Positive Research On Returns And Volatility Of China And Main International Stock Markets

Through Cointegration analysis, Vector Error Correction Model (VECM), Granger Causality Method, Impulse response function, Variance Decompositions and multivariable GARCH models, we research the correlation of the returns and volatilities among the stock markets of USA, UK, Japan, Hong Kong, Shanghai & Shenzhen in China.This study intends to find the Cointegration relations among the international stock indices and national indices. We want to find out whether the six stock markets can be integrated and have the same long-term trend. Then we set up VECM on the foregoing basis to study how Cointegration relation of stock markets affects every stock market's return and to observe if the coefficients are significant. At last, we want to know the correlation of return volatilities among DJIA, FTAE100, NIKKEI225, HSI, SS, SZ through BEEK model.With the exception of the first chapter, which is an introduction of this paper, the paper consists of three sections: theory basis and index choice, the model system and test method, the empirical study and results.1. Theory basis and index choiceAs economic globalization goes, the correlations among countries are more and more closed. One stock market's trend always affects other stock market, and the risk shows"domino effect". Research of correlation between major global stock markets and national stock markets, is greatly helpful for us to know the whole world's stock market, and use the successful experience to the national stock markets. We research the correlation of the Return and Volatilities among the stock markets of New York, Landon, Tokyo, Hong Kong, Shanghai & Shenzhen, in which New York and Landon are the global international financial centre. Tokyo and Hong Kong is the financial centre in Asia. And Shanghai and Shenzhen is our research centre. So it is practical for us to choose these countries'indices.In order to analyze the correlation between different countries'stock markets, we select six indices: DJIA, FTSE 100, NIKKEI225, HSI, SS, & SZ. The sample is daily index from 1996/12/26 to 2007/02/28. We deal with the series by the algorithm of ln(x) and dln(x), and we have the natural logarithm series and difference of natural logarithm series (the stock index's return series).From the point of view of international securities business, the stock market indices show the same movement trend more and more obviously. Because of the economic globalization, major stock markets in developed countries show evident characteristic of co-movement.2. Introduction to the model system and test methodThis paper chiefly uses Cointegration analysis, Vector Error Correction Model (VECM), the Granger Causality Method, Impulse response function, Variance Decompositions unitary GARCH (p,q) model and multivariate GARCH(1,1) model.Theory of Cointegration is effective to deal with non-stationary series. It resolves the problems of correlations among some I (d) series, where d is the order of integration.Granger Causality Method points out that if event Y is the cause of event X, then event Y can precede event X. The expression of"x Granger-caused y"doesn't means that y is the effect or result of x. Granger causality measures precedence and information content but does not by itself indicate causality in the more common use of the term.Non-stationary variables, which can be co-integrated, can be used to set up Vector Error Correction models. We introduce the theories of VAR and VECM. Time series models consider only stationary variables and classical econometrics ignores the false regression problem. VECM adopts the advantages of the two means and overwhelm the disadvantages. VECM sets up a model which involves both the long-term correlation and short-term adjustment.Cointegration vector and the coefficients of VECM can be estimated through Cointegration test and maximum likelihood method.Impulse response function describes the effect of one inner variable's impulse to other variable. And Variance Decompositions estimate the importance in different structure impulse by analyzing the contribution every structure impulse to inner variable's change.The class of GARCH models is suited to measure the volatility and correlation of financial time series. Especially, it can perfect measure the"fat-tail,volatility cluster and long memory". The correlation of multivariate GARCH (1, 1) model returns is the correlation of volatilities between two series.3. Empirical studyThis section is the most important part of the paper.First, through unit root test we confirm the order of integration of a series. The results show that all the series are I (1) and the corresponding return series are stationary series.Second, by Johansen Cointegration test and VECM, we find that stock indices has only one cointegration correlation, this means they have the same trend. On the basis of cointegration test, we set up VECM of stock indices. The cointegration equation and VECM equation of corresponding return are given.The cointegration relations among the stock indices show that their short-term movements are determined by the same stochastic item, and this item has the long-term affect on the six variables. Though every country's structure of economy and politics is different, but the stock indices can be integrated.From the point of view of cointegration equation, DJIA, FTSE100, HSI and SZ move in the same direction, and they have the same long-term trend. DJIA and NIK, DJIA and SS move in the opposite direction. DJIA and NIK move in the opposite direction which means that they have the opposite trend. From the degree of effect, effect of FTSE100 to DJIA is the biggest, and is positive. NIK has the opposite effect to DJIA, this is consistent to the difference between USA's and Japan's economy. SS has the opposite effect to DJIA because of the problem of trade surplus. This caused a lot of capital into China from USA, so it has the negative effect to US economy. SS has the opposite effect to SZ, because Shanghai stock market is mainly composed by big enterprises owned by government, but Shenzhen stock market is mainly composed by small enterprise. Recently, we reform the big enterprise, so the small enterprises can't develop quickly.Short-term volatilities in Stock indices are not the same in the VECM. Japan, Hong Kong, Shanghai and Shenzhen are affected significantly by the long-term balanced correlation, but USA and UK are not, because USA and UK have the leading status in the whole world.We test the daily stock index by Granger Causality. It is sensitive to the order of lag. We take the six series as one system to deal with the problem. So we set up the VAR model, through the Akaike Info Criterion (AIC) and Schwarz Criterion (SC), we choose the order of 7.The results of Granger Causality test show that there are significant Granger Causality correlations. DJIA has significant Granger Causality effect on other stock markets'indices except SZ, and this is consistent with the judgement of US stock market's leading role. Different stock markets correlate and affect each other. Every market's trend and volatility can be conducted to other markets, and this make four markets the characteristic of"co-movement".We use the impulse response function to analyze the time lag of effect from one stock market to another, we got: the return change of one stock market will supply the information to another stock market's trend in the future.Variance Decompositions give us the result: DJIA has 99% distribution to itself, other stock market has tiny shock to it, especially for SS and SZ. FTSE, NIK, HSI have great distribution to themselves, and DJIA also has distribution to them. For SS, it have great distribution to itself, to 96%, and HSI also have more distribution than others, for SZ, it is mainly effected by SS. SS has much more distribution to it than SZ to itself.At last, we use multivariable ARCH class models to test the effect of GARCH. Through multivariable GARCH, we estimate the volatility correlation between return series. We got: Except SS and SZ, other stock markets have strong volatility correlation, in which, New York and Landon have the strongest volatility correlation, these stock markets have certain volatility correlation to SS and SZ, but the correlation is more weak, and the volatility correlation between SS and SZ is very strong. There is some influence between our stock market and other major stock market of the world, but because stock market of China is new, not develops well enough, their influence is not strong, we have to work hard to improve it.