A Bayesian Research Of Chinese Stock Market In Structure Mutation

The volatility of financial time series has always been a hot issue economists study. For the modeling of financial time series analysis of volatility endless, but these studies put more focus placed on model building, ignoring the time sequence if subject to significant economic impact time, will make it deviate from the original trend, producing structural break this problem. For the existence of financial time series structure transformation point of great significance for the government to develop appropriate economic policies and address the impact of economic conditions on the stock market burst generated.This paper describes the full article, background and significance of background, research status of the stock market first structural break point system were reviewed, discussed the current situation sub-detection methods structural break point, pointed CUSUM, ICSS such as lack of algorithms. After the proposed research ideas and describes the advantages of the Bayesian approach. Then based on the study Inclan (1993) and others, taking ARMA model-based model, with the Shanghai Composite Index for the study, collect and collate adequate prior information and using Gibbs sampling method, China's stock market structure point mutation Bayesian analysis, according to the detected mutations after point, the data was segmented GARCH modeling analysis. The final results showed that (1) the existence of significant structural breaks in the stock market in our country, and successfully detected the location of the three structural break points, which correspond to our major structural changes in the stock market and in the mutant before and after the point of occurrence are also indeed have a significant economic event. (2) According to the test results, the Shanghai Composite Index return series is divided into four sub-sample sequences were segmented GARCH model for the full sample and sub-samples were analyzed analysis, empirical results show that segmentation model is better than the whole sample Modeling results.