Econometric Analysis Of Crop Revenue Risk Under "Insurance Plus Futures"

From the perspective of national policy and revenue insurance development,this thesis establishes subsystems of yield risk,price risk and revenue risk,which clearly describes the risk and premium rate of corn revenue insurance in Shandong Province.Specifically,calculating a fair and reasonable revenue insurance premium rate by establishing the model is helpful to protect farmers’ planting revenue and reduces the burden on insurance companies’In general,it provides empirical experience for winning the food defense war under new financial weapons and simultaneously has positive benefits for strategies such as decentralizing the country’s food security risks,agricultural modernization and rural revitalization.In the yield risk,it was determined that the time factors such as the year and previous yield,the meteorological factors such as the average temperature and sunshine duration in the previous period are similar to the trend of the yield change through the grey relational degree.Using the mixed model of GM(1,1)and PCR,the risk indicators including the average yield reduction rate,disaster frequency and coefficient of variation of 16 cities in Shandong Province from 2005 to 2019 were obtained.It is found that the three coastal cities of Weihai,Dongying and Rizhao have the greatest yield risk and the yield reduction rate of the mixed model is generally lower than the local standard thresholds of risk rating in Shandong Province.Finally,the spatial distribution of three different yield risks—light,medium and heavy in Shandong Province was divided by the yield reduction rate.In the price risk,the import volume of the previous period was removed from 10 factors such as supply and demand,upstream and downstream agricultural products.Taking September 2005-April 2020 as the training set and May 2020-December 2021 as the test set,we compared time series models,neural network of RBF and single-layer,one-way but Multi-step LSTM according to whether there are relevant factors.It is found that the increase of the expansion speed in the RBF has little effect on the model error but the number of interneurons is sensitive so that the network will overfit if the maximum number of neurons is more than 20.Finally,the ARIMA and the single-layer multivariate LSTM neural network with one-way and 6-step are more suitable for forecasting the monthly corn futures price in different dimensions.In the revenue risk system,the optimal joint distribution of annual corn yield,corn price and soybean price of the three risk levels after detrended by cubic natural spline regression in the above system was determined through the VineCopula structure.Then we calculated the corn revenue insurance premium rate by Monte Carlo simulation under different yield risks and different coverage rates in Shandong Province.It is found that vine structures is distinctly different at different risks,and this system obtains a more comprehensive premium rate because it considers the synthetic dependencies compared with traditional premium caculating.The innovation of this thesis is to establish a comprehensive ecological system of corn revenue insurance in Shandong Province by qualitative and quantitative methods.On the one hand,we obtained the relevant indicators of corn revenue and tried a variety of models to predict the trend of yield and price.On the other hand,the object of insurance premium is converted into risk division and we considered the upper and lower tail correlation between soybean price,corn price and corn yield with the help of VineCopula.The disadvantage is that the parameter adjustment and model comparison are limited due to the small amount of data.