This Letter to the Editor appears in the Dec. 7 print edition of Transport Topics. Click here to subscribe today.
Nonpreventable Crashes and Predictability
I disagree with the conclusion that removing nonpreventable crashes will result in significant changes in carriers’ CSA Crash Indicator BASIC percentile scores (11-16, ttnews.com).
The details of the American Transportation Research Institute report, “Assessing the Impact of Non-Preventable Crashes on CSA Scores,” actually shows that the relative percentiles which define whether a carrier exceeded the intervention threshold would remain largely unchanged.
The “Changes in Rank” table in the ATRI report shows that the logical statistical conclusion occurred, five carriers improve their rank, five carriers worsened and four experienced no change. At best, the most a carrier rose in rank among the 15 carriers examined was two places, at worst one carrier dropped three places. In the end, the carriers at the bottom of the class when considering all crashes nevertheless remained at the bottom of their class even after the non-preventable crashes were removed.
More importantly, these changes in rank were so insignificant that the bottom five carriers, who would be considered on “Alert” at the 65% threshold level in the Crash BASIC, did not change their position enough to move out of “Alert” status. Thus, while specific, individual scores of some companies may change, the general standing of those companies in comparison with their peers did not. Overall, the relative percentiles which define whether a carrier exceeded the intervention threshold would largely remain unchanged.
Your story incorrectly asserted that “ATRI estimated that not including the nonpreventable crashes in the 15 carriers’ safety measurement scores could save them a total of $68 million.” This number is the estimated total financial cost of crashes that ATRI considered nonpreventable, however, this does not mean that the cost would not still be borne by the industry and society. These crashes still would occur regardless of the assignment of preventability, and the vehicle and property damage costs, as well as worker’s compensation, still would be paid. Yet, the establishment of an ad hoc system for determining whether crashes were nonpreventable will itself cost millions of dollars, according to a January 2015 FMCSA study, and will not make any appreciable change in Crash BASIC scores, economic savings or, most importantly, real-world safety.
Moreover, the present version of the Crash Indicator, which includes all crashes, even nonpreventable ones, has consistently been shown to be one of the strongest indicators of future crash risk. A 2014 evaluation by the John A. Volpe National Transportation Systems Center provided significant support for the Crash Indicator as an accurate gauge of high future crash risk, concluding that the results indicated a consistent association between the Crash Indicator percentiles and future crash rates across carrier types.
In fact, ATRI determined in a 2006 study that involvement in a past truck crash increased the likelihood of a future crash by 87% and confirmed this finding in 2011, when another ATRI study found that involvement in a past truck crash, regardless of preventability, increased the likelihood of a future crash by 88%. In both cases, these findings were found to be statistically significant.
In sum, removing nonpreventable crashes from the CSA database, is a tempest in a teapot.
Shaun Kildare, Ph.D.
Director of Research
Advocates for Highway and Auto Safety