Dating and forecasting turning points by Bayesian clustering with dynamic structure: A suggestion with an application to Austrian data

The information contained in a large panel data set is used to date historical turning points of the Austrian business cycle and to forecast future ones. We estimate groups of series with similar time series dynamics and link the groups with a dynamic structure. The dynamic structure identifies a group of leading and a group of coincident series. Robust results across data vintages are obtained when series specific information is incorporated in the design of the prior group probability distribution. The results are consistent with common expectations, in particular the group of leading series includes Austrian confidence indicators and survey data, German survey indicators, some trade data, and, interestingly, the Austrian and the German stock market indices. The forecast evaluation confirms that the Markov switching panel with dynamic structure performs well when compared to other specifications.