The predictability of FOMC Decisions: evidence from the Volcker and Greenspan Chairmanships.

• Author: Lapp, John S.
• Position: Federal Open Market Committee - Federal Reserve Board Official Paul Volcker, and Chairman Alan Greenspan

1. Introduction

   About every six weeks, the Federal Open Market Committee (FOMC) meets to decide on the short-run course of U.S. monetary policy. Before each meeting, the financial press reports extensively on the likely decision and afterward attempts to explain the reasoning behind the decision. (1) Decisions to ease or tighten monetary policy are usually attributed to recently announced information about the state of the economy, such as the latest indicators of real activity and inflation. Moreover, asset price changes after these economic announcements are often interpreted as reflecting the market's expectation of the Federal Reserve's response to such announcements. This paper examines whether there has been a systematic connection between economic announcements and subsequent FOMC decisions that would potentially allow the public to predict the direction of these decisions. The press coverage given to the FOMC meetings suggests that FOMC decisions are considered important and are not highly predictable.

   There has long been a debate about whether policy should be predictable. William Poole (2001) has recently argued,

   The presumption must be that market participants make more efficient decisions--decisions that maximize economic growth by minimizing the wastage of resources from expectational errors--when markets can correctly predict central bank actions. (p. 9)

   On the other hand, to the extent that the central bank wants to surprise economic agents, as in the models following Barro and Gordon (1983), the central bank would prefer its decisions to be largely unpredictable. (2) This would require that policy changes be only loosely tied to the information available to the public.

   The goal of our investigation of the predictability of FOMC decisions is more modest than that of the traditional reaction function literature, which requires an accurate measure of the stance of monetary policy. We confine our attention to the decisions made at regular FOMC meetings, knowing that these do not capture all changes in monetary policy. (3) Our main question is how well the public might have anticipated FOMC decisions based on information that was available at the time. This question is related to the general issue of transparency in monetary policy. For example, Faust and Svensson (2001, p. 373) measure the degree of transparency by "how easily the public can deduce central-bank goals and intentions from observables."

   For each FOMC meeting we measure, based on summaries of the discussion, directives, and transcripts from FOMC meetings, and attempt to explain intended changes in FOMC policy. We characterize each decision as an easing of policy, a tightening of policy, or a decision to maintain the previous policy. Our unit of observation is the FOMC meeting rather than calendar quarters or months as has been the practice of most other researchers.

   While traditional reaction function studies have generally used revised data that were unavailable at the time of FOMC meetings, we are careful to use only real-time data, that is, data that could have been known by the public and the FOMC at the time of the meetings. (4) If the FOMC reacts to data that are not publicly available, say, its private forecasts, then FOMC decisions should be difficult for the public to predict. (5) We examine this issue by investigating the marginal explanatory power from adding the Federal Reserve's internal estimates of current conditions and confidential forecasts to the publicly available information set.

   We discuss our characterizations of FOMC decisions in section 2. Section 3 outlines our modeling strategy for estimating the probability of decisions to ease or tighten policy and the measurement of the variables in the models. Section 4 presents the estimation results. Section 5 explores the value of the Federal Reserve's internal forecasts, and section 6 discusses the implications of our results.

2. Characterizing FOMC Decisions

   We employ what has been called the narrative approach in which policy documents are interpreted to gauge the FOMC's policy stance. The advantage of this approach is that it can be used across the different operating target regimes that the Federal Reserve has employed. The drawbacks of this approach are the substantial room for subjectivity in document interpretation and in distinguishing between large and small policy moves. (6) We limit the subjectivity inherent in the narrative approach by focusing on the intended change in policy and forgoing any attempt to gauge its size. In contrast, authors of earlier studies using the narrative approach characterized monetary policy as expansionary, contractionary, or neutral (and to what degree; see Boschen and Mills 1995). This is no mean assignment as an attempt to define a "neutral" monetary policy will make clear. (7) Since the FOMC often uses a phrase similar to "the Committee seeks to maintain the existing degree of pressure on reserve positions" or "the Committee seeks to increase slightly the degree of pressure on reserve positions," assessing the intended change in policy from the preceding meeting is more likely to be successful than attempting to determine if policy is easy or tight in an absolute sense and to quantify the scale. (8)

