An important feature of the Rule is that simple application of these selected parameters to the 1984-92 period proved surprisingly successful in replicating the actual movement of the Fed's primary policy instrument, the nominal federal funds interest rate. By implication, the Federal Reserve was reacting to the economic environment as described by Taylor's Rule, or something like it, in establishing monetary policy. This feature of Taylor's approach contributed in no small measure to its immediate popularization in both research and policy studies.
The Taylor Rule's basic approach to policy inference can be applied to a more extended period of U.S. economic history. The period 1875-1998 was selected because of the availability of reasonably consistent data series and because of its divisibility into three roughly equal (over 35 years each) sub-periods of monetary control.4 The first period (1875-1913, or "Pre Fed") was characterized by the absence of a central monetary authority. Monetary discipline was exercised in rather severe fashion by way of the international gold standard. Deflation was a problem as much as inflation during this era.5 The connection between the relevant variables was enunciated by David Hume in the 18th century and recounted by Taylor (1998) in the context of the specie flow mechanism. Rising inflation in a country leads to a balance of payments deficit (prices of local goods become less competitive), gold outflow, and tighter monetary conditions as reflected in higher interest rates, which tend to initiate a reversal of the process. Deflationary pressures would make a country's goods more attractive to foreigners, resulting in gold inflow, monetary ease, and lower interest rates.
The Federal Reserve was created in December 1913 by Congress in reaction to a series of financial crises culminating in the panic of 1907. The purposes of the Federal Reserve Act "were to give the country an elastic currency, to provide facilities for discounting commercial paper, and to improve the supervision of banking." Coincidentally, the beginning of WWI and the end of the international gold standard occurred at about the same time as the Fed's arrival on the world scene.
The Early Fed years (1914-1951) were marked by institutional struggles, policy mistakes, and subsequent loss of policy independence. The Federal Open Market Committee and open market operations did not appear until the 1920's. Benjamin Strong, President of the Federal Reserve Bank of New York, helped develop the FOMC and was credited with some of the early success enjoyed by the young Fed. Strong's death in 1928 created a power vacuum and led possibly, to the ill-fated actions (or inactions) of the monetary authority in permitting the money supply to fall by approximately one-third between 1929 and 1933. The Banking Act of 1935 addressed some of the organizational flaws in the Fed's structure, but unfortunately, the first major act of the restructured monetary authority was to double reserve requirements in 1936 and 1937, thereby plunging the economy deeper into its Depression state. In April 1942, the FOMC announced a plan to peg interest rates at a very low level through the purchase of Treasury obligations to support the war effort. The Fed continued its policy of bond interest rate pegging (in effect, a loss of policy independence) until March 1951. The Treasury official who helped negotiate the Federal Reserve-Treasury "Accord", William McChesney Martin, was appointed chairman of the Federal Reserve Board in April 1951.
The period 1952-98 may be identified as the "Modern Fed", an era marked by monetary policy independence and generally robust economic activity. Although Martin's long tenure (1951-1970) and those of his successors, Arthur Burns, William Miller, and Paul Volcker were characterized by frequent controversy, it is clear that policy learning has occurred. The actions of the Alan Greenspan Fed (1987-present) are generally considered to have been among the most enlightened.
In order to apply the Taylor Rule to the three historical eras, it is necessary to specify parameters and weights for use in equation (1). Following Taylor, π and r f were set equal to 2.0. His choice is defensible since the long-term (1875-1998) means of inflation and the real (commercial paper) interest rate in our data set are 2.3 and 2.4, respectively. Also, as with Taylor, we set weights of g = 0.5 and h = 0.5, generating monetary "Rule 1". We also specify a monetary "Rule 2", with all values as noted except for altering the setting of the real output weight to g = 1.0.
Figure 1 (lower) depicts the actual annual percentage change in prices (π) and the GDP gap (y) for the Pre Fed period, and Figure 1 (upper) shows the actual policy (commercial paper) interest rate (r) along with the "recommended" policy rates generated by Rule 1 and Rule 2. Note that the actual rate changed little over the entire Early Fed Period despite wide swings in deflation, inflation and the GDP gap. For example, the longest contraction in U. S. history, stemming from the panic of 1873, ran 65 months into 1879, and was accompanied by falling prices; yet the nominal interest rate remained steady at about 5 percent. In contrast, the Rule 1 and Rule 2 policy rates fell to zero (truncated) in the late 1870s as well as for a period of time during the 1880s.
