Correlation of Results from the New Car Assessment Program with Real Crash Data

Monash University Accident Research Centre - Report #115 - 1997

Authors: S. Newstead & M. Cameron

Full report in .pdf format [4.4MB}

Abstract:

In recent years, several major initiatives have been undertaken to assess relative vehicle occupant protection performance for consumer information. Two of these initiatives undertaken were the New Car Assessment Program (NCAP), which measures relative occupant safety of current model vehicles by measuring dummy responses in controlled crash testing, and the Used Car Safety Ratings (also known as Vehicle Crashworthiness Ratings), which estimates the relative risk of severe driver injury for individual models of vehicles involved in real crashes by analysing mass data. This project assesses the relationship between NCAP test results and data from real crashes by comparing the results of crashworthiness ratings to the outcomes of NCAP testing for vehicle models whose relative occupant protection has been assessed in both programs.

The analysis indicates a number of relationships between the results of NCAP testing and the outcomes of real crashes. Firstly, correlation analysis showed the results from offset NCAP testing have a much stronger association with real crash outcomes than do the results of full frontal NCAP testing. NCAP test results were also found to have the strongest association with the injury severity component of the crashworthiness rating which measures the risk of serious driver injury given any injury. Correlations were generally stronger between NCAP results and two car head on crashes than with all crash types. Mass adjustment of the NCAP probability measures also improved their relationship with real crash outcomes. Detailed analysis of injury data by body region generally confirmed the results of the correlation analysis using a more detailed and specific method of analysis.

Logistic regression models of crashworthiness ratings and its components as a function of NCAP measures were built providing a direct functional relationship between the two programs as compatible and consistent measures of relative vehicle occupant protection.

Executive Summary

In recent years, several major initiatives have been undertaken to assess relative vehicle occupant protection performance for consumer information. Two of these initiatives undertaken were the New Car Assessment Program (NCAP) and the Driver Protection Ratings (also known as Vehicle Crashworthiness Ratings and later published as Used Car Safety Ratings). The first of these estimates the relative occupant safety of current model vehicles by measuring dummy responses in controlled crash testing. The second initiative estimates the relative risk of severe driver injury for individual models of vehicles involved in real crashes by analysing mass data.

The aim of this project is to assess the relationship between NCAP test results and data from real crashes by comparing the results of crashworthiness ratings to the outcomes of NCAP testing for vehicle models whose relative occupant protection has been assessed in both programs. Comparison has been made using existing crashworthiness ratings based on all crash types, limited crashworthiness ratings derived from crashes of specific types, and modelling of crash outcomes as a function of NCAP test results. A second stage of the project examined the relationship between injuries recorded in Transport Accident Commission claim data and the corresponding measurements taken from the crash test dummies in the NCAP test procedures.

The analysis presented in this report indicates a number of relationships between the results of NCAP testing and the outcomes of real crashes. Firstly, correlation analysis showed the results from offset NCAP testing to have a much stronger association with real crash outcomes than do the results of full frontal NCAP testing. For both NCAP test configurations, the NCAP test results and their associated measures have the strongest association with the injury severity component of the crashworthiness rating which measures the risk of serious driver injury given that some injury was sustained. Correlations were generally stronger between NCAP results and two car head on crashes than with all crash types. Mass adjustment of the NCAP probability measures also improved their relationship with real crash outcomes.

Detailed analysis of injury data by body region generally confirmed the results of the correlation analysis using a more detailed and specific method of analysis. Full frontal NCAP femur loading readings showed strong association with average maximum AIS to the leg region in real crashes. There was also weak association between HIC and real crash head injury severity for this NCAP test configuration. Offset NCAP scores and average maximum AIS recorded in real crashes show strongest association in the head regions, with weaker association in both the chest and leg regions.

Capitalising on these relationships, logistic regression techniques were able to successfully build accurate models of crashworthiness ratings and its components as a function of NCAP measures, providing a direct functional relationship between the two programs as compatible and consistent measures of relative vehicle occupant protection.

Sponsoring Organisation: Baseline Research Program - Department of Justice, Transport Accident Commission, Royal Automobile Club of Victoria (RACV) Ltd, VicRoads.