New Evidence on Targets for Monetary Policy

Abstract

Monetary policy involves Federal Reserve's use of select policy instruments to influence general economic conditions. Of major policy instruments (required reserve ratios, discount rate, and open market operations), open market operations are commonly manipulated to affect economic conditions. Open market purchases and sales are employed to achieve operational targets (i.e., levels of free reserves and/or monetary base). The operational targets influence market interest rates, monetary aggregates, and/or credit aggregates, so-called indicators or intermediate targets of monetary policy. The intermediate targets then impact upon general economic policy goals, (e.g., real growth, GNP, aggregate price level). Fellner [8] and Davidson-Hafer [6] argue that a variable is as an intermediate if it is closely related to objectives considered of substantial importance, is causally prior to such objectives or goals, and follows changes in policy instruments in a predictable manner (i.e., it is controllable). The appropriate choice of a variable as the intermediate target is source of great debate [20; 19; 1; 16]. This debate involves ability of Federal Reserve to control variable as well as relation of with goal variables. Time-series techniques are now utilized to examine both aspects of this debate. Timeseries examinations of control problem appear in Bomhoff [4], Johannes-Rasche [15], and Davidson-Hafer [6]. Time-series examinations of target-goal relationship are presented in Davidson-Hafer [6], Friedman [10], and McMillin-Fackler [18]. With exception of vector autoregressive (VAR) procedures reported in Friedman [10] and McMillin-Fackler [18], all researchers employ single-equation procedures. Bivariate techniques fail to permit possible feedback from a wide array of sources, many of which are believed to be important. This research employs a multiple autoregressive moving average (MARMA) technique to examine appropriateness of several intermediate candidates. The MARMA procedure is more general than, and hence generally preferred to, VAR or univariate procedures (see section III). Equally important, however, is our simultaneous examination of