|
Article Excerpt
I. INTRODUCTION
There has been a significant reduction in the macroeonomic volatility in the U.S. economy during the last 20 yrs. The standard deviation of the central bank's target variables like inflation and nominal gross domestic product (GDP) declined substantially; however, at the same time, the standard deviation of monetary aggregates has increased. (1) This unique macroeconomic phenomenon is also associated with the disappearance of correlation between inflation and money growth. This is in contrast to the monetarist argument that suggests that an important cause of economic instability is excess variation in the growth rate of money. The most famous monetarist prescription for monetary policy is Milton Friedman's proposal that the growth rate of money should be constant. There is, however, no evidence of falling variations in the growth rate of money over the last 20 yrs. In fact, the standard deviation of the rate of growth of monetary base has increased from 1.63% during 1960-1982 to 3.1% during 1983-2003.
Milton Friedman, in an interview with Taylor (2001), pointed out that there was a breakdown in the relationship between money and GDP after 1980, which came along with a dramatic reduction in the variability of growth rate of GDP. Friedman says, "I am puzzled about is whether, and if so how, they suddenly learned how to regulate the economy. Does Alan Greenspan have an insight into the movements in the economy and other shocks that the people don't have?" Friedman further says, "It looks as if somehow in 1991-1992 they were able to install a good thermostat instead of a bad one. Is Alan Greenspan a good thermostat compared to other Fed chairmen?" (Taylor 2001).
There is widespread consensus on the effect of monetary policy on inflation in the long run. However, there is no consensus on how monetary policy affects real output in the short run. There is now a consensus among economists that the original monetarist proposition that there is no long-run trade-off between inflation and unemployment is correct. Monetarism can also be viewed as the platform from which the rational expectation revolution took off. However, the monetarist proposition that economic fluctuation is caused by excessively variable monetary growth has largely been abandoned. Authors like Estrella and Mishkin (1997) and Feldstein and Stock (1994) find that monetary aggregates do not have a significant informational content after 1979 and, hence, cannot be used as an information variable for the stance of monetary policy. Their finding is also consistent with Dwyer (2002) and Dwyer and Hafer (1999), who find that there is little correlation of inflation and money growth across countries on quarterly and annual bases after 1979. The literature finds significant correlation (e.g., Lucas 1980) between money growth and inflation in the pre-1979 period, but this correlation has disappeared after 1979. (See Table 1 and Figures 1-4).
If stabilization policy makes inflation a constant, then there will be no correlation between inflation and policy as no variable is correlated with a constant. According to Kochin (1973), if the monetary authority is able to stabilize the economy by optimally setting the variability of the control variables (e.g., monetary aggregates and interest rate), then there will be no correlation between the goal variable (e.g., inflation and nominal GDP) and the control variable even if the policy is not perfect. If the policy is too variable and even if it is well timed, then a countercyclical policy adopted by the central bank might lead to greater variability in the goal variable, as argued by Friedman (1953). Poole (1993, 1994) and Tanner (1993) also argue that one predictable consequence of optimal monetary policy is that the correlation between monetary policy instruments and policy goals will be driven to zero. Poole further contends that it is obvious to any careful reader of Theil (1964) that optimally variable policy will give rise to a zero correlation between policy and goal variable. But, we will show in this paper that Kochin (1973) and Poole (1994) are correct only if the coefficient of effect of the policy variable on the goal variables is known precisely.
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
[FIGURE 3 OMITTED]
[FIGURE 4 OMITTED]
In this paper, we try to provide an explanation for the demise of monetarism in the United States since 1983. The main focus of this paper is to judge the effect of monetary policy on inflation and nominal GDP and to show how this relates to the demise of monetarism. Our primary focus is on inflation. For the sake of exposition, we have also shown the effect of monetary policy on nominal GDP. Current monetary policy literature talks about the response of the federal funds rate to stabilize goal variables (inflation rate and GDP) often in the framework of Taylor's rule. The improved performance of the economy is also related to the demise of monetarism in the policy-making and academic literature. Benjamin Friedman and Kuttner (1992) interpret this breakdown as a consequence of shifts in the structural equations, for example, instability in the demand for money equation, whereas we relate it to the success of monetary policy.
In The Effects of Stabilization Policy section we have shown that optimal policy leads to no correlation between policy and inflation if the effect of the policy is known precisely. We show in The Effects of Stabilization Policy under Coefficient Uncertainty section that, under coefficient uncertainty, optimal policy will lead to a negative correlation between the control variables and the goal variable. The negative correlation results from coefficient uncertainty; because the less certain we are about the size of a multiplier, the more cautious we should be in the application of the associated policy instrument. Therefore, although optimal policy leads to lack of correlation between monetary policy and inflation if the coefficient is known, it will lead to a negative relationship between the two if there is coefficient uncertainty. The higher the uncertainty, the more cautious will be optimal policy response. In The Effects of Stabilization Policy under Imperfect Control section we show the effect on the optimal correlation between the goal and the control variable, if the control variable is under imperfect control.
In this paper, we evaluate whether the fluctuations in the control variables were too large or too small to stabilize inflation in the U.S. economy after 1960 on the basis of our simple model. We also find out whether the movements in the control variable stabilized or destabilized the economy. However, as pointed out earlier, while there has been a widespread support for Taylor's rule, we know of no systematic study showing whether the deviations of the federal funds rate from Taylor's rule were stabilizing or not. The framework that we are going to follow in this paper will allow us to examine whether variation in monetary aggregates, the federal funds rate, or the deviations of the federal funds rate from Taylor's rule were too large or too small to stabilize inflation and GDP growth in the United States. We also give a more tentative reading as to whether monetary policy was stabilizing or destabilizing.
II. MODEL
A. The Effects of Stabilization Policy
If a central bank were eager enough to learn how monetary policy operates, it could perform a controlled experiment. Monetary policy instruction would be obtained from a random number generator and then monetary policy would be uncorrelated with all disturbing forces both known and unknown. But, stabilization policy works by counteracting other forces. Successful countercyclical policy will produce a bias tending to reduce the size and significance of the effect of the policy variable on the goal variable in our regressions. Imagine investigators who do not take into account any explanatory variables other than policy variables. First, we show exactly how the estimates of the true or ceteris paribus effect are biased and, second, how these biased results can be used for judging monetary policy.
Imagine investigators who attempt to determine the influence of a control variable on a goal variable. The true situation is represented by:
(1) Y = [[alpha].sub.s] S + X,
where Y is the goal variable, S is the control variable, and X is a sum of the true effect of all sources of movement in Y other than S. [[alpha].sub.s] is the ceteris paribus effect of the control variable on the goal variable. In this model we assume that as is known and positive. (2) We assume that the control variable is linearly related with the goal variable. The variance of Y when [[alpha].sub.s] is known is given as follows:
(2) [[sigma].sup.2.sub.Y] = [[sigma].sup.2.sub.X] + [[alpha].sup.2.sub.S] [[sigma].sup.2.sub.S] + 2 cov(X, [[alpha].sub.S]S),
which can be written as:
(3) [[sigma].sup.2.sub.Y] = [[sigma].sup.2.sub.X] + [[alpha].sup.2.sub.S] [[sigma].sup.2.sub.S] + 2[[alpha].sub.S] [r.sub.S.X] [[sigma].sub.S]...
|
