A0626
Title: Threshold autoregressive nearest-neighbor models for claims reserving
Authors: Tak Kuen Siu - Macquarie University (Australia) [presenting]
Abstract: Motivated by claims reserving in non-life insurance, a class of threshold autoregressive nearest-neighbor (TAR-NN) models is discussed. The TAR-NN model extends a major class of parametric nonlinear time series models, namely threshold autoregressive (TAR) models, by introducing a random field structure. It also generalizes nearest-neighbor (NN) models by introducing a regime-switching mechanism. Attention is given to a class of self-exciting threshold autoregressive nearest-neighbor (SETAR-NN) models and their applications to claims reserving. The (strict) stationarity and geometric ergodicity of the SETAR-NN model are discussed. The conditional least-square (CLS) method is used to estimate the SETAR-NN model and some of its nested models. Simulation studies on the parameter estimates are conducted. Applications of the SETAR-NN model and the nested models for projecting future claims liabilities are presented based on insurance claims data.
