How NHTSA Missed the GM Ignition Switch Defect

See All 3 Photos3

On February 13, 2014, General Motors issued a voluntary recall of 619,122 Chevrolet Cobalts and Pontiac G5s to correct a defective ignition switch. Less than two weeks later, it recalled more than 1 million more cars. Owners, who had been submitting thousands of complaints for years, probably weren't too surprised. No one was likely more surprised, though, than the National Highway Transportation Safety Administration's Office of Defects Investigation. Until that day, the office responsible for investigating potentially dangerous or deadly automobile defects had no idea the defective ignition switch was killing people.

How did this happen? A scathing audit of the National Highway Transportation Safety Administration's (NHTSA) Office of Defects Investigation (ODI) by the Department of Transportation's Office of Inspector General identified numerous cultural, procedural, and analytical failures over more than a decade that prevented investigators from ever connecting the dots between airbag failures and the defective ignition switch.

More than anything, this is a story of numbers. GM eventually recalled more than 2.6 million vehicles, including 20 models from six brands built over 15 consecutive model years. During that time, GM had submitted more than 15,600 relevant reports to ODI regarding vehicles that would eventually be recalled; 2,000 additional GM reports covered cases in which at least one person was injured or killed. During the same time period, owners filed 9,266 relevant complaints with ODI about vehicles that would later be recalled, 72 of which caused at least one injury and at least three of which caused at least one death. In total, ODI had in its possession more than 26,800 reports related to the defective ignition switch but never made the connection between the ignition switch and airbags failing to deploy.

See All 3 Photos3

GM discovered that a weak ignition switch cylinder allowed the key to accidentally drop out of "run" and into the "acc" (accessory) position when bumped or jostled. In acc mode, the vehicle's electrical system is still on but the engine shuts off. More important, the airbags are disabled because a car would normally be stopped before turning off the engine. When the engine accidentally shut off while the vehicle was in motion, the loss of engine power, braking assistance, and power steering in many cases led to an accident. In many of those accidents, the airbags should've deployed but didn't because the key was in acc, not run. So far, more than 100 deaths have been confirmed to have been caused by this defect.

Read about how the NHTSA's recall office is failingRIGHT HERE.

How all this information went overlooked can be attributed to two major faults within ODI: how it collected the data and how it analyzed the data.

The first issue is a result of a vague reporting process and weak scrutiny by ODI staff. Automobile manufacturers are required by law to send quarterly reports to NHTSA of every warranty claim, property damage claim, consumer complaint, dealer and non-dealer field reports, recalls of foreign vehicles similar to U.S. models, and all technical bulletins, consumer advisories, and warranty communications. Manufacturers must also send reports on every incident that involves an injury or death, even if it wouldn't otherwise be covered by regulations.

NHTSA provides 24 codes covering major vehicle components and systems, and all applicable codes must be assigned to each report. However, NHTSA offers no guidance on how to apply these codes and tells anyone who asks to use their best judgment. As a result, different GM employees coding different reports about the same issue over the years used whichever codes made sense to them, leading to inconsistencies in the data.

See All 3 Photos3

What's more, NHTSA doesn't scrutinize these reports for accuracy or even completeness, so investigators had no way of knowing whether they were getting good information. NHTSA doesn't even check to make sure automakers have good internal reporting processes in place. As a result, errors and inconsistencies occur regularly, and automakers have even been known to intentionally miscode and miscategorize reports to avoid scrutiny.

Whether intentional or accidental, a 2006 GM report coded a fatal crash involving a 2005 Chevrolet Cobalt as "not involving any of the systems, components, or conditions defined in regulations," according to the Department of Transportation's audit. Supporting documentation, however, included a report from the Wisconsin State Patrol that identified the ignition switch and airbags as being likely culprits in the crash.

See All 3 Photos3

It's not just incident reports that were miscoded, either. In February 2007, GM sent NHTSA an October 2006 bulletin on the ignition switch defect that noted the key could easily be knocked out of "run" and into "acc" and warned this would cause a loss of electrical power. It failed to mention this would also shut off the engine and deactivate the airbags. The bulletin did note the problem was more likely to happen if the vehicle was turning, and for this reason GM used the "Steering" code rather than "Electrical" or other applicable codes.

This apparently wasn't the first bulletin GM had put out regarding the ignition switch. In June 2005, a consumer sent NHTSA a copy of a letter she'd sent to GM in which she claimed her Chevrolet dealer gave her a GM bulletin on the problem. This alleged bulletin was not found in NHTSA's files.

GM wasn't alone in its inability to get the codes right. Even people working for ODI made mistakes. In September 2003, an owner complaint correctly identifying the ignition switch defect was miscategorized by ODI contractors who used the codes "Unknown or Other" and "Exterior Lighting: Headlights: Switch" rather than the appropriate "Electrical Systems: Ignition: Switch." The mistake happened again in June 2005 with another owner complaint.

Of course, the owners can hardly be blamed here. NHTSA's phone and online defect reporting tools require owners to choose up to three vehicle systems or components from a menu but offer no guidance, assistance, or even a glossary of terms to help them along. As such, many consumer complaints end up being of little value, as they incorrectly identify systems and components or don't provide enough detail. The systems also don't allow owners to upload supporting documents or instruct owners on which documents to keep.

See All 3 Photos3

Those who got through this system couldn't be certain their complaints would be read, though. NHTSA receives between 40,000 and 80,000 consumer complaints per year, but it only has one staff member to screen them, and he's only allowed to spend half his day on screening. That means he screens on average 330 complaints in half a day. He estimates only 10 percent of complaints make it to a second round of screening, where eight advanced screeners are overworked and in some cases underqualified to investigate the problem.

