Big Data Analysis of the Dynamic Relationship between Stock Prices and Business Cycles Via Bayesian Methods

In previous study, we proposed a Bayesian modeling technique to decompose a daily time series of Nikkei Stock Average (NSA) into three components which include a trend component, and analyzed the behavior of each estimated component. It was confirmed that there is time-varying correlation between the trend component and the coincident Composite Index in Japan (CIJ). In this paper, as an extension of the previous study we analyze the dynamic relationship between the trend component in the NSA and the CIJ using a regression model with a time-varying coefficient and a lag parameter. The regression model is constructed using the NSA as the dependent variable and the CIJ as the explanatory variable. Bayesian smoothness prior technique is applied to estimate the time-varying coefficient. Moreover, we explain the dynamic relationship between business cycles and stock prices based on the estimates of the time-varying coefficient and the lag parameter. As an empirical example, we analyze the daily time series of NSA closing values from January 4, 1991, to March 30, 2018, together with the monthly CIJ data over the same period.