基于Markov机制转换模型的中国外汇市场波动分析

发布时间：2018-02-26 19:34

本文关键词： 人民币汇率人民币NDF报价 MRS-GARCH模型最大似然估计法　出处：《东北财经大学》2013年硕士论文　论文类型：学位论文

【摘要】：30年来,我国汇率制度进行了多次调整与改革,人民币汇率也经历了多次复杂的变动。1994年,人民币与美元非正式地挂钩,但美元兑人民币汇率只能在8.27至8.28元之间浮动。2005年,人民币汇率以市场供求为基础、参考一篮子货币进行调节的、有管理的浮动汇率制度。2008年,美国次贷危机引发全球范围的金融危机,我国适当调整人民币波动幅度以应对此次危机。2010年,中国银行表示将进一步推进人民币汇率形成机制改革。这些原因给人民币汇率的波动状况带来不确定性,因此,对人民币汇率行为的描述变得尤其困难。随着近几年国际上对人民币升值压力的不断加大,对中国外汇市场的变动研究也显得越来越重要。本文在此研究背景下,以马尔科夫状态转移模型与GARCH模型为理论基础,采用最大似然估计法,首先使用GARCH模型、EGARCH模型以及GJR模型对中国外汇市场波动情况进行拟合估计,证实了其波动的非对称性,即利空利好消息拥有的不同冲击效应。接着建立MRS-GARCH模型对中国外汇市场的波动情况进行进一步分析,将中国外汇市场的波动分为高波动状态与低波动状态两个状态,通过马尔科夫状态转移过程将两个状态的波动状况进行拟合,并从中得出了相关转移概率及稳态概率。在此分析过程中,针对序列表现出的偏峰厚尾分布,对各个模型采用不同的残差项分布假定：正态分布、t分布、时变t分布以及广义误差分布。最后,根据AIC、BIC准则以及损失函数的对比,对不同分布下的各个模型进行比较,选出相对最优模型进行了进一步分析。本文在研究日度人民币汇率收益率的前提下,又引入另一序列,即人民币无本金交割远期合约报价收益率作为对比数据,旨在加入市场行为的影响来进行模型的比较,最终得到如下结论：①我国外汇市场的波动存在明显的非线性特征；②我国外汇市场的波动具有非对称性,即利好消息和利空消息对外汇市场等的冲击大小并不相同；③我国外汇市场存在显著的两状态转移过程,处在两状态的转移概率都较高,而且其各自的持续程度也较久；④人民币NDF收益率考虑了市场行为的影响,因此与汇率收益率在波动情况下存在一定差异,但仅从GARCH族模型中看不到这种区别的显著特征,而从最优分布下的MRS-GARCH模型能够将这种市场预测行为显著的区别出来。根据得出的实证结论,并结合我国外汇市场的实际特征,文章在最后以政府、金融机构以及企业这三个角度出发,对我国外汇市场体制改革、法律制度、衍生工具以及规避风险等各个方面进行了简单分析及建议,从而建立一个更加成熟稳定的外汇市场。
[Abstract]:Over the past 30 years, China's exchange rate regime has been adjusted and reformed many times, and the RMB exchange rate has undergone many complex changes. In 1994, the RMB was informally pegged to the US dollar, but the USD-to-RMB exchange rate could only fluctuate between 8.27 and 8.28 yuan. The RMB exchange rate is based on market supply and demand, with reference to a basket of currencies to regulate the managed floating exchange rate regime. In 2008, the US subprime mortgage crisis triggered a global financial crisis. On 2010, the Bank of China indicated that it would further promote the reform of the RMB exchange rate formation mechanism. These reasons have brought uncertainty to the volatility of the RMB exchange rate, so, It is particularly difficult to describe the behavior of the RMB exchange rate. With the increasing international pressure on the appreciation of the RMB in recent years, the study of the changes in China's foreign exchange market is becoming more and more important. In this background, based on Markov state transition model and GARCH model, the maximum likelihood estimation method is adopted. Firstly, the GARCH model and GJR model are used to estimate the volatility of China's foreign exchange market. The asymmetry of the volatility is confirmed, that is, the different impact effects of the good news. Then, the MRS-GARCH model is established to further analyze the volatility of China's foreign exchange market. The volatility of China's foreign exchange market is divided into two states: high volatility state and low volatility state. The volatility of the two states is fitted by Markov state transfer process. The correlation transfer probability and steady-state probability are obtained. In the process of this analysis, for the partial peak and thick tail distribution shown by the sequence, the different residual distribution assumption is adopted for each model: normal distribution / t distribution. The time-varying t distribution and generalized error distribution. Finally, according to the comparison of the AIC-BIC criterion and the loss function, the models under different distributions are compared, and the relative optimal model is selected for further analysis. On the premise of studying the daily rate of return of RMB exchange rate, this paper introduces another sequence, that is, the rate of return of RMB non-deliverable forward contract as the comparative data, in order to add the influence of market behavior to carry on the comparison of the model. Finally, we get the following conclusion: 1 the volatility of China's foreign exchange market has obvious nonlinear characteristics. Second, the volatility of China's foreign exchange market is asymmetric. That is, the impact of the good news and the bad news on the foreign exchange market is not the same. There is a significant two-state transfer process in our foreign exchange market, and the transfer probability of the two states is higher. Moreover, the NDF rate of return of RMB 4 has taken into account the influence of market behavior, so there is a certain difference between the rate of return and the rate of return in the case of fluctuation, but only from the GARCH family model there is no obvious characteristic of this difference. From the optimal distribution of the MRS-GARCH model can significantly distinguish this market forecast behavior. Based on the empirical conclusions and the actual characteristics of China's foreign exchange market, the article, at the end of the article, from the perspective of the government, financial institutions and enterprises, to the reform of the foreign exchange market system, the legal system, The derivatives and risk aversion are analyzed and suggested in order to establish a more mature and stable foreign exchange market.
【学位授予单位】：东北财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F224;F832.52

【参考文献】