Empirical Research on Value-at-Risk Methods of Chinese Stock Indexes

University essay from Lunds universitet/Nationalekonomiska institutionen

Abstract: The Chinese stock market has been established for more than 20 years. Although it is not as mature as the highly developed western securities markets, it has a huge influence on the global economy. It is significant to study the risks of the Chinese stock market, especially the risk of stock indexes. Affected by the economic globalization today, more and more financial derivatives and financial instruments appear which may lead to the increase of related risk so that the demand of research on the risk of the financial market is also getting higher and higher. Risk measurement is a key in risk management, and its measurement methods are constantly evolving. Value at Risk (VaR) method is one of the effective methods to measure the financial risk, which is widely used in domestic and foreign financial institutions. Compared with traditional models, it has much more accuracy and reasonability and is much easier to implement. As the two main indexes in Chinese stock market, the Shanghai Composite stock index and the Shenzhen Component index are selected as the research objectives. And the loss series of the two indexes are tested through normality test, unit root test, autocorrelation test and ARCH effect test. The outcomes of these tests indicate these loss series are skewed and stationary with the effect of ARCH. Hereby, the GARCH-type models are suitable to be used to estimate VaR. The TGARCH model and the EGARCH model under the hypothesis of Student’s t-distribution and generalized error distribution are employed for the six test periods from 2011 to 2016. And it can be concluded with backtesting that all these four models (the VaR-TGARCH-t model, the VaR-TGARCH-GED model, the VaR-EGARCH-t model and the VaR-EGARCH-GED model) are appropriate for the two indexes despite the fact several models fail the Kupiec test for the period 2015-2016.For the Shenzhen Component index, the VaR-TGARCH-t model may fit it most because all numbers of violations for the six test periods fall in the confidence intervals.

  AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)