Financial and Sentiment Shocks in a Large Bayesian VAR: Evidence from the US, 1970-2015

Abstract

This paper examines, empirically, the historical importance of financial and survey-based sentiment shocks in the US business cycle, and in the formation of US monetary policy over the period 1970-2014. A large Bayesian VAR (BVAR) model with a 'Minnesota Prior' is estimated on monthly US data using analytical and Markov Chain Monte Carlo (MCMC) methods. US inflation and monetary policy are found to be more sensitive to financial shocks, while fluctuations in the unemployment rate are influenced by both financial and sentiment shocks. The role of financial shocks in the unemployment rate cycle, however, is found to be increasingly important from the post-2001 recovery onwards. Results for the post-Lehman period indicate that financial and sentiment shocks had an equal effect on the scale of the employment contraction, and in the loosening of monetary policy in the aftermath of the Lehman collapse.