基于非线性依赖关系分析的人民币汇率多元描述与预测
发布时间：2018-07-13 13:59
【摘要】：从人类经济社会发展的历史进程来看，各界对汇率问题的关注由来已久。作为目前世界上最大的发展中国家，中国的人民币在国际货币体系中的地位也由于其对世界经济发展的重要作用变得越来越突出。本文拟在把握汇率系统非线性依赖特征的基础上，结合考虑不同国际外汇交易市场上相应的人民币汇率序列之间在数据信息结构等方面存在的差异，从多元角度精确描述和有效预测中国外汇市场的人民币汇率行为。这不但可以为全球金融一体化条件下评价人民币汇率政策的合理性和有效性提供参考，还能为帮助货币当局在坚持主动、渐进、可控的原则下，准确认识人民币汇率未来改革方向提供可行的分析工具。 基于上述现实和理论背景，本文沿着从基础分析到实证检验，再到理论推断的整体思路展开研究。首先，在回顾国际外汇市场的发展状况和总结中国外汇市场的改革实践的基础上，从理论上探讨汇率行为研究的范畴，评价有关汇率行为描述和预测的两类基本范式和思路，并对三大货币体系下有关汇率决定理论和汇率模型进行比较；接着对与汇率行为描述和预测相关的统计学习理论、方法进行全面的分析，并实证检验多个市场人民币汇率数据中的非线性依赖关系；然后从以往研究中有关汇率系统非线性依赖关系产生的两个假设出发，对来自中国、美国、英国等全球主要外汇市场的人民币兑美元汇率数据的非线性依赖性和低维混沌特征进行一系列实证检验；随后在对非线性非参数神经网络计算方法与模型进行全面分析的基础上，提出一个用来描述和预测中国市场人民币汇率收益率行为的多元动态Gauss径向基神经网络模型，，并对各个外汇市场上的人民币汇率收益率序列之间的相关依赖关系进行验证；再后以动态Gauss径向基神经网络模型为基础，比较不同汇率预测模型以样本内描述和泛化能力上的差异，对有关中国外汇市场开放程度、人民币汇率弹性与混沌特性等方面预期的理论假设进行检验和判断。 本文通过上述实证研究得到一下主要结论：一是人民币汇率价格序列的BDS检验结果表明，各市场在不同研究期的汇率价格序列不服从随机游走假设，原因在于汇率系统中存在明显的非线性依赖关系；二是4个市场上不同时间跨度的人民币兑美元汇率收益率序列中的确存在可捕捉和解释的条件异方差非线性依赖结构，且具体表现为波动群集和杠杆效应共存的特征；三是不同国家外汇交易市场上不同时间跨度内的人民币兑美元汇率收益率序列中的一般非线性关系和特定的GARCH类非线性依赖结构在本质上具有“插话式”和“瞬时性”非稳定特性；四是多元动态动态Gauss径向基神经网络模型在描述人民币汇率数据特征、刻画汇率数据间多样依赖关系结构以及样本外预测精度和能力方面均优于实证研究中选取的其它非线性参数模型；五是在相关理论检验方面，不同研究期内中国外汇市场的人民币汇率收益率序列中包括一定比例的其它未知类型的一般非线性依赖结构。同时，人民币兑美元汇率收益率序列中的非线性依赖关系可能产生于一个混沌机制的假设也具备一定的合理性，系统动力学视角、相空间重构理论和混沌性质与方法是构建人民币汇率行为描述与预测模型时必不可缺的分析工具。此外，中国外汇市场上的汇率数据与其它市场同名汇率数据之间存在一定的可预测相关关系，中国外汇市场的开放程度、市场间关联程度和信息传递效率等在不同时期表现出了不同的结构化特征。
[Abstract]:From the historical process of human economic and social development, all walks of life have been concerned about the exchange rate for a long time. As the largest developing country in the world, the position of China's RMB in the international monetary system is also becoming more and more prominent because of its important role in the development of the world economy. This paper is to grasp the nonlinear exchange rate system. On the basis of their dependence on characteristics, considering the differences in the data information structure between the corresponding exchange rate sequences of different international exchange trading markets, this paper describes and effectively predicts the exchange rate of RMB in the Chinese foreign exchange market from a multiple angle. This can not only be considered to evaluate the people under the condition of global financial integration. It provides a reference for the rationality and effectiveness of the currency exchange rate policy, and also provides a feasible analysis tool for the monetary authorities to understand the future reform direction of the RMB exchange rate accurately under the principle of insisting on the initiative, gradual and controllable.
