Quantile Regression And Its Application In Risk Measurement

August 2007 Subordinated Debt crisis sweeping the United States,the European Union and Japan which are the world's major financial markets,expanding its influence reach of the global,despite the influence different in regions,but it is sure that the financial crisis has now evolved the whole world into the global biggest crisis that the financial system has faced since 1929.The financial crisis has lead to the economy deterioration in many regions.Despite the constant introduction of positive policies,from the global perspective,the trend of deterioration has not yet to the bottom.From the U.S.sub-loan crisis of the Wall Street turmoil has evolved into a global financial crisis.Rapid development of this process,the number of large,the impact is huge,nobody has expected.Except for the current global financial crisis, financial markets have never been a no-wave world.With the different types and degrees of financial risk,as well as the "domino" effect brought about the trend of economic globalization,has been severely affected the global economy's health and prosperity,all the problems and the lack of financial supervision increasing people's attention.With the development of the financial sector,particularly the emergence of financial derivatives and innovation,brought to the investment and financing channels and more convenient,but also virtually the financial risk of new species,an increase of its occurrence.The financial environment has become more and more active and open,how to prevent and control the growing internal and external risk became an important issue,and how to grasp the timing of policy implementation. to fing the regulatory of crisis.to face the challenges of crisis.Therefore,we can see, it is precisely because the existence of financial risk,that the financial supervision system develop. Risk control and management attached more and more cares,management of risk lies in the measure,so the risk measure methods have enjoyed great development.To raise the level of risk management and control,people have to develop the risk and exploration methods,the current major comparative risk measure methods including sensitivity analysis,volatility method,the value-at-Risk VaR,stress tests Stress Testing,extreme value theory EVT.Value at Risk is currently the most widely used method.The key of Value at Risk method is calculated VaR,VaR was referring to the current market conditions,at the level of a certain probability (confidence level),the greatest possible loss of the financial position in a specific period of time in the future.Fluctuations in different models and the valuation model will correspond to different methods of calculating VaR,historical simulation, Monte Carlo simulation method,Ristmetrics risk measurement system,etc.the core of most methods is to estimate the future statistical distribution of financial position.In order to find the more reasonable measure of financial risk VaR calculation purposes,this article attempts to built Riskmetrics model based on GARCH,and semi-parametric quantile regression estimation to calculate VaR. Quantile regression have some statistics photogenic characteristics,this article attempts to compare Riskmetrics model and quantile regression methods in the calculation of VaR.In the confidence level of 95%and 99%,use the two models on the Shanghai Composite Index and Shenzhen Index component,modeling and estimate parameters,four sets of results obtained,do Kupiec likelihood ratio test to the results.Wo found the two modela all pass the test,what more both have good temperament.As can be seen from the results.Riskmetrics model based on GARCH depicts the nature of the sequence vary wellwhich has a fat-tail and the peak,the model is better than Riskmetrics model based normal distribution.We also see from the second model that Quantile regression method has statistical superiority.(1) Without the random perturbation of the distribution to make any assumptions,the entire regression model has strong stability;(2) quantile regression does not use a connection function to describe the dependent variable mean and variance therefore have better elastic properties;(3) Quantile regression for all points due to the median regression,it appears the date with outliers resistance;(4)Different from the ordinary least-squares regression,quantile regression for the dependent variable has a monotonic transformation;(5) Quantile regression with the large sample estimated parameters have good performances.In this paper,the second model is the most simple based on quantile regression methods autoregressive model,I believe that through further improvements of the structural design,the model using quantile regression technique will be much more better.Through the application of Quantile regression technology in practice,it is foreseeable that the application of quantile regression technique in the field of risk measure has a broad vision,what more,quantile regression technique open a new path for the risk measure.It is gratifying comparison,the software R,SRS,EIVIEWS6.0 has been provided the function of quantile regression,but dynamics forecast that contains the time window need for more complex manual programming.This papers can only be categorized as a study and practice of risk measure,for the depth realization of the VaR method.By,recalling the theory,determine the research methods and content,and the practice of model building and testing process, it is a more careful study of the VaR risk measure which is a very important theory.