Comparison And Optimization Of Different Calculation Methods Of Value At Risk

Value-at-risk is one of the risk measurement indicators.It is the application of financial knowledge in statistics and measurement methods,and is used to measure risk.Users can intuitively understand their maximum loss value in a certain period and a certain probability in the future through the value of risk,to help users face risks and respond to risks.There are many calculation methods for value at risk,which are simply divided into three categories: nonparametric,semiparametric,and parametric methods.Based on this,more calculation models have been derived to play a role in different financial markets.This article selects four calculation methods of value-at-risk for comparison.In this paper,the theoretical background,significance and innovation points are firstly explained in theory.Then introduces existing literature and the value-at-risk model,including its concept,calculation method,advantages and disadvantages,and applications,and then introduces four kinds of value-at-risk calculation methods.The standard historical simulation method and the nuclear density estimation method are traditional calculation methods,and the volatility weighted historical simulation method and the variable nuclear density estimation method are improvements of the first two methods,respectively.The last selected sample data is the closing price of the Shanghai Composite Index.The data range is 3662 consecutive trading days from January 4,2005 to January 23,2020.The data is processed to obtain 3151 historical sample data and 510 controls.According to the data,the four methods are empirically tested and the following results are obtained.First,neither the standard historical simulation method nor the nuclear density estimation method can show the market volatility.The estimated volatility curve is almost a straight line,which is inconsistent with the actual volatility.This is becausethere are 3151 estimated data of the selected historical samples,and each time the estimation is performed,only one data is moved backward,which has little effect on the distribution of the historical sample data,so the fluctuation of the estimation results is very small.The historical simulation method of volatility weighting is greatly improved compared with the traditional method.Because the weight is adjusted by volatility weighting,each estimation result is weighted.First,establish a GARCH model according to the characteristics of the sample data,then predict the sample volatility,compare the sample volatility with the actual volatility,generate weights,and multiply each estimated data with the weight coefficient to obtain new estimation results The standard historical simulation method has almost no change in the estimation results,and the estimation results can reflect the changes in market risks.Second,the accuracy of the standard historical simulation method and nuclear density estimation method is not high.Under the condition of 510 trading days,the accuracy rate is not within a reasonable range,and the estimation is not accurate.This is because the selected sample data has extreme values,which have an impact on subsequent estimates.These data make the results estimated by the standard historical simulation method and the nuclear density estimation method too low,thus overestimating the market risk.The volatility weighted historical simulation method and the variable kernel density estimation method reduce the weight of historical sample data with a longer time distance,thereby improving the results and the accuracy of prediction.