The partial relationship between car age and risk is studied by means of logistic regression as applied to a large insurance policy data set - more than 211.000 records. Annual mileage and car owners' gender, age and province of residence are controlled for. Contrary to expectations the analysis reveals a negative relationship between car age and accident risk. Older cars sustain fewer liability damage accidents and also fewer injure accidents. It is suspected that this is due to risk compensation (offsetting behaviour). Our findings may have important implications for the cost (in)effectiveness of a system with periodic motor vehicle inspection. (A)
Abstract