Structure Breakpoint Problems Of Unitroot Test For Consumer Price Index In Chian

Most of the macroeconomic variables are all fluctuating with time, and show non-stationarycharacteristics in the long-term dynamic changes. Researches of the dynamic characteristics of themacroeconomic variables have important theoretical and practical significance for the government to makedecisions. Generally speaking, the study on time series starts from the unit root test. If the time series is aunit root process, stochastic trend of the series is formed by the accumulation and aggregation of innovation,so that when we forecast the series, the future value will be impacted by the past accumulated innovation,having a persistent effect on series fluctuation.If the time series data is stationary or trend stationary, innovation only has temporary effect, and theeffect will weaken over time constantly to zero. But time series data can also be trend stationary withstructural breaks. If this case is ignored, the series will be mistakenly inferred as unit roots, and thesubsequent data analysis will be misled.Based on the theoretical research on time series stability of macroeconomic variables, this paperexplores deeply on the long-term dynamic structure change of inflation through inspecting the monthly dataof consumer price fixed base index. A dramatic increase is found through observing the data in its longterm volatility, while most of the literature and research often think that inflation is a unit root process.Currently there are few researches paying special attention on structural breaks. Seasonal adjustments arecarried out after taking logarithm on CPI, and then initial judgment of seasonal-adjusted series stationarityis made through ADF test, PP test and KPSS test. Most results show that it’s a unit root process, but therehas an opposite result, when test for the first order difference series with drift.This paper mainly conducts parameters and non-parameters test on this inflation series based on Zivot&Andrews (1992) model and state space model. ZA model develops an endogenous structure breakpointtest method, i.e. the structure breakpoints are unknown, thus avoiding subjectivity of empirically judgmenton the dates of breaks. Reasonable judgment on data generation process improves the test level and power.The state space theory is a non-parameters test model. This can be analyzed intuitively by the S-PLUSsoftware. Most of the time series can be represented as structural elements, namely, trend, cycle, seasonal andirregular disturbance factor. The collection of these elements forms the state of the observable variables insome specific moment. Usually, these unobserved variables are called state variables. But the state spacemodel is able to put these unobserved variables and observable variables together to establish some kind ofstructure relations. Each state will be extracted through state space model after the structure time seriesconverts to state space model. Two different types of test methods are used in this paper to further verifythat structure breaks does exist in the inflation series dates.Test results of ZA model show that the level breakpoint occurs in June1992，the intercept and theslope for break year of the series is in May1993. From the state-space model, the results show a number ofdifferent breakpoints, and occurs around year1994and in2007.which has a great relationship with theAsian financial crisis happened in structural breaks tend to be impacted by some events (such as thefinancial crisis, policy reform, disasters, etc.). So due to the existence of external shocks when researchingon the long time series data sample, it’s very necessary to consider whether the time series data containbreakpoints, as well as the problems such as form and number of breakpoints.Finally, according to the test results of this paper, combining with the change of policy system sinceour country’s reform, it’s found that the serious inflationary periods in China (including the end of80’s lastcentury, and around year1994), is the most key periods for China’s economic transition, and also periodswith relatively large and frequent reforms.In a word, considering the existence of structural breakpoints when applying unit root test to timeseries, is of critical importance for inference about the series’ DGP. Correctly establishing data generationprocess model is the premise for testing stationary and converting to stationary series, and also is the key todeal with the other processing. Considering structural breaks in unit root test is not only of great theoreticalsignificance, but also of practical significance for forecasting, making policy and so on.