I conduct inference on moral hazard in the Italian automobile insurance market. I disentangle moral hazard from adverse selection and state dependence by exploiting the nonlinearities in the penalties across driving records and companies, and a discontinuity in the cost of accidents in the last 60 days of the contractual year. I employ a unique matched insurer-insuree panel dataset, containing rich information on 4,316,647 auto insurance contracts underwritten by all Italian insurers. The results demonstrate that moral hazard is a pervasive feature of the market, although its magnitude varies across companies.