The Research Of Realized Volatility For China’s Stock Market Based On Heterogeneous Market Hypothesis

With the development of global economic integration and operation mechanism of the stock market, China’s stock market has become an important part of the global financial market. In the study of financial markets, the volatility has been a hot topic for years and after a lot of researches, scholars got three mainstream methods for solving volatility:the first one is the implied volatility according to the Black-Scholes equation; the second one is the historical volatility rate such as ARCH and SV Models; the last one is realized volatility based on the study of high-frequency data, which is the object of this paper, with the high-frequency data can be more convenient obtained, the ARCH and GARCH models can no longer meet the needs of the high-frequency data research. Compared to model volatility, realized volatility can be more direct and accurate description of the characteristics of the volatility. Financial research literature suggests that in the GARCH and SV models, daily returns as low frequency data will lose a lot of useful information which investors expect, while the inday high-frequency data can be made up for this shortcoming. Data sampling frequency is higher, we can obtain more information. Calculation of realized volatility by using high frequency data does not require complicated parameter estimation. In addition, the realized volatility is a nonparametric method and there are not estimation problem, so never brought the curse of dimensionality of modeling method, is a new kind of financial volatility measurement method.The theory of realized volatility includes Muller et al.(1993) theory of heterogeneous market hypothesis, Peters (1994) Fractal Market Hypothesis, Lux and Marchesi (1999) mixed market theory, based on the theory of heterogeneous market hypothesis and considering market microstructure noise and jumps, we use HAR-RV model and other related models(HAR-RV-GARCH,HAR-RV-J,HAR-RV-CJ) to estimate realized volatility in China’s stock market, and analysis to study impact of the short-term, medium-term and long-term types of transactions on the stock market volatility. We selected59,816data of the Shanghai Composite Index from April8,2005to April22,2010, and calculated1227realized volatility data; January4,2005to2011May23Shenzhen composite total77424, calculated the1613realized volatility as research data, they are all5minutes high frequency closing price data. Based on secondary power deteriorated theoretical, jump component is separated from volatility.We estimate short, medium and long term realized volatility by HAR-RV-J and HAR-RV-CJ model. This thesis’s empirical evidence suggests that short-term type of investment has the greatest impact on the day realized volatility, while the medium-and long-term type of transaction’s marginal contribution rate is low, but little difference. The three explanatory variables coefficients are significant, which indicates the presence of heterogeneity in China’s stock market, and validates the theory of heterogeneous market hypothesis that the short-term realized volatility is influenced by their previous value, and also by the medium-and long-term of the transaction, and for long-term traders,mainly by its own pre-value. This is similar to the conclusions of ’based HAR model realized volatility of the Chinese stock market ’by xiaolei Yu.After the least squares regression of HAR-RV model, we analyzed its residuals, and found the sequence of clustering phenomenon, so we examined the ARCH effect of the error term, and added GARCH items, by HAR-RV-GARCH model, further studied the influence of the three types of transactions on realized volatility. HAR-RV-GARCH model empirical evidence shows that, after adding GARCH item, there is little difference compared with the original HAR-RV model which verify the stability of the impact that the three types of transactions on the day realized volatility. Based on the theory of quadratic power variation, realized volatility can be decomposed continuous sample path variance and discrete jumps variance, considering the impact of market micro structure noise on realized volatility and in order to reduce the impact of market microstructure, we use the product of the absolute value of the staggered yield (original second power variation formula is the product of the absolute value of adjacent yield) to fix secondary power variation calculation. Then we can obtain the other jump variance of the sequences which considered market microstructure through the correction Z-statistics. At the confidence level of0.01, we obtained significant arrival jump of308times by the correction of the Z statistics inspection, compared with the previous276times,there are more32times, the capture rate also from22.49%in the pre-revised enhanced to25.1%, therefore, modified Z statistic test jump deteriorate much more accurate sequence content. According to Huang and Tanchen (2005) study, we first through HAR-RV-J model to get the results that jump items has a significant negative impact of realized volatility. Compared to HAR-RV model, we can know that HAR-RV-J and HAR-RV-CJ model which considering the jumps have better fitting results and forecasting performance.This innovation of this paper is that after fully considering the market micro-structure noise, we quantified the influence of jumps ingredients on different realized volatilities, used the modified Z statistics to detect the significant jumps effectively, improved the volatility capture, which are all better for us to analyze the impacts of factors to realized volatility.