The agency that finds vehicle defects destroys its own case files

Getting a dangerous vehicle defect off the road runs through two very different NHTSA processes: first, investigators have to find the defect, often by studying individual crashes; then, once a recall is issued, owners have to actually bring their cars in for repair.▸ Two recent oversight reports examined each end separately. DOT's inspector general⧉ found the Special Crash Investigations Program -- NHTSA's most detailed crash-investigation tool -- took an average of 25 months to publish a report against its own 3-to-6-month goal, then destroyed the case files afterward in violation of federal records rules. GAO⧉ found that once a recall is issued, NHTSA hasn't updated its research on what keeps owners from completing repairs since 2019, even though a typical recall only reaches 69% of vehicles against the agency's stated goal of 100%.

NHTSA's Special Crash Investigations Program has produced the agency's most detailed crash-investigation data since 1972, feeding directly into decisions about vehicle recalls and safety-system requirements.▸ To assess it, DOT's inspector general drew a random sample of 28 investigations. In 25 of the 28 (89%), NHTSA could produce some documentation of why the crash was selected for investigation. But of those 25, the office couldn't verify that 11 had actually been approved by the contracting officer responsible for assigning them -- meaning NHTSA couldn't show a case had been properly authorized. For the remaining 3 of 28, no documentation existed at all: not the original request, not an assignment email, nothing.

Add it up and half the sample -- 14 of 28 cases -- lacked verifiable proof that NHTSA properly selected and approved the investigation it was running.▸ NHTSA's own defense was that the program has "almost always proceeded to investigate the crashes" it's assigned, so it didn't see the missing paperwork as a real problem. The inspector general's response: without documentation, there's no way to confirm investigations align with the program's mission, or that new staff will select and approve cases consistently once the people who currently just remember the process move on.

NHTSA's own procedures manual sets a goal of 3 to 6 months to complete a final case report after a crash investigation is assigned.▸ Across the 28 published reports in the inspector general's sample, the actual average was 25 months from the start of the investigation to the report's publication -- more than five times the midpoint of the agency's own goal. Some delays trace to real obstacles outside NHTSA's control, like vehicles held by law enforcement or slow-arriving medical records. But the inspector general also found NHTSA had no process to track overall timeliness, flag investigations that were falling behind, or assess whether its 3-to-6-month goal was even realistic given how complex some crash investigations are.

Then, once a report was finally published, NHTSA destroyed the underlying case file.▸ Its own procedures manual told investigators to delete the source documents -- initial contact summaries, preliminary case summaries -- once a report went out, and for all 28 sampled investigations, those documents were gone. The inspector general found this violated National Archives and Records Administration rules: since NHTSA had never actually set a retention schedule for those specific documents, they were legally required to be kept as permanent federal records until a formal disposition was approved. NHTSA said it would stop the practice immediately.

The program's investigations have driven real safety rules -- its work on frontal air bag injuries fed into 2003-model-year air bag redesign requirements, its rollover-crash findings led to a rule requiring inflatable side curtains, and its research on backover crashes helped drive the 2014 rule requiring rear-visibility technology on new vehicles.▸ But NHTSA had no formal process to document which investigations actually drove which safety improvements -- meaning that beyond a handful of officials' institutional memory, the program can't systematically show its own impact.

About 22 million vehicles were recalled for safety defects in 2022 alone.▸ NHTSA's stated goal is a 100% completion rate on every recall -- every defect found, every car fixed. As of a 2021 report to Congress, a typical passenger vehicle recalled in 2018 had only a 69% chance of being repaired 15 months after the fix became available.

The best evidence of what's possible with an all-out effort is the Takata air bag recall -- 67 million defective inflators across tens of millions of vehicles from 19 manufacturers, tied to more than 400 injuries and 27 deaths in the U.S. Manufacturers, under NHTSA's direction, ran monthly outreach in multiple languages, offered loaner cars and mobile repair, and leaned hard on urgency messaging.▸ That campaign reached an estimated 80% completion rate -- the best NHTSA has ever achieved on a mass recall, and still 20 percentage points short of its own goal.

GAO found NHTSA hasn't updated its research into why owners skip repairs since 2019, and has no plans to.▸ Fourteen of the 27 outside entities GAO interviewed said the agency should be doing more of that research, particularly since new factors -- like the growing share of recalls fixable with an over-the-air software update rather than a dealership visit -- keep changing the picture. NHTSA's Recall Management Division, officials told GAO, is built for enforcement, not research.

NHTSA also hasn't systematically captured what it learned from the Takata campaign.▸ It analyzes Takata-related reports closely but, by its own account, doesn't apply that same scrutiny to other manufacturers' recall efforts. Six of the eight manufacturers GAO interviewed said NHTSA has never shared lessons learned from its collaborations with them, and four called the agency's public "Tips" webpage -- its main public documentation of Takata lessons -- outdated. One manufacturer told GAO that collaborating with third parties like independent repair shops and state DMVs can generate a 5 to 40 percent increase in its completion rate; without a real lessons-learned process, NHTSA has no systematic way to spread that kind of finding to the rest of the industry.