The objective of this study is to investigate the extent of biases in the police and occupant reported belt usage classifications (relative to a standard observer as represented by the NHTSA-sponsored investigating team), and the effects these biases have on the belt effectiveness estimate. The data used is the Restraint System Evaluation Program File (RSEP) provided by NHTSA. The study revealed that there is a significant amount of agreement (80%) between occupant and investigator concerning belt usage. There is also a significant amount of agreement (70%) between police and investigator on belt usage, but the agreement is significantly less than that between the occupant and investigator. Despite the presence of significant amounts of agreement between occupant and investigator, and between police and investigator, relative to the disagreements, the occupant consistently tends to overstate belt usage (bias = 0.6) whereas the police tendency to understate belt usage (bias = -0.17) is not as consistent. There is a stronger tendency for police to understate belt usage (bias = -0.44) when the occupants were injured than when they were uninjured (bias = -0.14). More specifically, when investigator reported belt not used, 18.2% of the occupants reported belt used and 26.3% of the cases were reported by police as belted. On the other hand, when investigator reported belt use, only 3.5% of the occupants reported belt not used, while 33.1% of the cases were reported by police as unbelted.
Abstract