基于高频数据的VaR金融风险度量的研究

发布时间：2018-03-02 05:37

本文关键词： 赋权“已实现”双幂次变差高频波动 VaR　出处：《武汉理工大学》2013年硕士论文　论文类型：学位论文

【摘要】：近年来,对金融高频数据研究已经成为了金融计量学的一个全新的研究领域和方向。金融风险度量也是在目前金融自由化下风险管理中最重要的部分。本论文主要研究了金融市场高频数据的特性、建模,并将它其应用在到金融风险度量VaR中。本文的主要工作可以概括如下： (1)本文在高频数据的基础上,提出“已实现”波动率及三个改进高频波动量,调整“已实现”波动率(ARV)、“已实现”双幂次变差(RBV)、赋权“已实现”双幂次变差(WRBV)。WRBV赋权“已实现”双幂次变差考虑了“日历效应”,具有稳健性,同时是有效的高频波动估计量。 (2)本文比较三种最优时间间隔的选择方法的思路,指出了方差法是获取最优抽样时间间隔是更简单易行的方法。并且用“已实现”波动率及其改进的三个波动率估计量为例,实证研究证明了WRBV是比已实现波动更有效的波动率估计量。 (3)建立基于WRBV的ARFIMA高频波动模型,将其与GARCH模型相比,用AIC和SBC准则对模型模拟效果进行评价,证明其为更好的拟合模型。 (4)用改进的高频波动率WRBV代替RV用于计算VaR,即建立WRBV-VaR模型。用股指期货的高频数据对WRBV-VaR模型进行实证研究,并将其与传统的风险度量模型GARCH-VAR进行风险估测效果的比较。用Kupiec失败率检验法对对风险价值模型的效果进行评价,得出WRBV-VaR模型对VaR有更好的估测。 VaR风险度量可以将不同市场因子、不同市场的风险集成为一个数。WRBV-VaR模型对金融风险的度量的有一定的综合性,但是它只能对市场正常情况下的风险进行预测和控制,而不能对市场上出现的极端事件进行预测和控制。因此,本文对金融风险度量有一定的局限性。如何对模型本身进行改进使得预测效果更好,有待今后进一步的研究。
[Abstract]:In recent years, The research on financial high-frequency data has become a new research field and direction of financial metrology. Financial risk measurement is also the most important part of risk management under the current financial liberalization. Characteristics of high-frequency data in financial markets, Modeling and applying it to the financial risk measurement VaR. The main work of this paper can be summarized as follows:. 1) on the basis of high frequency data, this paper puts forward the "realized" volatility and three improved high frequency fluctuations. To adjust the "realized" volatility rate (ARV), "realized" "double power variation difference" (RBV), to "realize" "double power variation", to realize "double power variation" has been realized. "double power variation" takes into account "calendar effect" and is robust, and is an effective high frequency fluctuation estimator at the same time. This paper compares three methods of selecting optimal time interval, and points out that the variance method is a more simple and feasible method to obtain the optimal sampling interval, and takes "realized" volatility and its improved three volatility estimators as examples. Empirical studies prove that WRBV is a more effective volatility estimator than realized volatility. (3) the high frequency fluctuation model of ARFIMA based on WRBV is established. Compared with the GARCH model, the AIC and SBC criteria are used to evaluate the simulation effect of the model, and it is proved that the model is a better fit model. In this paper, we use improved high frequency volatility WRBV instead of RV to calculate VaR, that is, to establish WRBV-VaR model. Using the high frequency data of stock index futures, we make an empirical study on WRBV-VaR model. Compared with the traditional risk measurement model (GARCH-VAR), this paper evaluates the effect of the risk value model by using the Kupiec failure rate test method, and concludes that the WRBV-VaR model has a better estimate of the VaR. VaR risk measurement can integrate different market factors and different market risks into a number. WRBV-VaR model has a certain degree of integration to measure financial risk, but it can only predict and control the risk under normal market conditions. Therefore, this paper has some limitations on financial risk measurement. How to improve the model itself to make the forecasting effect better, need further research in the future.
【学位授予单位】：武汉理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F830.91;F224

【参考文献】