Information asymmetry in fire insurance: a frontier approach

摘要

The composed error stochastic frontier model is used to separate random variations in fire insurance losses from systematic unexpected losses due to adverse selection or moral hazard. Net premiums are an ex-ante predictor of losses, based on information available to insurers. Losses due to information unknown to the insurer are picked up by the half-normal part of the error term. Loss data for 275 stock fire companies operating in 51 states and territories in 1917 are analyzed in a random effects panel data model. The mean cost per insurer of adverse selection and moral hazard amounts to 15% of fire losses. This is an upper bound estimate because some portion of this cost may be due to failure to operate efficiently rather than incomplete information.