Despite all this, some owners get through only to have their complaints go unanswered. From the report: "In June 2005, a consumer sent NHTSA a copy of a letter (attached to an ODI Vehicle Owner Questionnaire) that she sent to the GM customer service department describing how her 2005 Chevrolet Cobalt had turned off on three occasions while driving. The letter stated that the service manager tested the vehicle and was able to turn the ignition switch when his knee hit the bottom of the "opener gadget" on the keychain."

From the letter: "This is a safety/recall issue if there ever was one. Forget the bulletin. I have found the cause of the problem. Not suggested causes as listed in the bulletin. The problem is the ignition turn switch is poorly installed. Even with the slightest touch, the car will shut off while in motion. I don't have to list to you the safety problems that may happen, besides an accident or death …"

Then there was the formatting issue. Although ODI guidelines specify which file formats are acceptable when submitting a report, GM updated to the latest version of Microsoft Word in 2013 and began submitting reports with the .docx file extension, which NHTSA's software couldn't read. The error wasn't discovered until nearly a year later, and in that time, the software failed to process 13 relevant reports. ODI reviewed one report separately.

See All 3 Photos3

Even when ODI had good information in hand, missteps prevented investigators from making the critical connection. In the Wisconsin example above, NHTSA initiated a Special Crash Investigation that listed loss of power as a potential cause of the crash. The report was submitted to ODI, but ODI staff told auditors they never read it.

In many cases, though, we don't actually know what ODI did with the information in its possession, because until 2013, advanced screeners weren't required to document their reasoning when declining to pursue a complaint. Since 2013, we know 926 consumer complaints were received; only 27 made it to advanced screening, or 3 percent, and 11 of those 27 were declined with justifications such as "minimal hazard" and "no actionable trend indicated." This despite the fact that these cases include airbags failing to deploy in a crash.

Over the course of 11 years between the first report ODI received and the recall, ODI staff made only three proposals to the division chiefs to open a full investigation into Chevrolet Cobalts and Saturn Ions (the two cars that represented the vast majority of ignition switch defects). Those proposals, however, were to investigate power steering loss and/or airbags failing to deploy. None identified the ignition switch as the root cause even though one case included in one of the proposals had an owner's testimony that authorities had found the key in the acc position.

One of those proposals was sent in 2007 to the Defects Assessment Panel, which reviews proposals on the fence. The panel, described by staff as a "dog and pony show" and a waste of time, declined to open an investigation due to an apparent misunderstanding about when airbags should and shouldn't deploy, as well as a statistical analysis which showed Cobalt and Ion airbags failing at roughly the same rate as other vehicles in the segment. One proactive division chief assigned an advanced screener to monitor the issue in case it flared up again, but when that person left NHTSA the next year, no one was assigned to replace him as monitor.

Another opportunity was missed in 2009. A division chief ordered a special investigation into the crash of a 2005 Cobalt whose airbags didn't deploy, but ODI never followed up on the results of the investigation. ODI has no explanation for this. Worse, when two ODI staff members reviewed the investigation a year later, neither reported their findings. One said he wasn't responsible for screening safety issues. The other didn't recall reviewing the investigation but said he would have only reported on issues he specializes in. Those were engine, powertrain, and speed control.

Another miss occurred in 2010. A screener suggested revisiting the 2007 investigation on airbag non-deployments because new complaints were coming in. The screener dropped his suggestion after the airbag investigator produced an analysis that showed a downward trend in complaints.

In 2011, the Cobalt was included in an ODI analysis of 22 vehicles with known airbag issues. The 2005-2010 model year Cobalts ranked second in injuries and fourth in deaths.

Although ODI for years successfully identified airbag non-deployment as a serious safety issue, the office never identified a root cause and never linked it to the ignition switch. In fact, despite documented complaints specifically about the ignition switch, ODI staff told auditors "there were no discussions of the ignition switch defect prior to the February 2014 recall."

See All 3 Photos3

When asked how they missed the connection, ODI officials told auditors "they did not understand the safety consequences of the ignition switch defect before the GM recall." The report does not elaborate on how that could be the case, but one possible reason is insufficient training. Three screeners assigned to analyze data from airbag incidents said they "lacked training in airbags." Another screener was assigned to airbags in 2008 despite having no airbag training, no engineering background, and no automotive background; the screener was originally hired to review child seat restraint issues. In general, the audit suggests the investigations were often stymied by misunderstandings about when and under what circumstances an airbag should deploy, misunderstandings that were never corrected.

The final tally of missed opportunities is damning. NHTSA received its first report of a Saturn ION shutting off while driving and locking the steering column in May 2003 from a consumer. In September of the same year, NHTSA received its first report identifying the ignition switch defect as the cause of an engine shutting off while driving, which also came from a consumer complaint. NHTSA would go on to receive seven more reports from consumers correctly identifying the ignition switch defect over the years.

By January 2004, it had received its first report of an Ion's airbags failing to deploy in a crash. This, too, came from a consumer. The first report provided by GM did not come until March 2005. NHTSA did not launch any investigation of airbag non-deployment crashes involving Ions and Cobalts until August 2005. That investigation was the first to note the ignition switch was in the acc position. Three additional NHTSA investigations over the years would also note the ignition switch being in the acc position. Two of those investigations linked the ignition switch position to the airbag non-deployment, but ODI never acted on that information.

ODI continued reviewing airbag non-deployments in Cobalts and Ions through August of 2013 without ever making the connection or taking action on either the airbag non-deployments or the ignition switch defect. GM, to its credit, identified the problem on its own and initiated a recall in February 2014, which caught NHTSA by surprise.

Want to learn more??? Read about how the NHTSA's recall office is failingRIGHT HERE.