Chevrolet Cobalt

The FARS analysis didn't take into account fatal accidents where the airbags weren't supposed to deploy.

Earlier today, we reported that the actual death toll attributable to GM's ignition switch problem had crested the 300 mark according to new research, well up from the original reports of 12 to 13 deaths. Now, word is breaking that the US government database that informed the study that the report was based on may have significantly overstated the correlation between the study and the GM recall.

The initial study was conducted by Friedman Research on behalf of the Center for Auto Safety, and used something called the US Fatality Analysis Reporting System. To recap, the study claimed that over a 10-year period, 303 people were killed in Chevrolet Cobalt and Saturn Ion coupes and sedans when their airbags failed to deploy. These undeployed airbags were then linked to GM's ignition switch recall, which as we've explained before, can turn the ignition out of the "run" position and into the "off" or "accessory" position, disabling the airbags in the process.

Now, according to a report from The Detroit News, which cites research from the Insurance Institute for Highway Safety and the National Study Center for Trauma and EMS at the University of Maryland, the FARS analysis didn't take into account fatal accidents in conditions where the airbags weren't supposed to deploy (which isn't to say crashes and deaths weren't caused by loss of control from the ignition switching off in the GM vehicles). According to the report, this was a significant number of the cases.

There is another potential problem, too. According to that same report, the National Highway Traffic Safety Administration uses both FARS and another database on fatalities, called the National Automotive Sampling System/Crashworthiness Data System (NASS/CDS). Where FARS uses what the DetNews calls "not always reliable" police data to record vehicular deaths within 30 days of a crash, NASS/CDS relies on what's known as a probability sample. It collects data on 5,000 crashes each year – including some found in the FARS database – to calculate a probability figure.

According to a 2009 IIHS study, "Among crashes common to both databases, NASS/CDS reported deployments for 45 percent of front occupant deaths for which FARS had coded nondeployments." In plain English, FARS doesn't provide a reliable count airbag deployments.

"The bottom line is at least for the years we looked at it, the coding of airbag deployments isn't always accurate," the study's author, IIHS Senior Vice President for Research Anne McCaratt told The News. "It's frustrating because it seems like police reports would be able to code accurately whether an airbag deploys or not. It's not like trying to figure out if the driver fell asleep."

At this point, it remains unclear how many deaths are attributable to the GM ignition switch flaw. It's possible that the number of total deaths may be far fewer than the 303 claimed by the Friedman Research/CAS study we covered earlier (although it may climb back past that number once deaths across all models are included), though it's likely still higher than the 12 to 13 cases thus far acknowledged by GM. With this latest turn, though, the issue becomes cloudier and cloudier.