Predicting the used car safety ratings crashworthiness rating from Australasian New Car Assessment Program ANCAP scores.

The aim of this study was to investigate whether the correlation between vehicle secondary safety ratings developed from crash tests and those developed using real world crash data could be improved. Logistic regression was used to re-weight ANCAP test scores to improve their relationship with Used Car Safety Rating (UCSR) crashworthiness scores. TAC claims data were then used to investigate whether the re-weighted ANCAP scores reflected relationships that would be expected from existing knowledge of real world crashes and ANCAP crash tests. Sixty-nine vehicle models were included in the analysis. Analysis showed that the correlation between ANCAP scores and the UCSR crashworthiness estimates could be improved greatly by weighting the component ANCAP measures differently to the current ANCAP summary measure. Two broadly different approaches were utilised to achieve this. The first re-weighted the current summary scores from the offset frontal and side impact test, the second re-weighted each of the individual body region scores from each test. The second method produced slightly higher correlations with the UCSR crashworthiness measures however the first method had better face validity when validated against real world crash and injury distributions. Consequently, a new ANCAP summary measure based on the first method would be preferred for adoption in practice. Including bonus points given for performance in the optional side impact pole test and the presence of seatbelt reminder systems improved the relationship with the UCSR crashworthiness measure in all instances. Including vehicle mass in the new ANCAP summary scores investigated resulted in the highest correlations with the UCSR crashworthiness measure. However including mass effects in an ANCAP summary measure is problematic from the perspectives of the objectives of ANCAP and would need to be considered carefully. At best, the new ANCAP summary measures proposed could explain between 55% and 65% of the variation in UCSR crashworthiness scores highlighting that the current ANCAP protocols still do not reflect all important real world crash configurations and injury outcomes to key body regions. (Author/publisher)