Get access

Dynamic Limited Dependent Variable Modeling and U.S. Monetary Policy


  • I am indebted to James D. Hamilton for his advice and encouragement, and for many useful discussions. I am also grateful to Graham Elliott and to Allan Timmermann for their feedback and for useful discussions. I thank Kenneth D. West, three anonymous referees, Fabio Busetti, Carlos Capistrán, Takeo Hoshi, Bruce Lehmann, Valerie Ramey, Yixiao Sun, and seminar participants at UCSD, HEC Montréal, the Board of Governors of the Federal Reserve System, SUNY at Albany, SUNY at Binghamton, the 11th Conference of the Society of Computational Economics, the Seminar for Bayesian Inference in Econometrics and Statistics (SBIES 2005 conference), and the European Conference of the Econometrics Community (2006 EC2 conference) for useful comments. Finally, I am grateful to Jean Boivin and to Athanasios Orphanides for generously providing me with their real-time data—the unemployment and inflation data for the early years used in this study come from Boivin's data set. Any errors are mine.


I estimate a forward-looking, dynamic, discrete-choice monetary policy reaction function for the U.S. economy that accounts for the fact that there are substantial restrictions in the period-to-period changes of the policy instrument. I find a substantial contrast between the periods before and after Paul Volcker's appointment as Fed chairman in 1979, both in terms of the Fed's response to expected inflation and in terms of its response to the (perceived) output gap. In the pre-Volcker era, the Fed's response to inflation was substantially weaker than in the Volcker–Greenspan era; conversely, the Fed seems to have been more responsive to (inaccurate real-time estimates of) the output gap in the pre-Volcker era than later. These results, which carry through a series of extensions and robustness checks, provide support for the “policy mistakes” hypothesis as an explanation of the stark contrast in U.S. macroeconomic performance between the pre-Volcker and the Volcker–Greenspan periods.