Prices fell in the mid 1880s and throughout much of the 1890s. The economy also suffered contractions throughout much of the 1890s, during 1890-91, 1893-94, 1895-97, and 1899-1900 (NBER dating).13 The actual interest rate dropped somewhat at various times in the 1890s but hardly at all in comparison with the plunge recommended by the two Policy Rules. The lack of response relative to the severe rate swings suggested by the Taylor Rule call into question reliance on the specie flow mechanism as an automatic policy correction device. This point is more valid for the 1890s than the 1870s in that the United States formally returned to the gold standard in 1879, long after its Civil War suspension.
Prices rose somewhat and the economy expanded strongly the first part of the new century until about the time of the 1907 panic. The actual interest rate rose slightly during this period, but the two Rule rates (particularly Rule 2, which is more responsive to real output changes) move up in restrictive fashion to levels above 10 percent prior to the Panic. The Rule rates then swing sharply downward (in contrast to a mild decline in the actual rate) in response to the 1907 panic and its aftermath.
Figure 2 presents the same information for the Early Fed period as Figure 1 did for the Pre Fed era. Strong growth and increasing prices accompanied the expansive activities associated with WWI, prompting the Rule 1 and Rule 2 policy rates to rise to unrealistically high levels in 1916-20. Actual interest rates rose only slightly during this period. Although WWI ended in November 1918, the Federal Reserve did not undertake seriously restrictive policies until late 1919, increasing the discount rate sharply from 4 to 7 percent between October 1919 and June 1920 (causing a parallel move in the commercial paper rate). This contrast between actual and hypothetical Fed behavior points up two defects in applying the Taylor Rule to actual situations: 1) the Rule ignores Federal Reserve concerns such as war finance and balance of payments issues; and 2) the Rule does not accommodate lag effects.
The Fed was willing to tolerate inflation during WWI as they helped finance the war through low discount rates to member banks, using the loans to purchase bonds sold by the Treasury. The low interest rates subsequently encouraged gold outflows; thus, the Fed's restrictive policies immediately following the war were undertaken not only to slow inflation but also to reverse gold flows. Even granting the Fed leeway in the complexity of its decision making, it was still guilty of excessive ease during the war (in comparison to the Rule) and excessive restraint following the war. Early Fed seemed not to understand that low discount rates in comparison with other rates encouraged the member bank borrowing that brought higher bank profit and excessive growth to the monetary base and money supply. The Fed then swung policies too far in the other direction, causing negative money supply growth and contributing to numerous bank failures and a steep 18-month recession 1920-21.
Arguably erroneous actions of Early Fed 1928-37 included hiking the discount rate in August 1929, thereby contributing to the stock market crash, raising the discount rate in October 1931 to stem the flow of gold to Great Britain, a policy of virtually no open market purchases in the early 1930s, and the doubling of reserve requirements 1936-37. The Rules' interest rates of zero throughout the 1930's reflect the need for more simulative actions by the Fed. While actual interest rates were pushed below 1 percent for an extended period, other elements of monetary policy (such as the hiking of reserve requirements) were inappropriately concretionary.
The high (excessively so) policy Rule rates of the 1940s reflect the recommended reaction to both inflationary pressures and a GDP above potential for much of a decade marked by administrative controls and the low actual interest rates that accompanied the Fed's support of the war effort.
Figure 3 depicts actual macroeconomic variable movements and the policy Rule interest rates of the Modern Fed era. The initial indication of Fed policy "reasonableness" during this period is the fact that the spread between the actual policy rate and the Rule rates is never more than 10 percent, in contrast to the two earlier periods. The greater variance of the actual short-term rate may also be reflective of Modern Fed's intensified effort to employ interest rates as a countercyclical policy mechanism.
