中国物价指数单位根检验中的结构断点问题

发布时间：2018-01-15 00:07

本文关键词：中国物价指数单位根检验中的结构断点问题　出处：《河南大学》2014年硕士论文　论文类型：学位论文

【摘要】：多数宏观经济变量序列都是具有趋势性，在长期的动态变化中，表现出非平稳的特性。判断和确定宏观经济变量的动态特征对国家做出决策和制定措施有着重要的理论和实践意义。通常来说，对于时间序列的研究都是最先从单位根检验入手的，若时序数据存在单位根过程，则序列的随机趋势是由新息的加总累积形成的，这样在我们对序列进行预测时，未来值都会受到过去所累积新息的影响，对序列的波动有持续性影响；若时序数据是平稳或者趋势平稳的，那么新息只有暂时性效应，随着时间就会不断衰弱趋于零。但时序数据也可能是含有结构断点的趋势平稳的情形，所以在忽视这种情况下，而序列确实是含有结构断点的趋势平稳过程，这样就造成误判为单位根，对潜在DGP的判定出现错误，更会导致后序的数据处理出现偏误。基于对宏观经济变量时序平稳性研究的理论意义，本文通过考察居民消费价格定基指数月度数据，对通货膨胀的长期动态结构变化进行了较为深入的探究。先观察所研究的数据对象发现其长期波动过程中，出现了剧烈的大幅度的增长，而大多数文献及研究常常认为通货膨胀序列是单位根过程，目前关于深入挖掘结构断点的资料还很少。文中对消费价格定基指数数据取对数后，进行季节调整，然后通过常规的ADF检验、PP检验、KPSS检验对经过季调后的序列进行平稳性的最初判断，多数结论显示其为单位根过程，而KPSS检验中对于带有漂移的一阶差分检验，却出现了相反的结果。本文中主要基于Zivot和Andrews（1992）模型和状态空间模型对该通胀序列进行了参数和非参数的检验。其中，ZA模型理论是运用内生性结构断点检验方法，即结构断点是未知的，这样就避免了经验判断结构断点日期的主观性，当合理判断数据生成过程的情形下，，提高了检验水平和功效。状态空间理论是一种非参数检验模型，本文基于S-PULS软件，通过命令窗口实现对序列形态的直观判断。大多数时间序列都可以表示成结构性要素，即趋势、周期、季节和不规则扰动因素。这些要素的集合就构成了可观测到的变量的在某时刻的状态。在一般的模型里面，这些不能观测到的变量，被称之为状态变量。而状态空间模型却能够把这些不能观测到的状态变量与可观测变量建立某种结构关系，结构时间序列转化为状态空间模型以后，各状态量就通过状态空间模型被提取出来。本文用两种不同类型的检验方法进一步验证了通货膨胀时序数据确实有结构断点的存在。 ZA模型得出的检验结果显示截距上的断点日期为1992年6月，截距和斜率上都发生结构突变的断点日期为1993年5月。而状态空间模型则显示出该序列存在多个不同断点，在1994年前后和2007年。造成结构断点发生的成因，往往是受到一些事件的冲击（譬如金融危机、政策改革、灾害等等），所以在研究长期时序数据样本时，由于外部性冲击事件的存在，非常有必要考虑时序数据中是否含有结构断点，以及结构断点的形式和个数等问题。最后，根据本文得到的检验结果，并结合我国改革以来的政策制度的革新变化，发现在我国发生较为严重的通货膨胀时期（上个世纪80年代末和90年代初中期）是我国经济转轨最关键时期，也是改革力度相对较大、次数相对较频繁的时期。总之，当进行时间序列的单位根检验时考虑结构断点的存在问题，对于正确判定序列的数据生成过程（DGP）至关重要，正确建立数据生成过程的模型是检验平稳和向平稳序列转化的前提，而序列是否正确又是进行其他处理的关键。考虑结构断点的单位根检验除了具有重大的理论意义之外，对于国家进行经济预期、制定政策等有着实际的指导意义。
[Abstract]:Most macroeconomic variables are trend, dynamic changes in the long term, exhibit non-stationary characteristics. Judge and determine the dynamic characteristics of macroeconomic variables on the national decision-making and has important theoretical and practical significance to develop measures. Generally speaking, for the study of time series is the first from the unit root test to start, if the timing data are unit root process, stochastic trend sequence is composed of new information and the total accumulation, so that when we predict the sequence, the future value will be cumulative past innovation influence, have lasting effects on sequence fluctuation; if the timing data is stationary or trend stationary, so new information only temporary effects, as time will continue to decline to zero. But the sequence data may also be the breakpoint containing structural trend is stable, so ignore this In fact, the sequence is indeed a trend stationary process containing structural breakpoints, which results in misjudgement as unit root, and the decision of potential DGP is wrong, which will lead to errors in subsequent order data processing.
The theoretical significance of the research on the stability of time series based on macroeconomic variables, this paper investigates the consumer price index based on monthly data, the long-term dynamic changes of inflation is elaborated. The first observation data of the object of study found that the long wave process, there have been dramatic substantial growth, and most of the literature and research often think inflation sequence is a unit root process, there is little about the deep mining structure information. The breakpoint in the consumer price index data after logarithm, seasonal adjustment, and then through the ADF test, conventional PP test, KPSS test, the initial judgment of the sequence after seasonally adjusted smoothly the conclusion shows that the majority of unit root process, and KPSS test for a drift of one order difference test, there was an opposite result.
This paper is mainly based on Zivot and Andrews (1992) model and the state space model of the test parameters and non parameters of the inflation series. Among them, ZA model is a structure using the method of breakpoint test, the structural break is unknown, thus avoiding the experience judgment break date subjectivity, when reasonable to judge the data generating process conditions, improve the inspection level and efficiency. The state space theory is a non parametric test model based on S-PULS software, implementation of intuitive judgment sequence pattern through the command window. Most of the time sequence can be expressed as structural elements, namely, trend, cycle, seasonal and irregular disturbance factors set. These elements constitute at a time the state of the observable variables. The model in general, these cannot be observed variables, called state variables. The state space model is able to put these state variables can not be observed with observable variables to establish a structure, structure of time series into a state space model, the state through the state space model is extracted. The two kinds of inspection methods to further validate the inflation time series data does have a break there.
ZA model test results show that the intercept breakpoint date is June 1992, the intercept and the slope has breakpoint date structure change for May 1993. While the state space model shows that the sequence has many different breakpoints, before and after 1994 and 2007. The causes of structural break occurs, is often influenced by the impact of events (such as financial crisis, policy reform, disaster and so on, so in the study) time series data samples, due to external impact events are very necessary to consider whether the time series data structure contains a breakpoint, and structural form and the number of breakpoints.
Finally, according to the test results, combined with changes in our country since the reform of the policy system, found that the more serious inflation happened in China (the last century at the end of 80s and early 90s) is the most critical period of economic transition in China, the reform is relatively large, the number of relatively frequent periods.
In conclusion, considering the existing problems of the structure when the breakpoint for time series unit root test, to determine correct sequence data generating process (DGP) is to establish the correct data generation process model is the premise of smooth and inspection into stationary sequence, and the sequence is correct and is the key to consider the unit root of other treatments. In addition to the structural break test is of great theoretical significance, is expected for the national economy, and have practical guiding significance to formulate policies.

【学位授予单位】：河南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F726;F224

【参考文献】