Risk Management under Extreme Loss

This paper empirically analyzes tail behavior of rice loss due to typhoon. One of the issues of risk management is the choice of the distribution of loss data. Using peak-over-threshold approach to extreme value modeling, we fitted generalized Pareto distribution to agriculture natural disaster loss data of Taiwan from 1971 to 2005. By comparing with standard parametric loss modeling based on lognormal and gamma distributions, the appropriateness of the upper tail fitting to loss data and find generalized Pareto distribution outperforms classical parametric fits was evaluated. Finally, we computed tail risk measures and draw some implication of risk management.