NHTSA revising how it analyzes crash data

David Shepardson
Detroit News Washington Bureau

Washington — The National Highway Traffic Safety Administration is revising how it analyzes U.S. vehicle crash data — and changing where it samples crashes from.

The Government Accountability Office in a report Friday disclosed that NHTSA has sharply reduced the number of crashes it reviews annually as part of the National Automotive Sampling System Crashworthiness Data System, or NASS-CDS, which is a nationally representative sample of police-reported motor-vehicle traffic crashes.

On average, since 1988 NHTSA has reviewed about 4,700 crashes as part of the effort, but because of a flat budget the number of crashes reviewed in 2013 fell to about, 3,400. The number has steadily decreased since 2009. GAO said NHTSA blamed it on rising costs. “For example, funding for NASS-CDS has been flat-lined since 2010, whereas costs — including costs for labor, information technology, leases and fuel — have risen,” GAO said.

NHTSA is in the process of redesigning the NASS-CDS sample by reviewing the data elements and the statistical methodology behind selecting the sample for a new survey called the Crash Investigation Sampling System.

The current areas sampled include three places in Michigan: Washtenaw, Genessee and Muskegon counties, out of about two dozen areas, including Chicago, Los Angeles and Knox County, Tenn. The new sample areas include Comal and Tarrant counties in Texas, Maricopa County in Arizona and Hamilton County, Ohio. NHTSA reduced the areas included in the sample from 36 to 24 in 1991. It wants to add more areas if the budget allows going forward.

GAO found NHTSA’s approach to redesigning the sample “reasonable.” But it criticized NHTSA for not meeting congressional deadlines. A 2012 law required NHTSA to provide Congress with information on the types of analysis that can be conducted and conclusions drawn under the current sample size and an expanded sample size, and the number of investigations that NHTSA should conduct as part of the sample that would allow for optimal data analysis, and said the report had to be delivered by Oct. 1, 2013.

NHTSA missed the deadline and only as GAO was completing its review in January did it make the disclosures to Congress.

One issue is the study requires a lot of travel by investigators. In 2012, the White House directed federal agencies to spend at least 30 percent less on travel expenses compared with 2010 “and to maintain that level of spending through 2016.”

This could delay revisions because it “requires that NHTSA staff travel to train the new crash technicians as well as to gain the cooperation of police jurisdictions, tow yards and others,” GAO said. “NHTSA officials said they are currently working to obtain relief from this cap and hope to start implementing the new” sample this year.

Congress awarded $25 million for the Data Modernization Project in 2011 and has since given NHTSA another $3.5 million. NHTSA still has about $12 million available.

A bigger sample size could allow for better data on some less commonplace crashes like side impact crashes involving the death of a child under 15.

The budget for the study has remained at $12.5 million since 2010 “and NHTSA officials also told us that the future budget for CISS remains uncertain,” GAO said. The new sample is designed so 10 percent of the police accident reports selected for investigations will contain a newer vehicle and an serious injury, up from 6.9 percent in the current system.