| Investing in any financial product is a collection of risks and returns,and stocks are the representative of them.Investors always hope to obtain large returns with low risk,and scholars also hope to build better asset pricing models through stock volatility and returns.Therefore,forecasting stock returns and volatility has long been an area of concern for stock market participants and academics.At present,there are a large number of literatures in academia showing that stock market returns and volatility are predictable in-sample,but the research on the predictability out-of-sample is not very sufficient.In addition,although there are some methods in the literature that can predict future stock information to a certain extent,their prediction accuracy is low,which is not conducive to investors and industry people making decisions and pricing.This paper provides new evidence for its out-of-sample predictability by predicting stock returns and volatility through two types of investor sentiment,respectively.In this paper,the monthly data of investor fear Index(VIX)are characterized by Japanese candlestick,and the ULD is obtained by subtraction of the upper line of the candlestick.Then,this new index of investor sentiment prediction is used to forecast stock returns.Secondly,a newly proposed index of time-varying risk aversion is used as an explanatory variable to predict stock volatility.The empirical results in-sample show that ULD has a significant negative predictive effect on stock returns.Furthermore,ULD also has the best predictive power compared to macroeconomic variables often used in the literature.In terms of out-of-sample forecasting,adding ULD exogenous variables to the model can not only greatly improve the forecasting accuracy of stock returns,but also use its forecasted values for investment portfolios to obtain considerable economic benefits.On the other hand,all bivariate regression models that include ULD exogenous variables,whether in-sample or out-of-sample,have higher prediction accuracy than univariate regression models.Finally,after implementing a seies of robustness tests,including multiple portfolio models,business cycle,alternative portfolio prerequisites,rolling window estimates,different out-of-sample forecast windows,and alternative shading proxies,all robustness tests show that ULD is a powerful predictor,which can significantly improve the predictability of stock returns.This paper introduces time-varying risk aversion index exogenous variables into autoregressive models to predict stock volatility.In-sample analysis found that the time-varying risk aversion index had a significant positive impact on stock return volatility.Out-of-sample forecasts show that adding time-varying risk aversion to benchmark autoregressive models can produce more accurate forecasts of stock volatility.Our conclusions are robust when using autoregressive models with different lag orders and different forecast evaluation windows.Finally,this paper studies the relationship between the predictive ability of time-varying risk aversion and the volatility of other stock indices and two crude oil,and in most cases timevarying risk aversion can still improve the prediction accuracy of corresponding stock indices and crude oil volatility. |
