We investigate the probability forecasting performance of a three-regime dynamic ordered probit model framework suitable to forecast recessions, low growth periods and accelerations for the U.S. and Japan. In a first step, we apply a non-parametric dating algorithm for the identification of...