Font Size: a A A

Investigation On Extreme Events And Investors' Behavior Of Chinese Stock Markets

Posted on:2012-01-14Degree:DoctorType:Dissertation
Country:ChinaCandidate:G H MuFull Text:PDF
GTID:1119330332976312Subject:Applied Mathematics
Abstract/Summary:PDF Full Text Request
With the integration of mathematics, physics, finance, computers and the global economic in 1990s, "econophysics" was born at the right moment. Econophysics borrows the concepts, theories and methods from statistical physics, complex system theory, nonlinear science and applied mathematics to study the properties of financial markets. As the biggest emerging mar-ket, some behavior and regularities of Chinese stock market are similar with those in mature Western markets, but also are different due to the level of development, rules of market, struc-ture of market, and differences in behavior of traders. In this thesis, using tick-by-tick data and limit-order book data, we borrow the methods of econophysics to investigate the power-law distribution, temporal correlation, multifractality, dynamics of extreme events and behavior of traders in Chinese stock market.In Chapter 1, we briefly introduce the history and current status of financial markets, stock markets and Chinese stock market. Next, we give a general review of the origin, backgrounds and main research fields of econophysics. Finally, we outline the main content and structure of this thesis.In Chapter 2, we test the universality of the distribution of intraday return and find returns have power-law tailed distribution with tail exponents greater than 2, which are well outside the Levy stable regime for individual stocks. The tail exponents increase with time scales. When the stocks are grouped according to their turnover rates and market capitalization, the returns in each group can be well fitted by the q-Gaussian formula. The tail exponents logarithmically decrease with the turnover rate and increase with the market capitalization, and the market capitalization has a greater impact than the turnover rate. The slopes are significantly different from zero when time scales are 1 min and 5 min. When time scale is 30 min, there is not conclusive evidence showing that the slopes are different from zero. We conclude that the intraday return distributions at small time scales are not universal in Chinese stock markets and might be universal at large time scales.In Chapter 3, we study the distribution of trade sizes and trading volumes, temporal cor-relation and multifractal features in Chinese stock market. We find that traders prefer to place orders with the size being certain numbers. The empirical PDFs of trading volumes at different time scales(clock time ranging from△t = lmin to△t = 240 min, or trade number△n≤8) can be modeled by q-Gamma function. All the empirical PDFs exhibit power-law tails, with tail exponents great than 2. Trading volumes at different time scales possess non-universal long memory, whose Hurst index logarithmically depends on the average trading volume. The mean-variance unveils that the scaling exponent logarithmically depends on the time scale. The trading volume exhibits multifractal nature, which is mainly caused by the long memory and the probability distribution of the trading volume. However, the intraday pattern has negligible impact on the temporal correlations and multifractal nature of trading volumes.In Chapter 4, we study the dynamics around extreme events in Chinese stock market. We find that the cumulative number of aftershocks with magnitude exceeding a given threshold increases as a power law with respect to the time distance to the main shock. The power-law exponent is an increasing function with the volatility threshold. The exponents are significantly larger than 1, not consistent with the results of other markets. After selecting intraday extreme price changes under "relative filter" and "absolute filter", we find price evolution around ex-treme events has overreaction and reversal patterns. For the volatility (absolute returns), trading volume, bid-ask spread and volume imbalance, we find a significant peak around the extreme events, followed by a slow relaxation. It can be well fitted by a power law with relaxation ex-ponents are larger than those of LSE stocks in most cases. Considering on order direction and investor type, we find an increase of volume followed by a slow decrease, which forms a peak around extreme events. The volume peak of buyer initiated execution orders is found to appear earlier and higher than seller initiated executed orders around positive events. For negative events, the situation is just opposite. Moreover, the behavior of partially filled market orders, filled market orders and canceled orders is qualitatively the same:the volume peak of buy (sell) orders of each type appears earlier and higher than sell (buy) orders around positive (negative) events. However, the volume peak of sell (buy) limit orders shows up earlier and higher than buy (sell) orders around positive (negative) events. No qualitative difference has been observed in the volume dynamics around extreme events between individuals and institutions. There is a significant difference between individuals and institutions in the relative rates of different types of orders, which could be explained by the differences of strategies between individual and institutional investors. Moreover institutions are more aggressive and more informed.In Chapter 5, we analyze the relationship between the return and the trading frequency for individual traders and institutional traders respectively and compare the results with the simulations. We find that more trading frequency results in less return for investors in B-share market, while the return of individual investors in the A-share market is independent of the trading frequency. For individual or institutional winners, the return decreases with trading frequency as a power law. We also find that the return of real trading is worse than random trading, i.e. zero intelligence outperforms trading strategies in stock market.In Chapter 6, we study the cross correlation of inventory variation of investors. The mean value of cross-correlation coefficient is greater than 0, and the coefficient between individual investors is greater, while coefficient between individual traders and institutional traders is less than 0. The first and second eigenvalues of cross correlation matrix are outside of the bulk random matrix prediction, indicating that the two eigenvalues may contain special information of all investors or part of investors. The distribution of eigenvector corresponding to the eigen-value is similar to Gaussian distribution, while the distribution of eigenvector corresponding to big eigenvalue deviates significantly from Gaussian distribution, implying big eigenvalue only carries information of part of investors. The projection of inventory variation on the first eigen-vector has a linear relationship with return of stock. And the mean value of slope is 0.5, and the average slope of individual traders is 0.58, while the average slope of institution traders is 0.25. According to the trend in the transaction, investors are grouped into trending investor, reversing investors and uncategorized investors. We find individuals are more like reversing strategy, while uncategorized investors of institution take a higher proportion, indicating that the trading strategies of institutions are more complex, which can not be distinguished by the trend. Granger causality test is performed to investigate the causal relation between inventory variation and return. We find returns are Granger causing inventory variation for the major-ity of trending and reversing investors of individuals, while returns and inventory variation are Granger causing each other for the majority of trending and reversing investors of institutions.
Keywords/Search Tags:Econophysics, Extreme events, Random matrix theory, Behavior of investors
PDF Full Text Request
Related items