关键词: 因子Copula, 空间相依, 风险度量, 农业保险

Abstract: Agricultural risks are spatially dependent, influenced by climate and production clustering.Spatial correlation of yields is influenced by distance and weather. Spatial correlation is usually stronger in left tail than middle and right tail due to natural disasters. Therefore, this paper constructs a Gee factor Copula regression model, and uses Odd Pareto, Odd Log-Logistic-F as marginal distributions to improve the fitting of the left tail. Gaussian process is used to describe the spatial dependence of the middle part of data and two exponential distributions are used as common factors to adjust the tail correlation structure of Gaussian process to reflect the asymmetric tail correlation of yield. Temperature, rainfall, latitude and longitude are introduced into Gaussian covariance to make the dependence flexibly reflect climatic characteristics and geographical differences. An application to actual data shows that our model performs better in describing the tail correlation when fitting the spatial dependence of yield. The risk measures calculated by Monte Carlo simulation with our model have important implications for agricultural catastrophe risk management in practice.

Key words: Factor Copula, Spatial Dependence, Risk Measurement, Agricultural Insurance