The existing empirical literature on Taylor-type interest rate rules has failed to achieve a robust consensus. Indeed, the relatively common finding that the Taylor principle does not hold has fueled a degree of controversy in the field. We attribute these mixed estimation results to a raft of empirical issues from which many existing studies suffer, including bias, inconsistency, endogeneity and a failure to adequately account for the combination of persistent and stationary variables. We propose a new method of combining I(0) and I(1) series in a system setting based on the long-run structural approach of Garratt, Lee, Pesaran and Shin (2006). The application of this method to a long sample of US data provides modest support for the operation of a Taylor-type rule, albeit with considerable inertia. We argue that estimation across rolling windows may better reflect shifts in the underlying preferences of the monetary policymakers at the Federal Reserve. Such rolling estimation provides substantial evidence that the inflation and output preferences of the Fed have varied through time, presumably reflecting the prevailing economic and political conditions, its chairmanship, and the composition of the Federal Open Market Committee. Our most significant finding is that the Taylor Principle was robustly upheld under Volcker, often upheld pre-Volcker but rarely observed post-Volcker over any horizon.
System Estimation with Mixed I(0) and I(1) Variables Long-Run Structural Modelling Rolling Estimation Taylor Rule