Modelling outstanding claims liability for workers’ compensation, general liability and automobile liability using Bayesian MCMC simulation

• Main Authors: Pan, Jin, Nyon, Yan Zheng, Zheng, Yanxiong
• Other Authors: Li Ka Ki Jackie
• Format: Final Year Project
• Language: English
• Published: 2009
• Subjects: DRNTU::Business::Finance::Actuarial science
• Online Access: http://hdl.handle.net/10356/15259
• Institution: Nanyang Technological University

The group adopts the method proposed by Li [2008] to model historical claims run-off data for three lines of business; workers’ compensation, general liability and automobile liability. The group uses one Bayesian model which is similar to the gamma model within the generalized linear models (GLMs) and the WinBUGS software to run the MCMC simulation. Then, the group examines the residuals after fitting the model to find the best model for each line of business. Formal model criticism procedures are adopted to refine and adjust the model structures. In the modelling process, the group uses polynomials to smooth parameters from WinBUGS and tackles heteroscedasticity and superimposed inflation. Eventually, the group finds the final adjusted models which are reasonable in both quantitative tests and qualitative reasoning. The detailed tracks of how the group improves the models, conducts relevant tests and comparisons, are shown in the appendices.