我国股指期货的动态保证金设定水平研究

发布时间：2018-06-04 22:16

本文选题：股指期货 + 动态保证金　；参考：《山东大学》2017年硕士论文

【摘要】：保证金制度可以有效管控期货市场风险,设置合适的保证金水平,是提高期货市场运行效率、实现期货市场长期稳定与繁荣的基石。我国的静态保证金制度在运行过程中面临着调节机制僵化、设置偏高等弊端,而动态保证金可以根据市场风险调整保证金水平,既能有效管控市场风险,又可以提高期货市场的效率。因此,动态保证金是保证金制度发展的潮流,研究动态保证金设定方法,具有十分重要的现实意义。股指期货保证金水平的设定关键是对风险价值VaR的度量和计算,本论文采用理论分析和实证研究相结合的研究方法。理论部分主要介绍了参数法、非参数法、极值理论等VaR的计算方法,其中参数法基于收益率序列服从正态分布的假设,主要包括Delta—正态分布、EWMA、GARCH模型,非参数法主要是指历史模拟法,从历史数据来模拟计算VaR。极值理论主要介绍了 POT模型的基本原理,以及将GARCH模型和POT模型相结合的VaR计算。实证部分选取2010年4月16日至2017年4月18日的股指期货活跃合约的日结算数据,建立五种动态VaR模型,利用Kupiec检验方法对五种模型在不同时间窗口和置信水平情形下的准确性进行检验。在实证过程中对数据的正态性、平稳性和波动聚集性进行检验,检验结果表明我国的股指期货收益率序列是平稳序列,并不不服从正态分布,具有尖峰厚尾、波动聚集性和长时记忆性等特征。实证结果表明历史时间窗口的选择对Delta—正态分布和历史模拟法影响程度较大,较短的时间窗口 250天VaR的预测结果更灵敏,在模型预测准确的情况下击穿频率较低。在95%的低置信水平下,只有GARCH-P0T模型存在高估风险的状况,其余四种模型均预测准确。在97.5%、99%的高置信水平下,Delta—正态分布、EWMA、GARCH存在一定程度的风险低估状况,这与收益率存在尖峰厚尾的实际情况相符。历史模拟法在这两种高置信水平下,对VaR的预测最为准确,GARCH-POT模型在99%的最高置信水平下VaR预测结果准确,这是由于极值理论的基本原理和特性决定的。
[Abstract]:Margin system can effectively control the risk of futures market, set the appropriate margin level, is to improve the efficiency of the futures market, to achieve long-term stability and prosperity of futures market cornerstone. The static margin system in our country is faced with some disadvantages such as rigid regulation mechanism and high setting, while dynamic margin can adjust the margin level according to the market risk, which can effectively control the market risk. It can also improve the efficiency of the futures market. Therefore, dynamic margin is the trend of the development of margin system. It is of great practical significance to study the method of dynamic margin setting. The key to setting the margin level of stock index futures is to measure and calculate the value of risk VaR. This paper adopts the research method of combining theoretical analysis and empirical research. The theoretical part mainly introduces the calculation methods of VaR, such as parametric method, non-parametric method and extreme value theory. The parametric method is based on the assumption of normal distribution of return sequence, including Delta-normal distribution and EWMA-GARCH model. The non-parametric method mainly refers to the historical simulation method, which simulates and calculates VaR from historical data. The extreme value theory mainly introduces the basic principle of POT model and the VaR calculation which combines GARCH model with POT model. The empirical part selects the daily settlement data of the active stock index futures contracts from April 16, 2010 to April 18, 2017, and establishes five dynamic VaR models. The accuracy of the five models under different time windows and confidence levels is tested by using Kupiec test method. In the process of demonstration, we test the normality, smoothness and volatility aggregation of the data. The results show that the stock index futures yield series in China is a stationary sequence, which does not disobey the normal distribution, and has a sharp and thick tail. Volatility clustering and long-term memory and other characteristics. The empirical results show that the choice of historical time window has a great influence on Delta-normal distribution and historical simulation. The VaR prediction results of a shorter time window of 250days are more sensitive, and the breakdown frequency is lower when the model is accurate. Under the 95% low confidence level, only the GARCH-P0T model has overestimated risk, and the other four models are accurate. Under the high confidence level of 97.5%, Delta-normal distribution and EWMA-GARCH have a certain degree of risk underestimation, which is consistent with the actual situation that the yield has a sharp and thick tail. Under these two high confidence levels, the historical simulation method is the most accurate in the prediction of VaR at the highest confidence level of 99%, which is determined by the basic principle and characteristics of extreme value theory.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F724.5

【相似文献】