Time-varying X-efficiency: evidence from US property casualty insurers

The existing efficiency studies for the insurance industry rely on the time-invariant efficiency assumption, presumably because of the short duration of observation. However, temporal variation in firm efficiency could be substantial even for a short period of time. We demonstrate this point by examining the X-efficiency for the US property casualty insurers utilizing stochastic panel frontier models. Efficiencies are estimated under both time-invariant and time-varying efficiency assumptions using a large panel dataset from the National Association of Insurance Commissioners (NAIC) of years 2001-2006. It is shown that the two assumptions result in significantly different estimates. To test whether the estimated efficiencies correspond well to the properties that the true efficiencies should have, we investigate the relationship between the efficiency measures and other indicators of firm performance. The results suggest that the stochastic panel frontier model with time-varying efficiency provides more reasonable estimates.