Title: Two-part $D$-vine copula models for insurance claim data
Authors: Claudia Czado - Technische Universitaet Muenchen (Germany)
Lu Yang - University of Amsterdam (Netherlands) [presenting]
Abstract: Insurance claim data usually follow a two-part mixed distribution: a point mass at zero corresponding to no claim and an otherwise positive claim from a skewed and long-tailed distribution. In addition, insurance companies usually keep track of policyholders' claim over time, resulting in longitudinal data. We study the longitudinal mixed claim data using a two-part $D$-vine copula model. We build two $D$-vine copulas, one is used to study the dependence of whether or not a claim is occurred over time, and the other is used to study the dependence in claim size given occurrence. We then use our model to investigate the time dependence of insurance claim using a dataset from the local government property insurance fund in the state of Wisconsin.