| Stock market state is a comprehensive economic indicator.Accurate identification and prediction of stock market state are very significant in financial decision-making and economic analysis.In order to identify and predict bull and bear market states of China's stock market,this paper adopts the rule-based method,which first uses the algorithm of Pagan and Sossounov(2003)to identify,then establishes Markovian logit model and predicts stock market state probability.For identification,fixed-window-period identification of Shanghai composite index is achieved according to the setting rules of Pagan and Sossounov(2003)algorithm,and we statistically analyze number,duration characteristics,return mean and standard deviation of the identified bull and bear market phases.In prediction part,Markovian logit model is established.The identified bull-bear state sequence and a series of predictors are substituted into the model to obtain estimated parameters,so the conditional probability of being bull market state of next time point can be estimated.The data is expanded every one week to predict stock market state at every point from 2010 to 2018.To evaluate the role of predictors in predicting stock market states,this paper studies predictions with and without predictors respectively.To evaluate prediction effect of the rule-based method,this paper compares prediction results with fixed-window-period identification results,and measures prediction accuracy by calculating indicators such as bull correctness and bear correctness.The results show that the rule-based method performs well in identification,while there exists lag effects in prediction.Through model fitness test and robustness test analysis,we find that identification results are robust while prediction results are relatively sensitive to parameters used in the dating rules,the faster for dating rules to capture market transition,the better the rule-based method will be in prediction. |
PDF Full Text Request