Asymmetric Dynamic Factor Model

Abstract

We develop an asymmetric dynamic factor model that allows the impact of latent factors to vary depending on whether the factors are larger or smaller than their thresholds. An efficient Monte Carlo Markov Chain sampler is proposed to sample from the non-standard posterior distribution. The model is illustrated through an application to uncover the asymmetric impacts of economic conditions on macroeconomic variables. The results show that the asymmetries of the impact are significant. In particular, the asymmetric dynamic factor model suggests that the positive marginal impact of good economic conditions is smaller than its negative counterpart.