The Fed did not achieve perfect coincidence with the Taylor Rule during this era, but seemed to move closer to it as the period drew to a close. Policy Rule movements suggest that the Fed should have been more restrictive through most of the 1950s, more stimulative 1958-62, and considerably tighter during the 1963-80 period of inflationary pressures. The Volcker years of the 1980's into 1987 appear to have been too restrictive, based on the policy Rule comparison, but the Greenspan era, from 1987 on, achieves remarkable coincidence of actual shortterm rates and policy Rule rates.
Chapter III
3. The Enhanced Model
The usefulness of the Taylor Rule with its fixed coefficients and simultaneous reaction to events has been frequently demonstrated. Yet, its applicability for our purposes can be enhanced with simple modifications. Rather than accepting the assigned weights of 0.5 and 0.5 (or 1.0) for h and g respectively, it is possible through multiple regression to obtain coefficients for the variables for each separate monetary era. Also, it may be useful, and perhaps, more realistic, to allow for a lagged reaction of the monetary authority to changes in the inflation and output scenario.
From equation (1), Taylor's Rule is
rt = πt + gyt + h(πr -π*) + rf .
In a more regressable form:
where α = rf - hπ*
It is possible to estimate this equation directly (see Taylor [1998]) but that posits contemporaneous responses of the interest rate to changes in economic conditions. There is the possibility that the Federal Reserve faced lags in recognizing macroeconomic problems or was slow in reacting to those problems. Changes in y or π might not be realized correctly during the current year due to lags in collecting and analyzing data. That is, due to lags in data availability as well as subsequent data revisions, the monetary authorities would be making decisions on the basis of incomplete information. Since we are using annual data, this may not be a serious deficiency for recent decades, but could certainly be a difficulty for the early periods in our study, given the improvements in data collection and measurement that have occurred over time. Even if the Fed policymakers immediately perceived a problem causing them to change the target interest rate, they might not choose to move immediately to that interest rate.
To allow for the possibility that the Fed has lags in responding to inflation, πt in equation (2) is replaced by
Due to lags in either recognition or reaction the Fed responds to last year's inflation plus some fraction, X, of the change in inflation that occurred this year. Similarly, yt is replaced by yt'
The Fed only responds to some fraction, y, of the change in the GDP gap. Substituting π't and y't into equation 2 generates
Equation (5) only accommodates lags of a single year, but this is probably justifiable. Spencer and Huston (1998) and Falls and Hill (1985) found even shorter lags in the Fed's response to changes in unemployment and inflation.
Chapter IV
4. Empirical Results And Inter-period Comparison
The results from estimating equation (5) are shown in Table I.17 The model provides a good fit for the 1952 to 1998 period with an R2 of .87. The coefficients on the contemporaneous values of inflation and the GDP gap are both significant at the 1 percent level. The lagged values are not significant, which supports Taylor's implicit contention that all adjustment takes place rapidly.18 By using the estimated coefficients, it is possible to solve for (1+h) and g, the values corresponding to inflation and the GDP gap in Taylor's equation. We find a coefficient on inflation of .89 and a coefficient on the GDP gap of .33, both significant at the one percent level. While the coefficient on inflation is significant and of the right sign, its magnitude is not impressive. The coefficient is insignificantly different from one. A coefficient of one would suggest that while the Federal Reserve does respond to inflation by increasing nominal interest rates, it raises them only enough to keep real interest rates constant.
The results for the Early Federal Reserve period reveal a pleasingly high R2 of .86, but that is misleading since nearly all of the explanatory power comes from the correction for serial correlation. The coefficient on lagged inflation comes closest to being significant, but none of the coefficients are significant at the 10 percent level. Thus the derived values of g and (1+h) are also insignificant. Note that an inability to reject the hypotheses that g and (1+h) equal zero is not due to large standard errors; in fact, the standard errors are quite small compared to the other periods. Rather it is because the coefficients are so small. In this case the coefficients are found to be insignificant because it can be stated with a large degree of confidence that their true values lie close to zero. This is rather damning evidence against early Fed policy. The results imply that the Fed did not respond to either the crushing GDP gaps of the 1930s or the war-induced inflation of the 1940s. That the 1+h coefficient is significantly below one implies that the Fed did not change nominal interest rates sufficiently in response to inflation to move real rates in the same direction. In theory this leads to instability and in practice this may have contributed to the greater observed macroeconomic instability in this period.
