The Research On Catastrophe Loss Assessment Of Forest Fire Insurance

In recent years, the comprehensive effects of frequent natural disasters, increasinglosses and human factors lead to raising frequency and intensity of forest fire extremeevents. In this context, selecting the appropriate methods and data for forest firecatastrophe loss assessment is particularly important. Meanwhile, the forest fire risk isconcentrating. Once the event of fire occurs, it tends to span multiple regions,endangering tens of thousands of farmers and large areas of woodland. The feature offorest fire easily developing to catastrophe makes the forest fire catastrophe riskassessment become more and more important.Currently, parametric method can distort catastrophe losses a lot and is not properfor the study of forest fire insurance. Correspondingly, the development of modernnonparametric statistical theory provides another route for catastrophe insurance los sassessment. However, the defect of tending to underestimate the probability of thepeak and overestimate the probability at trough is inevitable. Data sharpening methodcan preprocess the data so that it will be pressed slightly by sharpening thanbefore. The sharpening data will replace the original data and then follow the normalnon-parametric kernel density estimation method for evaluation. This method remainsand even enhances the benefits of common estimation. It also could make up somedrawbacks of estimation by preprocessing the data.In this thesis, based on statistical theory, the non-parametric kernel densityestimation method and data sharpening technology are used to solve the problem offorest fire catastrophe risk assessment and risk curve fitting error reduction. The thesisaims to technology innovation and formation of forest fire loss assessment modelbased on the theory study. It will help to solve problems of lack of scientific basis,human factors leading loss assessment, so as to improve our country’s underwritingrisk management level in the field of forest fire insurance. This thesis has collected alarge domestic insurance company’s forest fire loss data as an example, and has appliedthe data sharpening technique firstly to reduce bias. From the perspective of theory, wecan demonstrate the bias reduction of obtained risk curve compared to the traditionalrisk curve; from the data we can see that the evaluation results change to the directionaccording to the theory. With the expanding of coverage, the fitting degree between insurance data and social data is rising. All these improve the accuracy of the lossassessment from both the data and method perspective.