Based on the above realistic and theoretical background, this paper begins with the overall thinking from the basic analysis to the empirical test and then to the theoretical inference. First, on the basis of reviewing the development of the international foreign exchange market and summarizing the reform practice of China's foreign exchange market, the study of exchange rate behavior is discussed theoretically and the related exchange rate behavior is evaluated. Two basic paradigms and ideas of describing and predicting, and comparing the exchange rate decision theory and the exchange rate model under the three major monetary systems; then the statistical learning theory, which is related to the description and prediction of the exchange rate behavior, is analyzed comprehensively, and the nonlinear dependence of the RMB exchange rate data in multiple markets is tested. Then, starting from the two hypotheses about the nonlinear dependence of the exchange rate system in the previous study, a series of empirical tests on the non linear and low dimensional chaotic characteristics of the RMB exchange rate data from China, the United States and the United Kingdom are carried out, and then the nonlinear non parametric neural networks are used. On the basis of a comprehensive analysis of the calculation method and model, a multi dynamic Gauss radial basis neural network model is proposed to describe and predict the exchange rate of RMB exchange rate in the Chinese market, and the correlation between the exchange rate sequence of RMB exchange rate in each foreign exchange market is verified, and then the dynamic Gauss is used. Based on the RBF neural network model, the different exchange rate prediction models are compared with the differences in the description and generalization ability, and the theoretical hypotheses about the opening degree of China's foreign exchange market, the elasticity of the RMB exchange rate and the chaotic characteristics are tested and judged.
The main conclusions of this paper are as follows: first, the BDS test results of the exchange rate series of RMB show that the exchange rate sequence of each market does not obey the random walk hypothesis in different period of study. The reason is that there is a clear nonlinear dependence system in the exchange rate system; two is the different time span in the 4 markets. There is a conditional heteroscedasticity nonlinear dependence structure that can be captured and explained in the exchange rate sequence of RMB against US dollar, and it is characterized by the characteristics of the coexistence of volatility cluster and leverage effect. Three is the general nonlinearity of the exchange rate of RMB against US dollar in the different time span of different countries. The relationship and the specific GARCH class nonlinear dependence structure are essentially "interset" and "transient" instability. Four the multivariate dynamic dynamic Gauss radial basis neural network model describes the characteristics of the RMB exchange rate data, depicts the diversity of the exchange rate data, and the accuracy and ability of the sample prediction. It is better than other nonlinear parameter models selected in the empirical study. Five in the related theoretical test, the exchange rate sequence of RMB exchange rate in China's foreign exchange market includes a certain proportion of other unknown types of general nonlinear dependence in the different research period. Meanwhile, the nonlinearity of the rate of return of the people's currency to the dollar is nonlinear. The assumption that the dependence may arise from a chaotic mechanism is also reasonable. The system dynamics perspective, the phase space reconstruction theory and the chaotic properties and methods are indispensable analytical tools for the construction of the RMB exchange rate behavior description and prediction model. In addition, the exchange rate data in the Chinese foreign exchange market are similar to the other markets. There is a certain predictability correlation between rate data. The openness of China's foreign exchange market, the degree of inter market correlation and the efficiency of information transfer show different structural characteristics in different periods.
