Time-varying response of monetary policy to macroeconomic conditions.

• Author: Shen, Chung-Hua

1. Introduction

   A reaction function empirically relates a monetary policy indicator (often the federal funds rate or a monetary aggregate) to the macroeconomic objectives of monetary policy (usually national output, inflation, and unemployment). Assuming that the underlying macroeconomy is stable, estimated coefficients from a reaction function reveal information about the monetary authority's response to macroeconomic conditions (Chappell, Havrilesky, and McGregor 1993).

   Beyond the assumption described above, most reaction function studies further assume that the Federal Reserve's response function is linear in the sense that it contains fixed coefficients (see the survey by Barth, Sickles, and Weist 1982, and, more recently, Chappell, Havrilesky, and McGregor 1993); that is, most studies implicitly assume that the Fed responds to a change in inflation, for example, with the same vigor regardless of whether the level of inflation is unusually high or low and regardless of the timing of the response with respect to the business cycle, elections, and so on.

   Although it is generally recognized that a fixed-coefficient estimate of the Fed's reaction function overly homogenizes the Fed's actual response to macroeconomic conditions, solutions to the problem have been ad hoc. Studies that seek to expose a political business cycle impose structural break-points on the Fed's reaction function based on pre- and postelections periods and then estimate a fixed-coefficient reaction function within each subperiod (Hakes 1988a; Chappell, Havrilesky, and McGregor 1993; Gamber and Hakes 1997). Other studies impose structural break-points based on political or policy regimes and then estimate a fixed-coefficient model within each regime (Havrilesky, Sapp, and Schweitzer 1975; Potts and Luckett 1978; Hakes 1988b, 1990; Hakes and Gamber 1992).

   In this short paper, we test the hypothesis that the Fed's reaction to changes in output, prices, and unemployment over the post-Accord period of 1953-1994 can be accurately represented with a fixed-coefficient model. We reject the hypothesis. Because of this finding, we model the time-varying response of the Fed to the macroeconomy with a Kalman filter A Kalman filter allows us to map the evolution of each reaction function coefficient over time.

   This paper is organized as follows. In section 2, we propose a reaction function for estimating the Fed's response to the macroeconomy. In section 3, we describe the application of a Kalman filter to our model and report the results of the estimations. Section 4 contains some concluding remarks.

2. Model

   We model the Fed's response to the macroeconomy with the following discrete choice model:

   [Mathematical Expression Omitted], (1)

   where

   P = policy indicator (0 = tight money, 1 = easy money),

   U = civilian unemployment rate (unemployment objective),

   PPI = producer price index (inflation objective),

   IP = industrial production index (growth objective),

   [Epsilon] = error term,

   and the (+) and (-) denote the expected signs of the coefficients based on the assumption that the Fed engages in countercyclical policy. This particular reaction function has been employed by Hakes (1990), and similar reaction functions have been utilized by Hakes (1988a, b), Wallace and Warner (1984), Luckett and Potts (1980), and Potts and Luckett (1978). Hence, we include only a brief description of the model below.

   The dependent and independent variables in the reaction function are the policy indicator and policy objectives, respectively. We employ a dummy variable representing tight money and easy money as the policy indicator This dichotomous series was first generated by Potts and Luckett (1978), who classified monetary policy intentions as tight or easy from their reading of the Record of Policy Actions of the Federal Open Market Committee (FOMC) for the period 1956-1976. This series was extended by Wallace and Warner (1984) and by Hakes (1990) to include the period of 1953-1987. We extend the series to include the entire post-Accord period of 1953-1994.(1)

   We choose a narrative policy indicator for a number of reasons. First, Boschen and Mills (1995) show that there is a greater persistent relationship between any of the many discrete narrative policy indicators and M2 than there is between money market indicators (such as the federal funds rate) and M2. This result suggests that interest rate indicators may contain greater nonpolicy shocks than narrative indicators. Second, Boschen and Mills report that there is great similarity across all narrative indicators, which suggests that narrative indicators lack subjectivity. Finally, unlike monetary aggregates or money market indicators, a discrete choice indicator is independent of the operating procedures employed by the Fed. Thus, it may be a superior indicator of Fed behavior over lengthy time periods that are certain to overlap many different operating procedures.

   Table 1. Results of Probit Estimates Independent Variable Coefficient t-ratio Constant -0.13(*) -1.95 U 1.11(**) 3.17 PPI -17.28(*) -2.03 IP -13.51(*) -1.92 * Significant at the 0.05 level. ** Significant at the 0.01 level. The independent variables in the reaction function are proxies for the ultimate objectives of monetary policy as established by the employment act of 1946 - high employment, stable prices, and stable growth. We employ the civilian unemployment rate, the producer price index, and the industrial production index from the CITIBASE data set. Consistent with...