Can Vehicle Maintenance Records Predict Automobile Accidents?

Abstract

This article proposes that vehicle maintenance records can provide useful information for predicting the probability that an owner will have an automobile accident. To test the hypothesis, we use a unique data set that is merged from an insurance company and a vehicle manufacturer in Taiwan. We find weak evidence to support our hypothesis. Among all the proxies for proper maintenance, we indicate that proper maintenance defined by the recommended kilometers is significantly negatively correlated with the loss probability in compulsory automobile liability insurance. The average loss probability decreases by 0.23 percent when the insured vehicle is properly maintained according to the recommended number of kilometers in the previous years, whereas the average loss probability for the overall sample is 0.49 percent. We further find that proper maintenance is insignificantly correlated with loss severity.