Bound VaR And EVT Risk Management Model And Its Empirical Research

There are different types of financial risks, e.g., market risk, credit risk, operational risk, liquidity risk, business risk, etc. Managing these risks to minimize potential losses is essential to ensure viability, profitability and good reputation for a financial institution, especially when the financial crisis occurs. Value at Risk (VaR) method provides a single risk measure, which can measure the overall risk for financial institutions. However, most of the time financial returns exhibit fat-tails, skewness, kurtosis and other departures from normality. The traditional VaR measure, based on the assumption that portfolio returns are normal distributed, not only overestimates the market risk during market calmness, but also has difficulty in controlling the market risk during market crashes. In order to deal with the fat-tail problem, there are three common approaches. The first one is called nonparametric technique, including historical simulation and Monte Carlo simulation methods. The second is parametric technique. This method constructs a conditional-normal model, based on ARCH/GARCH model. The third is Extreme Value Theory (EVT) method.In recent years, the trading accounts at large financial institutions have grown rapidly and become more complex progressively. In this situation, a portfolio in its trading book may includes hundreds of trade positions, which suffer huge potential losses from hundreds of risk factors. At this time, computing the profit and losses (P&L) distribution of the portfolio will be a tough task. Since the suggested approaches mentioned above are always complex and computationally heavy, we want to find a new approach, which is fast, straightforward and computationally easy. Luciano and Marena presented a quick-to-compute VaR bounds as an alternative to these three methods for VaR assessment. This approach can cope with any distribution for marginal returns, including the fat-tailed ones. We do not require hypothesis on the joint distribution or its dependence structure. This method not only requires little information, but is also easy to compute. In this paper, the VaR bound method and EVT method are applied to study the risk of domestic stock market and the risk of domestic future market. Corresponding policy proposals for risk management are discussed.The innovation in this paper lies in three parts: First, we evaluate the performance of VaR risk model in domestic market. At the same time, we provide a detailed analysis on the violation clustering phenomenon of domestic market, and discuss the corresponding model back-testing procedure, which is also used to test the performance of bound VaR method during market crisis. In the second part, we incorporate the liquidity risk with the market risk and then evaluate the margin level setting in domestic future exchange by the EVT and bound VaR approaches. In the third part, we consider the diversity effect in the portfolio of future contracts. Considering the diversity, we find a balance for the future margin level setting, in the trade-off between little risk and higher efficiency of the capital.