On a new family of composite models with generalized log-Moyal tail for modeling insurance loss data

Abstract

In this paper, four new families of composite models are developed for modeling claim severity of mixed sizes. These models are developed, considering generalized log-Moyal distribution for the tail using Mode-Matching techniques. The proposed models are applied to fit well-known Danish fire insurance data set. The comprehensive study of model selection suggests that the proposed composite models with generalized log-Moyal distribution for the tail are an appropriate choice as compared to other heavy-tailed distributions. In addition to this, the behavior of risk measures viz. Value-at-Risk and Limited expected value, for proposed models, are assessed and compared with other existing models.