Markov-Switching Common Dynamic Factor Model with Mixed-Frequency Data

In this paper, we consider a coincident economic indicator model with regime-switching dynamics with the time series observed at different frequencies, for instance, at monthly and quarterly frequencies. Until now the only situation was to drop the lower frequency series and to estimate the model based only on the higher frequency series. This approach leads to the significant information losses. We propose an approach allowing to overcome this problem and to estimate a nonlinear dynamic common factor with the missing observations taking advantage of all the information available.