A multidimensional Bayesian model to test the impact of investor sentiment on equity premium

Prior research hasn't provided convincing empirical support for two crucial issues: how investor sentiment can be measured, and can it be used to forecast stock market performance? To bridge this divide, we present a novel empirical technique based on investor sentiment and Bayesian vector autoregressive (BVAR) models to estimate stock returns. Sentiment has always had a substantial impact on the determination of stock prices in US financial markets, even during the subprime crisis. Impulse responses show that a positive shock to investor sentiment pushes equity returns above their equilibrium positions, with the maximum deviation being achieved around two months after the trigger event followed by a return to equilibrium after 6 months. Compared to BVAR models with priors commonly used in the literature, our model enables a very efficient restriction of the coefficients to be estimated without reducing the explanatory power of the model. Moreover, it provides better forecasting of stock returns than other BVAR models beyond 3 months.