Calculate the Financial Risk According to Extreme Value Theory

Over the past few decades, the financial market turmoil frequently shocking , people dedicated to research deeper financial markets operating mechanism while also strengthening financial regulation efforts. Financial risk measure theory , portfolio theory and the capital asset pricing theory laid the cornerstone of modern financial management theory . Three strain, are inseparable, the former is based on the latter two . Financial risk measure for each generation and effective solution to the problem will have a profound impact on them . On this basis, this study summarizes the main issues of financial risk measurement and problem solving . Including the distribution of financial asset returns thick tail, the conditions of extreme value distribution , the distribution of extreme integrity, and return on assets in the portfolio distribution and other issues. In this study, the first two issues of re- summarize and effective solution for China 's securities market ( mainly the stock market ) is also one of the practice of undeniable stroke. Specifically, the sample block maxima theoretical models and valve apex (Peaks over threshold, POT) under the framework of two theoretical models , the article focuses on the model parameter estimation. The former includes both the model parameters include the maximum likelihood estimate of the median point of maximum likelihood estimates and log- likelihood contour interval estimation. The latter includes both based on the generalized Pareto distribution (GPD) maximum likelihood parameter estimation , but also including moment estimation method . In the Moment estimation, this paper shape parameter moment estimate 1 / ( ? ) Practice , practice counting only analyzed the first two moments estimator causes and conditions given quadratic Hall buffet basis points digit random number algorithm flowcharts and procedures . Empirical analysis , first with GARCH (1,1) model fit the logarithm of the Shenzhen Component Index gains its standardized error term fitting generalized Pareto distribution (GPD). In calculating the standardized error term and logarithmic returns on the basis of the median , and finally against the number of binomial test, the test results show that the model has good predictive effect estimates .