Identifying business cycle turning points with sequential Monte Carlo methods: an online and real‐time application to the Euro area

We propose a new approach for detecting turning points and forecasting the level of economic activity in the business cycle. We make use of coincident indicators and of nonlinear and non‐Gaussian latent variable models. We thus combine the ability of nonlinear models to capture the asymmetric features of the business cycle with information on the current state of the economy provided by coincident indicators. Our approach relies upon sequential Monte Carlo filtering techniques applied to time‐nonhomogenous Markov‐switching models. The transition probabilities are driven by a beta‐distributed stochastic component and by a set of exogenous variables. We illustrate, in a full Bayesian and online context, the effectiveness of the methodology. We also measure its ability to identify turning points and to forecast the European business cycle on both realtime and last‐revised data