Business Cycle Turning Points, A New Coincident Index, and Tests of Duration Dependence Based on a Dynamic Factor Model With Regime Switching

摘要: The synthesis of the dynamic factor model Stock and Watson (1989) regime-switching Hamilton proposed by Diebold Rudebusch (1996) potentially encompasses both features business cycle identified Burns Mitchell (1946): (1) comovement among economic variables through (2) nonlinearity in its evolution. However, maximum-likelihood estimation has required approximation. Recent advances multimove Gibbs sampling methodology open way to approximation-free inference such non-Gaussian, nonlinear models. This paper estimates for U.S. data attempts address three questions: Are empirically relevant? Might implied new index coincident indicators be a useful one practice? Do resulting regime switches show evidence duration dependence? answers all would appear yes.