deaths in low-speed collisions. But most of these occupants died in 1997 and earlier models, and the problem has been dramatically reduced. Only one death from an inflating airbag has been confirmed in 2001 and later models (see "Occupant deaths from inflating airbags have been all but eliminated," Aug. 6, 2005). In contrast, the National Highway Traffic Safety Administration estimates more than 10,000 lives have been saved by airbags since 1994.

"Meyer and Finney violated two basic tenets of scientific research, and this is what led to their erroneous finding," says Susan Ferguson, Institute senior vice president for research. "The first violation was using incomplete data without checking whether the remaining data were unbiased. Then when Meyer and Finney arrived at a conclusion that was out of line with a wealth of published research on the same subject, they didn't question their own work as scientists routinely do."

Using a weighted sample from the National Automotive Sampling System/Crashworthiness Data System (NASS) of more than 20,000 people involved in crashes, Meyer and Finney concluded that death rates were higher for occupants of vehicles with airbags than for those without. But an analysis by the Institute's chief statistician, Charles Farmer, indicates a fundamental flaw — eliminating 50 percent of the data because of missing information. Most of these records were excluded because the impact speeds of the crashes were unknown. However, speed was a major factor in Meyer and Finney's results. In fact, their principal explanation for their finding was that airbags were a problem in collisions occurring at lower speeds.

Farmer took advantage of additional information in NASS to add back more than 10 percent of the crash records Meyer and Finney had excluded. These data plus corrections of some misclassifications by Meyer and Finney dramatically shifted the results, even though Farmer used the same analytical techniques. Taking into account other factors such as crash type shifted the results more, finding airbags beneficial.

Even this larger sample was incomplete and probably biased. Farmer found that including still other variables in the analysis could shift the results back to finding airbags harmful.

"Meyer and Finney's work along with Farmer's analyses should serve as a warning to others about the limitations of using NASS data for some purposes," Ferguson says. "NASS is a carefully designed, nationally representative sample of crashes. We use these data all the time for details about injuries and vehicle damage sustained in crashes. But NASS isn't right for this study because the sample was grossly limited."

©1996-2016, Insurance Institute for Highway Safety, Highway Loss Data Institute | www.iihs.org