【学位授予单位】：湖南大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：F832.6
本文编号：2119606
[Abstract]:From the historical process of human economic and social development, all walks of life have been concerned about the exchange rate for a long time. As the largest developing country in the world, the position of China's RMB in the international monetary system is also becoming more and more prominent because of its important role in the development of the world economy. This paper is to grasp the nonlinear exchange rate system. On the basis of their dependence on characteristics, considering the differences in the data information structure between the corresponding exchange rate sequences of different international exchange trading markets, this paper describes and effectively predicts the exchange rate of RMB in the Chinese foreign exchange market from a multiple angle. This can not only be considered to evaluate the people under the condition of global financial integration. It provides a reference for the rationality and effectiveness of the currency exchange rate policy, and also provides a feasible analysis tool for the monetary authorities to understand the future reform direction of the RMB exchange rate accurately under the principle of insisting on the initiative, gradual and controllable.
Based on the above realistic and theoretical background, this paper begins with the overall thinking from the basic analysis to the empirical test and then to the theoretical inference. First, on the basis of reviewing the development of the international foreign exchange market and summarizing the reform practice of China's foreign exchange market, the study of exchange rate behavior is discussed theoretically and the related exchange rate behavior is evaluated. Two basic paradigms and ideas of describing and predicting, and comparing the exchange rate decision theory and the exchange rate model under the three major monetary systems; then the statistical learning theory, which is related to the description and prediction of the exchange rate behavior, is analyzed comprehensively, and the nonlinear dependence of the RMB exchange rate data in multiple markets is tested. Then, starting from the two hypotheses about the nonlinear dependence of the exchange rate system in the previous study, a series of empirical tests on the non linear and low dimensional chaotic characteristics of the RMB exchange rate data from China, the United States and the United Kingdom are carried out, and then the nonlinear non parametric neural networks are used. On the basis of a comprehensive analysis of the calculation method and model, a multi dynamic Gauss radial basis neural network model is proposed to describe and predict the exchange rate of RMB exchange rate in the Chinese market, and the correlation between the exchange rate sequence of RMB exchange rate in each foreign exchange market is verified, and then the dynamic Gauss is used. Based on the RBF neural network model, the different exchange rate prediction models are compared with the differences in the description and generalization ability, and the theoretical hypotheses about the opening degree of China's foreign exchange market, the elasticity of the RMB exchange rate and the chaotic characteristics are tested and judged.
The main conclusions of this paper are as follows: first, the BDS test results of the exchange rate series of RMB show that the exchange rate sequence of each market does not obey the random walk hypothesis in different period of study. The reason is that there is a clear nonlinear dependence system in the exchange rate system; two is the different time span in the 4 markets. There is a conditional heteroscedasticity nonlinear dependence structure that can be captured and explained in the exchange rate sequence of RMB against US dollar, and it is characterized by the characteristics of the coexistence of volatility cluster and leverage effect. Three is the general nonlinearity of the exchange rate of RMB against US dollar in the different time span of different countries. The relationship and the specific GARCH class nonlinear dependence structure are essentially "interset" and "transient" instability. Four the multivariate dynamic dynamic Gauss radial basis neural network model describes the characteristics of the RMB exchange rate data, depicts the diversity of the exchange rate data, and the accuracy and ability of the sample prediction. It is better than other nonlinear parameter models selected in the empirical study. Five in the related theoretical test, the exchange rate sequence of RMB exchange rate in China's foreign exchange market includes a certain proportion of other unknown types of general nonlinear dependence in the different research period. Meanwhile, the nonlinearity of the rate of return of the people's currency to the dollar is nonlinear. The assumption that the dependence may arise from a chaotic mechanism is also reasonable. The system dynamics perspective, the phase space reconstruction theory and the chaotic properties and methods are indispensable analytical tools for the construction of the RMB exchange rate behavior description and prediction model. In addition, the exchange rate data in the Chinese foreign exchange market are similar to the other markets. There is a certain predictability correlation between rate data. The openness of China's foreign exchange market, the degree of inter market correlation and the efficiency of information transfer show different structural characteristics in different periods.
【学位授予单位】：湖南大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：F832.6
本文编号：2119606
本文链接：https://www.wllwen.com/jingjilunwen/guojijinrong/2119606.html
