GM's first missed sign of fatal flaw came in late 2003

DETROIT -- The first death linked to faulty ignition switches in millions of General Motors cars happened more than a decade before the automaker recalled them, according to the report on GM’s internal investigation.

The death toll from the switches was two by the time GM began building the Chevrolet Cobalt in the fall of 2004 and three when a top engineer on the Cobalt program decided in early 2005 not to improve ignition switches because of the cost and time required.

The crashes were among the earliest signs missed by GM in its failure to recognize that the ignition switch contained a fatal defect and to begin treating the matter as more than a mere inconvenience to customers.

The three people each died when 2004 Saturn Ion sedans equipped with the same switch that GM used for the Cobalt crashed head-on into roadside obstacles but the airbags failed to deploy. Several years later, GM concluded that at least one of the victims would have survived if her airbag had worked properly.

The first of the crashes happened in “late 2003,” the report says, at least eight months before any of those previously known. The government’s traffic-fatality database lists only one fatal crash in 2003 involving any Saturn vehicle from the 2004 model year, in which an Ion drove off a residential street outside Miami and into a building. The 19-year-old woman identified by Automotive News as the driver died after 17 days in a hospital, during which she accumulated nearly $300,000 in medical bills that went unpaid, according to public records.

Based on available data, the Miami incident fits the criteria that GM used to identify whether fatal crashes are linked to the defect: cars going off-road, where bumps or jarring terrain may have helped jostle the ignition out of the “run” position, followed by a frontal crash in which the airbags didn’t deploy despite indications that they probably should have.

But it’s unclear whether the Miami death is among the 13 that GM has linked to the recall. The only fatal Ion crash in 2003 that GM reported to regulators is listed as occurring in Connecticut, two weeks after the Miami crash. No such incident appears in either the federal Fatality Analysis Reporting System database or Connecticut’s traffic-crash database.

Lawyer warns of pattern

The following year, two people died in similar, single-vehicle, off-road crashes involving 2004 Saturn Ions. In July, an Ion jumped a curb in Visalia, Calif., hitting a utility pole and killing the driver. In November, another Ion drove off the side of a rural road in Van Zandt County, Texas, crashing into a tree and killing the front passenger while injuring the driver.

In those crashes, as in the Miami crash, the traffic-fatality database categorizes the point of impact as the center of the car’s front end and shows that each car was traveling at a speed experts say would normally trigger airbag deployment.

Both of the 2004 crashes perplexed GM engineers advising the legal team, but the company chose to settle the cases arising from them rather than act on the red flags they raised, the Valukas report shows.

After the family of the California victim, 37-year-old Shara Lynn Towne, sued GM in 2006, engineers noted that the car’s airbag sensing and diagnostic module did not record a crash, suggesting some type of power loss. A GM lawyer expressed concern that the damage to the car was “remarkably similar” to what happens during a certain GM crash test in which airbags are meant to deploy.

“Despite extensive analysis, the engineers have no solid technical explanation,” GM staff attorney Doug Brown wrote in October 2006. “The engineers agree 1) that the airbags … should have deployed … and 4) it is reasonably likely that deployment of the driver airbag would have prevented [Towne’s] death in the accident.”

Based on Brown’s case summary, a committee of GM lawyers known as the Roundtable agreed to settle. Though the exact amount is sealed in court files, Roundtable authorization allowed a settlement of $100,000 to $1.5 million, the report says.

Brown also handled two lawsuits related to the Texas crash. The GM engineer assigned to investigate it said he “had never seen a situation like this” and determined that the impact was “clearly severe enough to warrant deployment of the vehicle’s airbags,” according to a January 2008 evaluation prepared by an outside law firm hired to defend the case. But despite making that determination, another GM engineer asserted in a March 2008 court filing that the airbags wouldn’t have made a difference and that the car wasn’t defective.

GM agreed to settle the cases later that year for undisclosed amounts.

Engineering study

GM knew of at least one of the 2003 and 2004 deaths at the time it ended an internal inquiry into the ignition switches, but it doesn’t appear to have linked any of them to the ignition switch until much later. GM reported the November 2004 crash to the National Highway Traffic Safety Administration through the agency’s Early Warning Reporting system less than two months after it happened, citing an allegation that the airbags did not work properly.

At the time GM submitted that report, its engineers already were studying why the ignition switch in the Chevrolet Cobalt was so flimsy that it could be inadvertently turned off when bumped by the driver’s knee. But in March 2005, after GM informed dealers of the problem through a routine technical service bulletin, Gary Altman, the program engineering manager for the Cobalt, ordered the Problem Resolution Tracking System inquiry closed “with no action.” Altman determined that the lead time required, estimated cost and effectiveness of all proposed solutions meant they did not represent “an acceptable business case,” documents show.

Altman was among 15 employees dismissed by GM this week.

“The engineers who opened and worked on the PRTS report didn’t know of any accidents or fatalities that might have been attributable to problems with the ignition switch in the Ion or Cobalt, even though GM has now identified at least several crashes that might be attributable to the safety defect prior to the end of 2004,” Valukas wrote. Engineers interviewed by his team said that knowledge “would have changed their view as to whether the problem they were seeing in the Cobalt represented a safety issue.”

The Valukas report says GM opened its first legal file related to an airbag nondeployment claim in any of the recalled cars in January 2004. It says GM rejected the claim, which involved a 2004 Ion. Because NHTSA redacted victims’ names before publicly releasing the report Thursday, it’s unclear whether the claim related to the Miami crash or to the Connecticut death GM reported that quarter, which was categorized as airbag-related.

Extensive efforts to reach the family of the Miami crash victim were unsuccessful, and police there say they have purged the accident report because it happened so long ago.

It’s possible that the entry about the Connecticut crash, as posted on NHTSA’s Web site is an error. The July 2004 crash, which GM reported to NHTSA in mid-2006 after being sued, is listed as involving one death and two injuries, even though the driver was the only occupant.

'Basic design flaw'

At the same time that GM’s legal staff received its first clue about the airbag problem, an engineer driving an Ion flagged the car’s ignition as having a “basic design flaw” because the keys could hit the driver’s knee. A month after that, another engineer reported that he had accidentally turned the Ion off by bumping the ignition with his knee. But GM proceeded with plans to use the same switch in the Cobalt when production began that fall.

NHTSA’s acting administrator, David Friedman, has said it’s “likely” that the actual death toll from the defect is higher than 13. A Texas law firm representing relatives of four victims on GM’s list says it has evidence that the toll includes at least 60 deaths and nearly 300 injuries, 80 of which it describes as “catastrophic.”

But GM officials say they stand by their number, which was revised in March from 13 down to 12 and then back to 13 as the company discovered that one victim was counted twice and confirmed that a June 2013 death in Canada was connected. All of the other deaths linked to the faulty ignition switches occurred no later than 2009.