| Once a stock market crisis occurs,the investors are destined to suffer heavy losses of wealth,and the stability of a country’s financial market will also be greatly damaged.Therefore,finding a timely and effective warning method of stock market crises is of great importance to investors as well as to financial regulators.This paper which inspired by the serious Chinese stock market crisis happened in June of 2015 focuses on the early warning of stock market crises.As an extreme condition of the stock market with violent market price fluctuation,a crisis is also defined as a financial anomaly that can not be explained merely by traditional financial theories.And investor sentiment,at the same time,can exert influence on asset pricing to some degree,which has already been a consensus reached by many researchers.So this paper choose to study the early warning of security market crises based on investor sentiment.First of all,in order to guarantee comprehensiveness and effectiveness of the proxy index of investor sentiment,all the original single index go through rigorous screening process and the five index,i.e.turnover rate of securities,price earning ratio,price to book ratio,new high-new low index and balance of financing turn out to meet the screening assessment requirements.Then the principal component analysis（PCA）is used to get a comprehensive proxy index of sentiment.One thing worth being pointed out about this paper is that it emphasizes timeliness of the sentiment index and warning method,all the data adopted are on a daily basis.And the target sample interval is from January 4th 2011 to June 30th 2017,with 1577 samples.The empirical study shows that the investor sentiment proxy index SENT is quite effective because it has a trend presents a highly consistency with that of CSI 300 index.In addition,the correlation coefficient is 0.878 and it passes the test of significance with the level of significance equals 0.05.Secondly,this paper extend Ensemble Empirical Mode Decomposition（EEMD）to the analysis of CSI 300 index time series and the constructed investor sentiment index（SENT）time series.The two original time series are decomposed into 9 independent intrinsic modes（IMF）and a residual each.After that,the intrinsic modes are reconstructed according to their fluctuating characteristics,with the first to the fifth IMF composed to get a new data series representing the short-term fluctuating process,the sixth to the ninth IMF composed to get another new data series representing the mid-term fluctuating process,and the residual data series left unchanged in represent of the long-term trend.Next,the cross correlogram analysis is involved to dig the relationship of SENT and CSI 300 index in different time ranges which provides good evidence for the fellow-up empirical study.The conclusions drawn from the empirical study are as follows:While SENT and CSI 300 index are positively correlated in all of the 3 time ranges,in different time ranges,we can observe different lead-lag relations between SENT and CSI 300 index.In short-term,there exists no obvious lead-lag relationship between SENT and this market price index.However,we can observe that SENT evidently leads the CSI 300 index in both mid-term and long-term,with a larger correlation coefficient in mid-term.Lastly,logistic regression model which is a fairly classical model for crisis warning is used to get the final conclusion.The crises warning prediction index is dependent variable and the investor sentiment index SENTt-1 is the independent variable.The paper concludes that the variable SENTt-1 is a significant predictor for stock market crisis under the 0.05 significance level and the regression coefficient is 13.372,which means the contribution to the occurrence of the stock market crisis made by investors’highly positive sentiment is exp（13.372）times as much as investors’negative sentiment does.In other words,we may impute the eruption of the stock market crisis mostly to investors’positive sentiment.In addition,the model proves to pass the test of robustness. |
