Choosing a Policy Instrument

reviews some caveats bearing on any final determination of the success of a particular rule. Results suggest that monetary base rules do not consistently produce greater levels of price stability—performance is model dependent. In contrast, the M2 rule consistently met the four criteria listed above during the sample period 1964:Q1–1995:Q4.

II. Choosing a Policy Instrument

Two operational concerns arise if the policy instrument is not directly andor accurately controllable by the Fed. First, according to McCallum 1988, for a rule to be operational, there must be a way to monitor compliance. Because data on the components of the monetary base are available to Fed staff daily, the Fed can maintain more precise short-run control over the base than M2. The public can more easily monitor Fed compliance with base rules because money control error is unlikely to be a source of missed monetary base targets. 7 Although short-run M2 control is less precise, compliance with an M2 rule might be gauged by establishing a narrow error band for the specified, quarterly M2 target level, as is commonly done with bilateral exchange rates under fixed exchange-rate regimes. Levels of M2 could be allowed to deviate from any given quarter’s target by the absolute value of a pre-specified, small percentage. Or, compliance could be measured by requiring the Fed to limit deviations from rule-specified M2 growth measured in quarterly loga- rithmic units to be zero on average with a mandated, relatively narrow variance. 8 The specified compliance period could be a longer period of time e.g., one year. The size of the error band in the first measure of compliance or the variance and compliance period in the second measure must be large enough to accommodate the normal range of money control error, but small enough to prevent discretionary policy-making. A second operational issue which arises from using M2 is that deviations of money growth from the quarterly, rule-specified targets may generate substantially greater deviations of nominal GDP from its target path. However, this problem might be mitigated in two ways. If money control errors exhibit a systematic, negative, contemporaneous correlation with GDP shocks, an M2 rule might not produce large deviations of simulated GDP from the target path. Alternatively, deviations of GDP from the target that are likely to result from the presence of money control errors may be small—if the impacts of the money control errors are mitigated by the rule’s feedback properties. Feldstein and Stock 1993 noted that another important condition affecting the performance of a policy instrument is the existence of a sufficiently strong and stable relationship to nominal income. Although absolute stability in velocity growth is not a necessary condition for McCallum’s rule to be successful, the more stable the relationship between a policy instrument and GDP, the greater the degree of price stability. Although there are thresholds of instability which rule out the use of particular policy instruments, 7 The monetary base is assumed to be controllable by the Fed over periods of time as short as one quarter. This assumption is arguable because the Fed doesn’t perfectly control all of the items in its balance sheet. However, the Fed can control Federal Reserve Bank Credit FRBC, the major item affecting the supply of reserves. Control of FRBC, combined with timely information on the most recent values of other items in the balance sheet and Fed staff forecasts of daily values of these other variables, should allow the Fed to maintain close, albeit imperfect, control of the monetary base over short time periods. 8 This is analogous to the way the Fed measures bank compliance with reserve requirements. Suitable Policy Instruments for Monetary Rules 381 such large degrees of instability would also hamper the formulation of discretionary policy. The strength and stability of the relationship between some policy instruments and income has been extensively examined. Friedman 1988 suggested that the relationship between the monetary base and nominal GNP has weakened as a result of surges in the growth rate of unusual types of currency demand e.g., foreigners living in unstable economies, drug traffickers, etc.. More recent studies in the literature on money-income causality have utilized Granger causality tests to determine if interest rates, the monetary base or M2 exhibits significant predictive power for either nominal or real income. These studies have often yielded conflicting results because test outcomes were highly sensitive to the variables included, lag length, sample period, temporal aggregation of the data, and the use of vector autoregression VAR versus vector error correction models VECM for estimation of Granger-equations. Recent studies include Friedman and Kuttner 1992, Becketti and Morris 1992, Feldstein and Stock 1993, Abate and Boldin 1993, and Dotsey and Otrok 1994. With the exception of Friedman and Kuttner, these authors generally have found significant predictive power for M2 in equations modeling nominal or real income. Feldstein and Stock 1993 found that: 1 M2 is a useful predictor of nominal GDP; 2 the coefficients linking M2 to GDP appear to be stable over time; 3 M2 has more predictive content for nominal GDP growth than interest rates do, and 4 the relationship between the monetary base and nominal GDP appears to be highly unstable. The presence of various deficiencies in the policy instruments has caused questions regarding instrument efficacy to be addressed empirically by comparing rule performance in simulations utilizing different policy instruments in a variety of plausible macro models. Judd and Motley 1991, 1992, McCallum 1987, 1988, 1990, and Thornton 1993 found that monetary base rules can be used successfully to target nominal income. This result is robust for different time periods, different economic models, different start-up conditions, and the Lucas Critique. However, Friedman 1988 argued favorable results for the monetary base were a statistical artifact, based on the strong relationship between currency and GDP, rather than the relationship between the total reserves component of the base and GDP. Results regarding M2 are mixed. Feldstein and Stock 1993 developed an M2 rule from a simple VAR model and found that the rule would probably reduce the average ten-year standard deviation of annual GDP growth by over 20, thus reducing the long-run average inflation rate assuming no M2 money control error. Thornton 1993 found that between 1964 and 1989, an M2 version of McCallum’s rule didn’t deliver substantially different levels of success in attaining the GNP target than a monetary base rule. This result was robust for different economic models, but simulations were con- ducted assuming no money control error. In contrast, Dotsey and Otrok 1994 found that the rule specified by Feldstein and Stock 1993 reduced the variance of nominal GDP only marginally and not in all cases when M2 money control errors were included in GDP simulations. Several factors determine the likely success of a policy instrument, but analysis suggests that both M2 and the monetary base exhibit some problems. The monetary base is more controllable, but M2 exhibits a stronger and more stable linkage to nominal income. Which characteristic is more important in rule-based policy? Section III addresses this question. 382 S. R. Thornton

III. Evaluating the Performance of Monetary Base and M2 Rules