Topic: Contributed on Regime change modeling in economics and finance I Title: The stock market and real economy: A Bayesian nonparametric approach Authors:  Qiao Yang - ShanghaiTech University (China) [presenting]
Abstract: The study of the joint dynamics between the stock market and real economy has a long history. A bivariate infinite hidden Markov model (IHMM) is proposed to investigate the joint dynamics of monthly SP500 returns and industrial production growth rates from 1926 to 2014. The bivariate IHMM is a nonparametric Bayesian approach which extends the vector autoregression (VAR) model to include Markov switching of infinite dimension. Comparing to conventional approaches, the bivariate IHMM is able to accommodate multivariate modelling, regime switching as well as structural changes in unified framework. The bivariate IHMM shows significant improvements in out-of-sample density forecasts as well as evidence of capturing structural change after 1984. In addition, we find past stock returns only have significant relation to future economic growth rates when regime-dependence is allowed. We show both jointly modelling and regime-dependence are two essential building blocks for fully capturing the joint dynamics of stock returns and real growth rates.