The Pre Federal Reserve period has an R2 of .46. The explanatory power here comes primarily from the coefficient on the lagged GDP gap. (The contemporaneous coefficients are insignificant) This generates a significant g coefficient, which while larger than the Early Fed period is far smaller than the Modern Fed's. The pattern of seeing a larger and quicker response in the more recent data is consistent with the results of Spencer and Huston (1998) who showed that the Federal Reserve responded more rapidly and with greater force in the 1980s and 90s than it did in the prior three decades.
Table l: Enhanced Model Regression Results
Modern Federal Reserve Period
R- square = .87 R-square adjusted = .85
Log Likelihood Function = -69.90, Durbin's h21 = 1.44
As demonstrated above, the Taylor Rule is one method of judging an earlier period's monetary policy by modern standards. To obtain a more precise evaluation, take the estimated equation for the Modern Fed and substitute in the values of inflation and the GDP gap for the other time periods. This generates fitted values that describe how the Modern Fed would have reacted to economic scenarios of the past. The same procedure is followed to generate predictions of how the policies of the Pre Fed and Early Fed would have been employed outside their time periods.
As indicated by Figure 4, Modern Fed would have varied its interest setting policy considerably in reaction to the challenging economic environment of the Pre Fed era. It would have employed particularly expansive polices in the late 1870s, mid 1880s and mid 1890s in consonance with the prevailing deflationary conditions of the times. It would have become more restrictive in the early part of the new century, but loosened policy in response to the 1907 panic. Note, however, that the Modern Fed would not swing rates as low or as high as the Taylor Rules would prescribe for this period (see Figure 1). Early Fed would have altered its policy rate very little throughout the Pre Fed era. The level of its policy rate would have remained below the actual interest rate the entire period, which would seem appropriate during the deflationary pre-1900s, but inappropriate in the early years of the new century.
Figure 5 indicates that the Modern Fed would have pursued restrictive policies during both World Wars (in the absence of any concern for war bond finance) and would have been quite simulative during the Great Depression. As with the earlier period, the interest swings are dampened considerably for Modern Fed in contrast to the Taylor Rules (Figure 2). Pre Fed appears to move policy generally in the same direction as Modern Fed, but employs smaller interest rate changes.20 For example, Pre Fed's interest rates fall to near-zero in the mid 1930s and rise to a maximum of 10 percent in the mid-1940s compared with a rate about double that for Modern Fed.
During the Modern Fed era (Figure 6), there appears to be some consonance in movement of Pre Fed's policy rate and that of the actual rate, but the rate remains between 4 and 6 percent the entire period. The Early Fed rate varies even less, remaining between 3.0 and 3.5 percent throughout. A more stable economy 1952-1998 than the two earlier periods helps account for the lack of policy change by these two hypothetical Feds. However, it is clear that both Pre Fed and Early Fed react less to changing economic conditions than Modern Fed.
Another way to compare reaction functions with prevailing economic conditions is found in Table 3. The table shows the standard deviation of interest rates recommended by the three policy authorities across the three regimes. As expected, based on Figures 4, 5, and 6, the standard deviation of the (nominal) policy rate recommended by Modern Fed exceeds the standard deviation of the rate recommended by the other two authorities for each era as well as for the entire period.
A more complete picture of regime comparisons is obtained by inspection of the "real" policy Rule rates in Table 3. The standard deviation of the "real" policy Rule rate recommended by Modern Fed is less than that of the other two regimes for each era as well as for the entire period. Since the Modern Fed advocates more aggressive nominal rate responses to inflation it achieves a smoother path of real interest rates.
A number of caveats apply to this out-of-sample simulation exercise. Very different economic conditions prevailed in each era. Inflation and output targets shared priority status with other economic objectives. The central bank was either non-existent or operating at a very formative stage of development. The Taylor Rule itself is subject to critical comment. These concerns notwithstanding, the empirical results reflected in the three figures and Tables 2 and 3 suggest a policy authority that has become more active and appropriately responsive in dealing with varying macroeconomic problems across time.
