Stock Market, Investment and Sentiment in the Framework of Bayesian DSGE Models

Abstract

We investigate the interactions among consumer preference, firm investment, stock market activity, investor sentiment and monetary policy in Bayesian Dynamic Stochastic General Equilibrium (DSGE) models for the U.S. economy. We design a framework in which household turnovers, firm turnovers, equity risk premiums, investment, preference and sentiment jointly influence stock price misalignments and macroeconomic fluctuations. These are not only due to households’ interactions with the stock market through financial wealth, consumer preference and investor sentiment, but also induced by firms’ interactions with the stock market through financial resources, firm investment and equity risk premiums. Our objectives are fivefold. We disentangle between stock price fluctuations induced by risk premiums and by animal spirits. We model risk premiums using the financial shock and animal spirits using the sentiment shock. We identify influence channels of the real economy on stock price fluctuations. We investigate propagation mechanisms of stock price fluctuations to the real economy and evaluate monetary policy responses to stock price misalignments. Our methodologies include Bayesian estimation, historical shock decomposition, impulse response analysis, forecast error variance decomposition and Bayesian model comparison. Our main findings are fourfold. Equity risk premiums, sentiment, investment and preference make substantial contributions to explaining stock price fluctuations, consumer sentiment variations, investment fluctuations, output fluctuations and inflation variations. Financial, sentiment, investment and preference shocks propagate through stock market index pricing rule, stock market bubble evolution, intertemporal substitution of investment and intertemporal substitution of consumption respectively. Higher household turnover rate increases stock market wealth effect and aggregate demand, whereas higher firm turnover rate contaminates stock market wealth effect and financial shocks’ impacts. Monetary policy responds counteractively and significantly to financial slack at business cycle frequency. We then combine the artificial data generated by the DSGE model and the actual data originating from the unrestricted VAR model to formulate DSGE prior for Bayesian VAR (BVAR) model. Furthermore, we apply the DSGE model implied cross-equation restrictions to BVAR model and generate DSGE-VAR model in two forms. One form is combination of DSGE model implied prior mean and Normal-Inverse Wishart prior, and we define it as DSGE-VAR model with N-IW prior. The other form is combination of DSGE model implied prior mean and Stochastic Search Variable Selection (SSVS) in Mean-Inverse Wishart prior, and we define it as DSGE-VAR model with SSVS in Mean-IW prior. Finally, we estimate and assess relative forecasting performance of BVAR models with three types of priors, which are DSGE-N-IW prior, DSGE-SSVS in Mean-IW prior and Minnesota Prior. We find DSGE-VAR model with SSVS in Mean-IW prior has identical forecasting performance of BVAR model with Minnesota prior. Therefore, we have demonstrated that the DSGE model does not incur serious model misspecification problem.