A regime-switching model for the federal funds rate target

This paper develops an ordered choice model for the federal funds rate target with endogenous switching among three latent regimes and possibly endogenous explanatory variables. Estimated for the Greenspan era (1987-2006), the new model detects recurring switches among three policy regimes (interpreted as loose, neutral and tight policy stances) in response to the state of economy, outperforms the Taylor rule and the existing discrete-choice models both in and out of sample, correctly predicts out of sample 90% of the Fed decisions during the next thirteen years, successfully handles the zero lower bound period by a prolonged switch to a loose policy regime with no-change to the target rate (while the Taylor rule and the conventional ordered probit model predict further cuts), and delivers markedly different inference. The empirical results suggest that the endogeneity of explanatory variables does matter in modelling monetary policy and can distort the inference: the marginal effects on the choice probabilities can differ by several times and even have the opposite signs.