Chapter V
5. Conclusion
The second-guessing of Federal Reserve policies has been a popular pastime for analysts and politicians since the Fed's inception. This paper develops an approach adapted from recent contributions by professor John Taylor that have proved useful for this and other purposes. Taylor's Rule links the Fed's short-term policy interest rate with a combination of inflation and output goals. It suggests, for example, that the Fed raise the policy rate by certain magnitudes dependent on whether inflation is above an arbitrarily chosen target and/or if output is above its potential or trend. Taylor chose parameters for the Fed policy Rule (or reaction function) based on characteristics of the 1984-92 economy.
We applied Taylor's Rule to three long periods in U.S. economic history, 1875-1913 (Pre Fed), 1914-1951 (Early Fed) and 1952-1998 (Modern Fed) to study the actual movements of a short-term interest rate relative to the change in the rate suggested or "recommended" by Taylor's Rule. Since the long-term (1875-1998) means of certain parameters were not decidedly different from those selected by Taylor, initial policy comparisons were developed on the basis of Taylor's basic Rule.
The actions of Modern Fed, as reflected in changes of the actual short-term rate, were determined to be much closer to actions suggested by Taylor's Rule (or a slight variation thereof), than those of either Early Fed or Pre Fed. This result was not surprising in that movements of the short-term rate can only be indirectly associated with the "automatic" component of the specie flow mechanism of the 1875-1913 international gold standard years, and the Early Fed's policy difficulties in association with two World Wars and the Great Depression are well chronicled.
To provide a more sophisticated analysis of how historic monetary actions measure up to modern standards, we constructed an enhanced model that corrects for serial correlation and permits a lagged response by the monetary authority to changing economic conditions. Coefficients obtained from the estimated equation for the Modern Fed were used to generate predicted policy rates for earlier periods. The results demonstrated that Modern Fed would have responded with vigor and speed to the difficult inflation and output situations that characterized the Pre Fed and Early Fed periods. Also, policy interest rate swings appear more moderate and reasonable than those associated with the basic Rule.
Neither Pre Fed nor Early Fed would have changed the policy rate much to counter the economic problems that surfaced in any era, although Pre Fed's reactions came closer to the corrective policies of Modern Fed than did those of Early Fed. Modern Fed appeared considerably more aggressive than the other two regimes in terms of manipulating the nominal rate to achieve inflation and output targets for each period. Such action generates smaller swings in the real interest rate. Finally, the enhanced model's empirical components indicate that Modern Fed responded in nominal terms strongly, and within the same year to both inflation and output targets, Pre Fed responded with a one-year lag to a real output gap, and Early Fed responded in no significant or consistent fashion to either inflation or output challenges.
REFERENCES
Balke, N. S. and Gordon, R. J., (1989), The Estimation of Prewar GNP: Methodology and New Results, Journal of Political Economy, 97, 38-92.
Ball, L., (1999), Policy Rules for Open Economies, in J. B. Taylor, ed., Monetary Policy Rules, Chicago: University of Chicago Press, 203-246.
Benhabib, J. Schmitt-Grohé, S. and Uribe, M., (1999), The Perils of Taylor Rules, Center for Economic Policy Research, Discussion Paper No. 2314, December.
Bernanke, B. and Woodford, M., (1997), Inflation Forecasts and Monetary Policy, Journal of Money, Credit and Banking, 24, 653-684.
Board of Governors of the Federal Reserve Systems, (1963), The Federal Reserve System: Its Purposes and Functions, Washington,
D.C. Bryant, R. C., Hooper, P., and Mann, C. L., (1993), Evaluating Policy Regimes, Washington, D.C.: The Brookings Institution.
Chandler, L. V., (1958), Benjamin Strong, Central Banker, Washington, D. C: The Brookings Institution.
Clarida, R., Galí, J. and Gertler, M., (2000), Monetary Policy Rules and Macroeconomic Stability: Evidence and Some Theory, Quarterly Journal of Economics, 115: 147- 180.
DeLong, J. B., (1999), Financial Crises in the 1890s and 1990s: Must History Repeat?, Brookings Papers on Economic Activity, 2, 253-279.
Falls, G. A. and Hill, J. R., (1985), Monetary Policy and Causality, Atlantic Economic Journal, 13, 10-18.
