The early warnings that GM missed
Drivers got blame before switch was quietly changed
- First signs of GM ignition switch flaw emerged in 2001
- GM offers $500 discount, free loaners to owners of recalled cars
- Justice Dept. exploring GM's handling of fatal switch recall
- House panel to probe GM recall of faulty ignition switches
- U.S. House committee to probe GM recall
- For GM, a need to show sense of urgency
- Bankruptcy terms limit GM liability in crash lawsuits
- GM uses social media to respond to customer gripes
DETROIT — Ten years ago, two General Motors engineers reported that the ignition switch on the Saturn Ion was so flimsy and so low on the steering column that the driver’s knee could easily bump the key and turn off the car.
"This is a basic design flaw and should be corrected if we want repeat sales," one of the engineers wrote in January 2004 as part of GM's Company Vehicle Evaluation Program.
But not only did GM fail to correct the flaw; it installed the same switch on the Chevrolet Cobalt that the company introduced that year, proclaiming a new era of higher-quality small cars. It kept using the switch until 2006, when it was quietly redesigned.
For years, as drivers complained that their Cobalts and Ions were stalling repeatedly, GM treated it as a matter of customer satisfaction, not safety. Documents show that the company either didn't grasp the significance of the problem or didn't consider it worthy of resources.
A 2005 entry in GM's complaint-tracking system blamed drivers' bad habits: "They hit the ignition key slot."
Proposals considered before the Cobalt went on sale were vetoed "after consideration of the lead time required, cost, and effectiveness of these solutions," GM said in a detailed timeline filed Feb. 24 with federal regulators.
The decade-long lead-up to a recall, during which at least 13 people died in crashes linked to the defect, appears, in the best case, to be a tragic byproduct of GM's famously siloed culture, which stymied the sort of interdepartmental communication and data sharing that could have prompted quicker action. Compounding the chances for a serious defect to go unresolved for so long was the fact that GM eliminated one-third of its white-collar work force from 2005 through 2009.
'Acted without hesitation'
Mary Barra, who became GM's CEO in mid-January after three years as head of product development, said she learned of the problem just before the first stage of the recall was issued last month.
Barra: Learned about problem just before recall
"When this was brought to my team a few weeks ago, we acted without hesitation to go well beyond the decision by the technical experts," Barra wrote in a message to GM employees last week.
GM announced an initial recall on Feb. 13 then expanded it Feb. 25. Overall, GM is recalling 1.6 million vehicles worldwide, including 1.4 million in the United States. The models are the 2003-07 Ion, 2005-07 Cobalt, 2006-07 Chevrolet HHR, 2006-07 Pontiac Solstice, 2007 Saturn Sky, 2007 Pontiac G5, 2005-06 Pontiac Pursuit (Canada only) and 2007 Opel GT (Europe only).
The National Highway Traffic Safety Administration, which is investigating whether GM acted quickly enough, has given the company until April 3 to answer 107 questions about its handling of the recall. GM says it is cooperating with the investigation and has hired an outside law firm to conduct an internal review.
Before issuing the recall, GM communicated little with customers about the issue unless they complained.
At least 12 times, GM bought back Cobalts from customers who reported frequent incidents of stalling that dealers could not fix, according to a June 2013 deposition of Victor Hakim, a senior manager in GM's Field Performance Assessment department. The company sent a service bulletin to dealers in 2005 and developed a small plastic key insert designed to reduce the chances that a heavy dangling key ring could inadvertently turn off the engine. The key inserts were made available to customers who complained. GM said its warranty claims data show that only 474 customers received one.
In other cases, GM and its dealers essentially told customers that their expectations of a $13,000 car were too high.
"There is nothing mechanically wrong with the vehicle," reads an entry in GM's complaint-tracking system following up on an October 2005 complaint from a Cobalt owner in Gettysburg, Pa., according to the deposition of Hakim. "It is the customer's driving habits. They hit the ignition key slot."
Airbags won't deploy
The Cobalt recall is linked to reports of 23 frontal-impact crashes in which airbags failed to deploy. Six of the crashes -- two of which involved drunken driving -- resulted in the deaths of eight front-seat occupants, at least three of whom were age 16 or younger. At least one crash also resulted in the death of a back-seat passenger.
The most recent fatality that GM has linked to the Cobalt recall occurred in December 2009. GM doesn't consider the 2010 death of 29-year-old Brooke Melton to be connected to the recall because hers was not a frontal-impact crash, though black-box data suggest the ignition switch in her 2005 Cobalt stopped airbags from deploying.
Five additional deaths are linked to crashes involving the Ion, which GM added to its recall on Feb. 25. GM expanded the recall six days after Lance Cooper, the Georgia lawyer who represented Melton's estate in a lawsuit against GM, asked NHTSA to investigate GM's handling of the recall and pointed out that more vehicles used the same ignition as the Cobalt.
GM said its employees learned of many of the Cobalt crashes now linked to the recall within a month of each one happening. But that information didn't always spread within the company, as shown by the handling of a July 2005 crash in Maryland that resulted in the first fatality now linked to the recall.
NHTSA's Special Crash Investigations unit began examining the Maryland crash in mid-August 2005, and GM said its legal staff opened a file related to the crash in September 2005.
