A time-varying threshold STAR model with applications

Abstract Smooth-transition autoregressive (STAR) models, competitors of Markov-switching are limited by an assumed time-invariant threshold level. We augment the STAR model with a time-varying that can be interpreted as ‘tipping level’ where mean and dynamics VAR shift. Thus, latent level serves demarcation between regimes. show how to estimate in Bayesian framework using Metropolis step unscented Kalman filter proposal. To allowing time variation affect results, we present two applications: natural rate unemployment regime-dependent government spending.