Forecasting The Direction Of Stock Markets Based On Time-Varying Coefficient Probit Models

"The direction prediction of stock markets" has been a hot topic in academic fields.In this paper,we consider the predictive ability of the binary dependent time-varying probit model in predicting the direction of monthly excess stock returns.Then we analyze the change trend and the reasons of the time-varying coefficients in the sample period.This paper starts from forecasting the recession obtained from the model for binary recession indicator and the CPI obtained from the ARMA model,then we select the four-month-ahead recession and CPI forecast as optimal explanatory variable.Then,we estimate probit models based on time-varying coefficients,which appears that the recession and CPI forecast are the most useful predictive variables,and once they are employed respectively in the US and Chinese markets,the sign of the excess return is predictable.Compared with the traditional results,the time-varying model can successfully capture the relationship between explanatory variables and the return;also solve the problem of "the direction prediction of stock market".Meanwhile,the results can accurately predict the “bear” and“bull” states of the further stock market,which help investors avoid risk by reasonable investment.