A 16-year-old girl, Amber Marie Rose, died when the 2005 Cobalt she was driving hit a tree at the end of a residential cul-de-sac. Rose was drunk, not wearing a seat belt and driving nearly three times the 25 mph speed limit, according to the NHTSA report on the crash.
The airbags did not deploy, and black-box data showed that the ignition was in "accessory" mode, instead of "on," when the crash occurred. Her mother, in an interview last month with radio station WTOP in Washington, said paramedics told her at the time "if the airbags had gone off, [Rose] would have been alive today; she would have been injured, but she would have been alive."
NHTSA issued its report on the crash in February 2006. When NHTSA brought up its findings during a March 2007 meeting with GM about airbags, the GM employees present were unaware of the crash.
At that point, GM assigned an engineer to track frontal-impact crashes involving Cobalts in which the airbags failed, and it had discovered 10 such incidents by the end of that year. It learned that the ignition had been turned from "on" to "accessory" mode in four of the 10 crashes.
New key in 2009
GM's timeline shows no further action on the issue until February 2009, when it opened an internal inquiry that resulted in a redesign of the ignition key for the 2010 Cobalt, the car's last model year. Engineers determined that the new key design -- a change the company had approved four years earlier, then canceled -- would "significantly reduce downward force and the likelihood of this occurrence."
By the time GM filed for bankruptcy protection on June 1, 2009, it had collected black-box data from 14 frontal-impact Cobalt crashes showing that the ignition was in "accessory" mode when seven of them occurred. Of the 23 crashes that GM now links to the recall, it has data showing that the ignition was in "run" when just nine of them occurred, in "accessory" for 12 of them, and off for one.
GM has not released details about the crashes it knows about, so it's unclear whether the numbers include the crash involving Chandra Smallwood, who was trying to beat an approaching hailstorm as she headed home from work on the beltway around Fort Worth, Texas, in April 2008. She said Interstate 820 was still clear and dry when she lost control of her 2007 Cobalt and slammed into a concrete median at about 65 mph. She suffered bruises and back injuries, and her brother, in the front passenger seat, was knocked unconscious and taken to a hospital.
"It smashed up the whole front end of the car, all the way to the tires, but the airbags never deployed," Smallwood, 28, told Automotive News last week. "The insurance agency said there's no way it couldn't be a defect."
But the dealership at which she had bought the Cobalt a year earlier told her there was no recall for her car. Smallwood said she tried contacting GM directly but never received a reply.
Secret design change
It appears the recall would have included far more than 1.6 million vehicles if not for a design change in 2006 that almost no one at GM knew about until last October.
GM said its design engineer responsible for the Cobalt ignition switch, which was supplied by Delphi Mechatronics, signed a document on April 26, 2006, approving a new detent plunger and spring that created more torque, making it harder to switch the key position.
But because the part number did not change, GM for years was unable to figure out why the reports it was getting of crashes without airbag deployment involved vehicles only from 2007 or earlier. (Delphi itself was in bankruptcy proceedings from 2005 through 2009 and cutting thousands of jobs.)
Charlie Miller, the owner of a Mississippi auto repair shop who was hired by the Melton family's lawyer, discovered the design change while performing a mechanical analysis of the Cobalt involved in the 2010 crash. A new switch Miller bought from GM had almost double the torque as the one in Melton's Cobalt, he said in an interview.
A 2012 GM internal review failed to find an explanation for the low torque in 2007 and earlier Cobalts, but an outside expert hired by GM in April 2013 found the same differences Miller had noticed. On Oct. 29, about a month after Melton's family reached a confidential settlement with GM, Delphi provided GM with records documenting the 2006 design change.
GM subsequently tested and analyzed the old switch, and the matter reached the team of senior executives who decide on recalls in December. The team, known as the Executive Field Action Decision Committee, requested more analysis, then approved a recall on Jan. 31, 2014.
GM said it would begin sending recall letters to customers today, March 10.
Until the affected vehicles have their ignition switches replaced, GM is urging owners to remove everything from their key rings but insists the vehicles are safe to drive.
"People have been driving them all along," GM spokesman Alan Adler said last week. "There should be no issues with driving the vehicles."
GM had almost the same message in 2005 when several journalists test driving the Cobalt reported experiencing the car stalling. A service bulletin GM sent to dealerships in 2005 said owners "should be advised of this potential and should take steps to prevent it -- such as removing unessential items from their key chain."
"When this happens, the Cobalt is still controllable," Adler told New York Times reviewer Jeff Sabatini for a story published June 19, 2005, explaining why GM did not consider the problem a safety issue.
"The engine can be restarted after shifting to neutral. Ignition systems are designed to have on and off positions, and practically any vehicle can have power to a running engine cut off by inadvertently bumping the ignition from the run to accessory or off position."
Gary Heller, writing in The Daily Item in Sunbury, Pa., said the engine on the Cobalt he tested shut down four times in a week.
"I never encountered anything like this in 37 years of driving," Heller wrote. "I hope I never do again."
GM cited both articles in its timeline filed with NHTSA, as well as a June 26, 2005, column in the Cleveland Plain Dealer skewering GM for trying "to pretend that turning off the engine by mistake isn't a safety issue."
You can reach Nick Bunkley at firstname.lastname@example.org. -- Follow Nick on