基于半参数分位数回归法的地震损失研究

Study on Seismic Loss Based on Semi-parametric Quantile Regression Method

为预测在险价值以帮助测算地震巨灾基金规模,文章基于中国1990—2015年的地震损失数据,论证了半参数分位数回归法的预测效果。结果表明:当置信度不够高时,半参数分位数回归法(方法1)的预测效果略好于单纯使用该方法选取节点并使用所选节点进行常规分位数回归法(方法2)的预测效果,反之则不然。建议在设置地震巨灾基金规模时,按照监管规定分别使用方法 1和方法 2对地震灾害的95%条件在险价值和99.5%条件在险价值进行预测。

For predicting the value at risk(VaR)to help calculate the size of funds for earthquake catastrophe,this paper demonstrates the prediction effect of semi-parametric quantile regression(SQR)method based on the seismic loss data of China from 1990 to 2015.The results show that when the confidence degree is not high enough,the prediction effect of SQR method(i.e.Method 1)is slightly better than that of the conventional quantile regression method(i.e.Method 2)which is simply used for the selected nodes and makes use of the selected nodes,and otherwise the result reverses.Finally the paper suggests using Method 1 and Method 2 to predict the 95%conditional VaR and 99.5%conditional VaR of earthquake disaster respectively according to the regulations when setting up the scale of funds for earthquake catastrophe.