A Study On Chinese Stock Market Bubble State And Crash Early Warning

Bubbles refer to the explosive growth of asset prices,the crashes are rapidly decline of stocks or stock index after the bubble bursts.There is no warning signal for the burst of bubbles,and the crash will have a huge impact on the economy.From the perspective of global financial events,the occurrence and frequency of asset price bubbles has increased rather than decreased.A series of asset prices plummeted in a short period of time causing a sharp oscillation in the stock market,which in turn damaged the asset interests of investors,affected the operating efficiency of the financial market,and disrupted the orderly development of the capital market.Therefore,it is essential for investment trading activities and risk management activities to effectively identify bubbles in the stock market and quantify the risks of bubbles in financial activities.The CSI 300 Index is one of the most representative indices of China's mainland A-Shares.This paper uses the closing price of the CSI 300 Index from January 4,2002 to November 22,2019 as the research object.Based on the Log-periodic Power Law Model(LPPL)to identify the bubble state of the stock market and analyze the bubble state transition,we use the "rolling window" and "fixed starting point and moving end point" to resample the sample,then make out-of-sample prediction daily to calculate the critical point of the bubble bursts and construct the dynamic confidence interval.On this basis,a risk level early warning signal is constructed to effectively quantify the bubble risk.First,the BDS method is used to test the non-linear characteristics of the CSI 300 Index,and the R/S analysis method is used to test its fractal characteristics to determine whether the CSI 300 Index is a non-linear market with fractal characteristics.Then,the wavelet transform modulus maxima(WTMM)was used to identify the singularity of the wave in the sequence,the sample windows were divided by the singularity point,the LPPL model was used to identify the state of bubbles and analyze the transition in each window.There were 3 bubbles,2 anti-bubbles,and 2 negative bubbles in the CSI 300 Index,and it was found that the crashes event was accompanied by the transition of the bubbles state.In order to effectively avoid the huge loss caused by the transition of the bubbles state,this paper constructs an LPPL dynamic confidence interval from an empirical perspective to track the most possible region of the bubbles burst dynamically.On this basis,we use the time difference between the sample endpoint and the dynamic confidence interval as a standard to discriminate the strength of the bubbles burst signal and set the risk level warning signal.The results show that the dynamic confidence interval can overcome the randomness of the LPPL critical time point forecasting work as well as display the risk level of the trajectory of the critical interval of the stock market bubble.In the same time,the risk level early warning signal effectively quantifies the intensity of the bubble risk faced by investors in trade activities.