【作者】 李睿；

【导师】 董晨昱；

【作者基本信息】 山西大学 ， 统计学， 2022， 硕士

【摘要】 在信息化、科技化、全球化的当今世界,国家地区之间的关系逐渐加强,其中任一方经济的重大变动必定会引发世界范围内的蝴蝶效应,经济学上自然也离不开对变量间关系的研究。关系一词定性化可以表现为股票黄金期货市场上的联动性等,定量化可以表现为回归、相关等,而这些传统的方法通常都是基于时间序列平稳这一过高要求使得理论上的结果往往不具实际解释能力。协整的提出打破了关于平稳时间序列间关系研究的瓶颈,而非线性协整检验又为变量间长期均衡关系提供了一种全新的思路。针对传统非线性的秩方法具有检验效率低、过程繁琐且无法通过结果准确判断系统中具体存在的协整关系等缺点,本文提出非线性Johansen协整秩检验方法,通过非平稳序列的秩构建误差修正模型,可同时获取非线性协整关系的数量及相应协整变量。将所提方法应用于国际股票指数实证分析中。结果表明非线性协整检验能够更加有效地反映非平稳经济变量之间的长期均衡关系和变化机制,且验证了基于Johansen的非线性协整秩检验能在多变量情况下达到提高检验效率和准确获得协整关系的优化预期效果。传统的经济模型大都基于同频时间序列建立,而在现实生活中各类经济数据往往是不同频率的数据,传统的插值法、均值法等降频方法中大量的人为操作容易改变数据的先验知识,使得原始信息无法被充分利用。文中基于MIDAS混频回归模型提出全新的混频数据频率对齐方法对传统降频方法进行改进,并将之应用于混频宏观经济变量的实证分析中,取得了良好的协整检验结果;将变量间的长期均衡关系与短期影响关系相结合,根据协整检验结果建立货币与GDP的CoMIDAS模型与3年内滚动系数进行比较,以此全面地看待变量间的长短期均衡关系,最后通过与几类传统自回归模型进行比较证明混频协整CoMIDAS建模的有效性。更多还原

【Abstract】 In today’s world of informatization,technology,and globalization,the relationship between countries and regions is gradually strengthened.A major economic change in any one of them will inevitably trigger a worldwide butterfly effect.In economics,the study of the relationship between variables is also indispensable.Qualitative of the term relationship can be expressed as linkage in the stock gold futures market,and quantitative can be expressed as regression and correlation.However,these traditional methods are usually based on the high requirement of time series stability,which makes theoretical results often not have practical explanatory power.The proposal of cointegration breaks the bottleneck of research on the relationship between stationary time series,and the nonlinear cointegration test provides a new way of thinking for the long-term equilibrium relationship between variables.Aiming at the shortcomings of the traditional nonlinear rank method,such as low test efficiency,cumbersome process,and inability to accurately judge the cointegration relationship in the system through the results,this paper proposes the nonlinear Johansen cointegration rank test method,which is based on the rank construction error of the non-stationary sequence.By modifying the model,the number of nonlinear cointegration relationships and the corresponding cointegration variables can be obtained simultaneously.The proposed method is applied to the empirical analysis of international stock indices.The results show that the nonlinear cointegration test can more effectively reflect the long-term equilibrium relationship and change mechanism between non-stationary economic variables,and it is verified that the nonlinear cointegration rank test based on Johansen can improve the test efficiency and accuracy in the multivariate case.Obtain the desired effect of optimization of the cointegration relationship.Most of the traditional economic models are built based on the same frequency time series,but in real life,all kinds of economic data are often data of different frequencies.In the traditional interpolation method,mean method and other frequency reduction methods,a large number of human operations are easy to change the prior of the data knowledge,so that the original information cannot be fully utilized.Based on the MIDAS mixing regression model,a new frequency alignment method of mixing data is proposed to improve the traditional frequency reduction method,and it is applied to the empirical analysis of mixing macroeconomic variables,and good cointegration test results are obtained.Combining the long-term equilibrium relationship between the variables and the short-term impact relationship,establish a CoMIDAS model of currency and GDP according to the cointegration test results and compare them with rolling coefficients within 3 years,so as to comprehensively look at the long-term and short-term equilibrium relationship between variables.Comparison with traditional autoregressive models proves the effectiveness of the mixed-frequency cointegration CoMIDAS modeling.更多还原