Dependence modeling of non-life insurance using the Bernstein copula

Copula is distribution functions which is used in modelling dependency between random variables. For this reason, it is frequently used in dependency studies. In this study, the effectiveness of the modeling of non-life insurance risks with the Bernstein copula was examined by comparing with other copula models. Study's data got from an insurance company which is operating in the field of non-life insurance in Turkey. This data includes paid incurred for 72 months in the area of traffic insurance, car insurance, liability insurance (imm), residential fire insurance, residential theft insurance, residential glass breakage insurance, personal accident insurance, seat personal accident insurance and personal health insurances. First of all, car insurance-traffic, traffic-personal health and car insurance-imm were selected which have high correlation between them. Than the distribution and parameters of these data pairs were determined. In order to select the model, Euclidean distances were calculated. Finally, for each copula model Risk exposure values (VaR) were calculated. The results of the study showed that it would be useful for insurance companies and regulators to consider the Bernstein copula model in risk models.