基于Skewed-T Realized GARCH模型的沪深300指数波动性研究

发布时间：2018-03-28 18:02

本文选题：“已实现”波动率　切入点：Realized　出处：《天津商业大学》2017年硕士论文

【摘要】：自改革开放以来,随着市场机制的不断完善,我国资本市场有了较大的发展。作为国有企业改革的主要途径,股票市场已经成为国有企业进行机制改革和融资的重要场所。股票市场不仅在推动国民经济持续增长上发挥了不可替代的作用,同时也对世界经济一体化影响巨大。与传统的市场相比,股票市场是充满了不确定性的市场,信息的披露、资本的流动会导致股票市场的价格出现波动,使投资者们对其难以把控。剧烈的股票价格波动增加了投资者的投资风险,会给投资者们带来损失。因此,对股票市场价格波动的研究是十分重要且十分有意义的。本文使用数理工具和计量方法,对我国的证券市场特征进行描述和预测,揭示其内在规律。通过对股票价格波动率的建模分析,加深对股票价格波动实质的认识,把握未来股票市场的趋势,同时发现我国股票市场有可能存在的问题,并提出相应的改进措施。再进一步,通过理论结果与现实结果相比较,发现理论与实际不一致的地方,进而对所使用的模型可能存在的缺陷进行改进。其中,在对收益率的描述性统计分析中,除了对其主要指标包括均值、标准差、偏度、峰度的分析外,还发现其波动具有聚集性和周期性。从结果中可以看出,收益率序列呈现了尖峰厚尾的特征。在对“已实现”测度进行比较时,不同频率所得到的“已实现”方差也呈现出不同的特征,时间间隔越小,频率越大,序列的均值就越小。其中,5分钟时间间隔下的“已实现”方差与“已实现”核估计的各项特征十分接近。收益率序列的波动具有聚集性,即在一个较大幅度的波动后会跟随着另一个较大幅度的波动,在一个较小幅度的波动后也同样会跟随另一个较小幅度的波动。而其周期性,主要体现在与宏观经济周期相吻合。在Skewed-T Realized GARCH模型的实证分析中,我们通过模型对数据的拟合优度进行比较,确立了最佳的模型滞后阶数和最优分布,同时也利用信息冲击曲线等工具,分析了波动的非对称效应。在对模型的预测能力的分析上,通过与SkewedNormal Realized GARCH和传统EGARCH模型的比较分析,分别从样本内预测和样本外预测两个方面验证了Skewed-T Realized GARCH的优越性。Skewed-T Realized GARCH模型的优越之处在于新引入的“已实现”测度,而其原因在于波动率本身是剧烈波动的,而传统的EGARCH模型在收益率层面上是一个简单的ARMA过程,对于迅速变动的波动率序列拟合能力弱。而“已实现”测度可以对日内波动率的变动做出迅速的反应,使得参与拟合的序列能够包含日内波动率变化的信息。
[Abstract]:Since the reform and opening up, with the continuous improvement of the market mechanism, China's capital market has developed greatly. The stock market has become an important place for state-owned enterprises to carry out mechanism reform and financing. The stock market has not only played an irreplaceable role in promoting the sustained growth of the national economy. Compared with the traditional market, the stock market is full of uncertainty. The disclosure of information and the flow of capital will lead to the fluctuation of the price of the stock market. Make it difficult for investors to control. Sharp stock price fluctuations increase investors' investment risk and bring losses to investors. It is very important and meaningful to study the price fluctuation of stock market. In this paper, we use mathematical tools and measurement methods to describe and predict the characteristics of securities market in China. Through modeling and analyzing the volatility rate of stock price, we can deepen the understanding of the essence of the fluctuation of stock price, grasp the trend of stock market in the future, and find out the possible problems in the stock market of our country at the same time. And put forward the corresponding improvement measures. Further, by comparing the theoretical results with the practical results, we find that the theory is inconsistent with the practice, and then improve the possible defects of the model used. In the descriptive statistical analysis of the return rate, in addition to the analysis of the mean, standard deviation, deviation and kurtosis, it is also found that the volatility has aggregation and periodicity. When comparing the "realized" measure, the "realized" variance obtained from different frequencies also shows different characteristics. The smaller the time interval, the greater the frequency. The smaller the mean value of the sequence is, the closer the realized variance is to the kernel estimation at 5-minute intervals, and the volatility of the return series is convergent. That is, it follows another larger fluctuation after one larger fluctuation, and also follows another smaller fluctuation after one smaller fluctuation, and its periodicity, In the empirical analysis of the Skewed-T Realized GARCH model, we compare the goodness of fit of the data, and establish the best model lag order and optimal distribution. At the same time, the asymmetric effect of volatility is analyzed by means of information shock curve. The prediction ability of the model is compared with that of SkewedNormal Realized GARCH and traditional EGARCH model. The superiority of Skewed-T Realized GARCH in Skewed-T Realized GARCH model is verified from two aspects of intra-sample prediction and out-of-sample prediction, respectively. The advantage of Skewed-T Realized GARCH model lies in the newly introduced "realized" measure, and the reason is that volatility itself is highly fluctuating. On the other hand, the traditional EGARCH model is a simple ARMA process at the return level, which is weak for the rapidly changing volatility series, and the "realized" measure can react rapidly to the intraday volatility changes. So that the series involved in the fitting can contain information about intraday volatility changes.
【学位授予单位】：天津商业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F832.51;F224

【参考文献】