Stock returns and real growth: A Bayesian nonparametric approach

Abstract

Abstract This study constructs a Bayesian nonparametric model to investigate whether stock market returns predict real economic growth. Unlike earlier studies, our use of an infinite hidden Markov model enables parameters to be time-varying across an infinite number of Markov-switching states estimated from data rather than fixed like a prior. Our model exhibits significantly greater accuracy in out-of-sample density forecasts. We uncover strong evidence of the time-varying power of lagged stock returns to predict economic growth.