Modelling Federal Reserve Discount Policy

Christopher F. Baum
Boston College
Baum@bc.edu

Meral Karasulu
Bosphorus University
Karasulu@boun.edu.tr

Abstract

During the past three decades, the Federal Reserve has altered its reliance on alternative tools of monetary policy on several occasions. The efficacy of the discount rate, the Fed's original policy tool, has varied throughout monetary regimes, but has remained an important instrument in the current policy environment. Changes in the discount rate influence market participants' expectations formation, and, under some circumstances, directly affect commercial banks' behavior. Although an extensive literature on the interactions of monetary policy instruments has developed, the particular characteristics of discount policy warrant analysis with current econometric methodologies.

In this paper, we explicitly take account of the infrequent nature of Federal Reserve discount policy actions and the asymmetric effects that those actions have on the market for bank reserves. A band threshold autoregressive (Band-TAR; Tong, 1990) model is used to capture these characteristics, and the technique of threshold cointegration (TCI; Balke and Fomby, 1994) is used to model the joint evolution of the discount and Federal funds rates in the post-October 1979 policy regimes.

The post-October 1979 period, which has been characterized by reserve targeting behavior of the Fed, implies a cointegrating relationship between the discount and Federal funds rates. However, standard linear cointegration techniques are not designed to capture the discrete adjustments to which the discount rate has been subjected. The Band-TAR TCI model captures these discrete adjustments by modelling the equilibrium error of the cointegrating relationship. Intuitively, this suggests a long-run relationship between these two rates which is inactive inside a band implying no adjustment either in the discount rate or the instruments directly affecting the Federal funds rate for observations within this band. However, once the system deviates from the long-run equilibrium beyond the limits of this band, the cointegration relationship drives it back inside of the band. The band limits in a Band-TAR model, specified as lagged values of the equilibrium error, are estimated as part of the model. The estimated lag parameter indicates the lag in the reaction of the Fed to deviations of the discount and Federal funds rates from the equilibrium relationship.

References

1

Balke, N. and Fomby, T. 1994. Threshold cointegration. Unpublished working paper, Southern Methodist University.

2

Tong, H., 1990. Nonlinear Time Series: a Dynamical System Approach. Clarendon Press, Oxford.