Fishe, R. P. H. and Wohar, M., (1990), The Adjustment to a Change in Regime: Comment, American Economic Review, 80, 968-976.
Friedman, M. and Schwartz, A. J., (1963), A Monetary History of the United States 1867- 1960, Princeton, NJ: Princeton University Press.
Gordon, R. J., (2000), Macroeconomics, Reading, Mass: Addison Wesley.
Greider, W., (1987), Secrets of the Temple: How the Federal Reserve Runs the Country, New York: Simon and Schuster.
Hetzel, R. L., (2000), The Taylor Rule: Is it a Useful Guide to Understanding Monetary Policy?, Federal Reserve Bank of Richmond Economic Quarterly, 86 (2), 1-33.
Hicks, J. R., (1937), Mr. Keynes and the 'Classics', Econometrica Inder, B.A., (1984), Finite Tests for Autocorrelation in Models Containing Lagged Dependent Variables, Economics Letters, 57, 53-68.
Judd, J. P. and G. D. Rudebusch, (1998), Taylor's Rule and the Fed: 1970-1997, Federal Reserve Bank of San Francisco Economic Review, 3, 3-16.
Krugman, P. R., (1998), It's Back: Japan's Slump and the Return of the Liquidity Trap, Brookings Papers on Economic Activity, 2, 137-187.
Mankiw, N. G., Miron, S. A. and Weil, D. N., (1987), The Adjustment of Expectations to a Change in Regime: A Study of the Founding of the Federal Reserve, American Economic Review, 77, 358-374.
Mankiw, N. G., Miron, S. A. Weil, D. N., (1990), The Adjustment of Expectations to a Change in Regime: Reply, American Economic Review, 80, 977-979.
Meyer, L. H., (1997), Monetary Policy Objectives and Strategy, Business Economics, 32, 17-20.
Onatski, A. and Stock, J. H., (1999), Robust Monetary Policy Under Model Uncertainty in a Small Model of the U. S. Economy, Paper presented at Conference on Monetary Policy and Monetary Institutions, sponsored by Federal Reserve Bank of San Francisco and the Stanford Institute for Economic Policy Research, March 5-6.
Orphanides, A., (1997), Monetary Policy Rules Based on Real Time Data, Board of Governors of the Federal Reserve System, Mimeo.
Patrick, S. C., (1993), Reform of the Federal Reserve System in the Early 1930s: The Politics of Money and Banking, New York and London: Garland Publishing, Inc.
Peersman, G. and Smets, F., (1999), Uncertainty and the Taylor Rule in a Simple Model of the Euro-area Economy, Paper presented at Conference on Monetary Policy and Monetary Institutions, sponsored by Federal Reserve Bank of San Francisco and the Stanford Institute for Economic Policy Research, March 5-6.
Rogoff, K., (1998), Comment on 'It's Back: Japan's Slump and the Return of the Liquidity Trap, Brookings Papers on Economic Activity, 2, 194-199.
Rudebusch, G. and Svenson, L. (1999) Policy Rules for Inflation Targeting, in J.B. Taylor, ed., Monetary Policy Rules, Chicago: University of Chicago Press, 203-246.
Spencer, R. W., (1996), Monetary Policy at Work: Lessons from the FOMC Transcripts, Journal of Economic Education, 27, 309-322.
Spencer, R. W. and Huston, J. H., (1998), Monetary Lags: Changing or Constant, Journal of Business and Economic Perspectives, 24, 81-94.
Sprague, O. M. W., (1997), History of Crises Under the National Banking System, New York: A. M. Kelley.
Svensson, L. (1999), Inflation Targeting as a Monetary Policy Rule, Journal of Monetary Economics, 43, 607-654.
Tallman, E. and Moen, J., (1998), Gold Shocks, Liquidity, and the United States Economy During the National Banking Era, Explorations in Economic History, 35, 381-404.
Taylor, J. B., (1993), Discretion versus Policy Rules in Practice, Carnegie-Rochester Conference Series on Public Policy, 39, 195-214.
Taylor, J. B., (1998), An Historical Analysis of Monetary Policy Rules,"
J. B., (1998), An Historical Analysis of Monetary Policy Rules," Paper presented at the NBER Conference on "Monetary Policy Rules", January 15-17.