   Each FOMC meeting results in a short-run directive for monetary policy that is the basis for any dynamic open market operations between meetings. For each meeting, we studied the summary of the discussion and the directive and determined if the FOMC had intended to tighten, loosen, or maintain policy. When available we tried to resolve any ambiguity by consulting the verbatim transcripts of FOMC meetings. (9)

   As anyone who has read these documents can attest, the FOMC is not always clear as to its objectives. (10) Thus, ambiguities are unavoidable. (11) Sometimes the FOMC directive does not clearly indicate a policy change. If the transcripts did not clarify the FOMC's intention or if they were not available, we relied on changes in targets for monetary growth and the federal funds rate. Occasionally, the targets announced in the directives appeared to be conflicting, forcing us to choose between the targets. In these instances, we relied on the change in the federal funds rate target to define the change in policy. (12)

   During the Volcker chairmanship the FOMC held 67 regularly scheduled meetings. The committee voted to ease 14 times, to tighten 13 times, and to leave policy unchanged 40 times, or about 60% of the meetings. This percentage rose to almost 80% during the Greenspan period. Of the 92 FOMC meetings Greenspan chaired through 1998, there were only 10 votes to ease and 10 votes to tighten. Figure 1 plots the cumulative values of our policy change variable, initialized to zero and coding an easing as -1 and a tightening as +1. (13) The graph indicates the FOMC tightened in the first two years of the Volcker chairmanship but then generally eased afterward, moving to a tighter policy just before the end of Volcker's tenure. During the Greenspan chairmanship, our policy change measure indicates an initial tightening until mid-1989 and then an easing until late 1991. There was no change in policy during 1992 and 1993. The FOMC often voted to tighten policy during 1994 and then voted to ease three times from July 1995 through the first meeting in 1996. With the exception of one vote to tighten in March 1997, FOMC policy remained unchanged from spring 1996 until the two decisions to ease in late 1998. The figure supports the common perception that changes in monetary policy are applied in stages. This pattern is particularly apparent over the past 15 years, reflecting either a cautious approach to policy or the serial correlation in economic conditions.

   [FIGURE 1 OMITTED]

3. Models and Data

   To assess the relationship between FOMC policy decisions and available economic data, we estimate models of the probability of the alternative decisions as a function of recent economic information. We assume there is an unobservable policy index, I[P.sub.t], that depends on a set of observable variables, [X.sub.t], and an error term, [[epsilon].sub.t], so that I[P.sub.t] = [X.sub.t][beta] + [[epsilon].sub.t]. We assume the FOMC chooses its intended change in policy by comparing I[P.sub.t] to two threshold values. Values of I[P.sub.t] below the lower threshold result in a decision to ease policy, while values above the upper threshold result in a decision to tighten policy; otherwise, the FOMC votes to maintain the previous policy. Because we assume a normal distribution for [[epsilon].sub.t], the ordered probit model is an appropriate estimator of [beta] and the threshold values, given the ordered nature of the policy alternatives. (14) In the ordered probit models, the signs of the estimated coefficients give the direction of change in the probability of a decision to tighten. Thus, a positive coefficient on a variable indicates that an increase in that variable raises the probability that the FOMC will decide on a tighter policy. Throughout the following discussion, we describe the effects of variables on the probability of a tighter policy. Each description can be reversed for the probability of an easier policy.

   Presumably, the FOMC chooses to change policy when it believes the current policy would produce undesirable values of its ultimate goals. This may be indicated by undesirable current values of the goal variables, by values of intermediate targets that diverge from their desired paths, and by values of information, or indicator, variables that imply unacceptable future developments. (15) The goals of U.S. monetary policy are usually taken to be stable prices, low unemployment, and sustainable economic growth. In addition, the FOMC often mentions the trade balance in its discussions, with large imbalances considered undesirable. Quantitative targets corresponding to the goals are not available, and as a result, we...