Study On The Day-of-the-week Effect From China A Share Stock Market

Since December 19,1990,the Shanghai Stock Exchange has been officially traded in the twists and turns in the vigorous development from the initial eight stocks to the scale of more than 3,000 stocks.With the gradual expansion of the scale of China's securities market,the study on the effectiveness of China's securities market must also keep pace with the times.The week effect of the stock market has long been paid attention to as a kind of financial vision in the financial and economic field.The research on this issue will be of great significance to China's securities market policy-making,trading behavior of securities market and academia's understanding of China's securities market efficiency.Based on the GARCH model and the rolling window regression method,the full-sample and rolling window regression tests are performed on the week effect of the Shanghai Composite Index.By introducing the auto-regressive term and the generalized error distribution,this paper improves the traditional GARCH model and establishes the AR-GARCH-GED model,which leads to the conclusion that the week effect of Shanghai Composite Index is more significant than the previous study.By introducing the rolling window regression method,this paper further investigates whether the week effect under full-sample regression is only a pure opportunistic behavior of data mining.The results of rolling window regression show that in 1059 windows,the significant proportion of week effect is only 13% to 24%.Full-sample regression and rolling window regression in two ways,gradually and step by step in-depth test of the Shanghai Composite Index week effect,thus the conclusion of our study will have a comprehensive and robust result.This paper uses the data of the Shanghai Composite Index from December 19,1990 to September 7,2016 to analyze the week effect of the Shanghai Composite Index.The empirical results show that in the case of full sample regression,the Shanghai Composite Index exists significant week effect.However,once the rolling window regression method is introduced,this significant week effect is not robust and is only a statistical illusion depending on data mining.