Taylor Rules and the Great Inflation Alex Nikolsko-Rzhevskyy and David H. Papell San Francisco Fed November 9, 2010
Monetary Policy in the 1970s Result was the Great Inflation Did the Fed Conduct Bad Policy in the 1970s? If so, Hope for Good Policy in 2010s If not, What Should we Hope for? What is Good Policy Stipulate Answer from 1985 to 2002 The Taylor Rule and Taylor Principle Page 2
The Taylor Rule and Taylor Principle The Taylor Rule: The Taylor Principle Coefficient on π > 1 Real Interest Rate Increases when Inflation Rises Page 3
Central Questions of the Paper Did the Fed Follow a Taylor Rule in the 1970s with a Two Percent Inflation Target? Did the Fed Follow a Taylor Rule in the 1970s with an Increasing Inflation Target? Did the Fed Follow a Taylor Rule in the 1970s with a Higher Output Gap Coefficient? Does Forward-Looking Policy (Inflation Forecasts versus Inflation Rates) Affect the Answer? Page 4
Output Gap with Revised Data Taylor (1999) and Clarida, Gali, and Gertler (2000) Used Revised Data Quadratic or Hodrick-Prescott Detrended GNP Congressional Budget Office (CBO) Data Page 5
Revised output gaps Page 6
Actual and implied federal funds rate Page 7
Real-Time Data Revised Data Criticized by Orphanides (2001, 2003a, 2003b, 2004) Not Available to Policymakers When Decisions Were Made Use Real-Time Data Inflation Log Change in GNP Deflator for Previous Four Quarters Output Gap Council of Economic Advisors Estimates Page 8
Real-time and Revised Inflation Page 9
Revised and real-time output gaps Page 10
Actual and implied federal funds rate Page 11
Output Gap with Real-Time Data Orphanides Data Criticized by Taylor (2000) Assume that the Fed used the series produced by the White House Analogous to assuming a can opener Politicized as early as the late 1960s Serious economic analysts like Burns and Greenspan paid no attention to it Page 12
Output Gap with Real-Time Data Detrended Output Gaps from Philadelphia Fed Data Vintages Start in 1965:4 Data Starts in 1947:1 Don t Use ex post Data to Compute Trends First Observation for 1947:1 1966:1 Add Observations Each Quarter until 1979:2 Cecchetti et al (2007) and Levin and Taylor (2009) Use Real-Time HP Detrended Data Page 13
Revised and real-time output gaps Page 14
Actual and implied federal funds rate Page 15
Output Gap with Real-Time Data Real-Time Hodrick-Prescott (HP) Detrending Could Not Have Been Used in the 1970s Technology Didn t Exist Until 1981 Cecchetti et al (2007) Technology to Compute Simpler Trends Did Exist Linear and Quadratic Detrending Are the Output Gaps the Same as with HP Detrending? Is the Taylor Rule the Same as with HP Detrending? Page 16
Linear real-time output gaps Page 17
Quadratic real-time output gaps Page 18
Actual and implied federal funds rate Page 19
Actual and implied federal funds rate Page 20
Real-Time Output Gaps in the 1970s Which Real-Time Output Gap Reflects Real-Time Research Available to Policymakers? Okun s Law y = c (U U * ) U * = Full employment unemployment rate U * = Natural rate of unemployment Use Okun s Law to Calculate Real-Time Output Gaps Need c and U * c = -3.0 in 1970s c = -3.3 in 1962 c = -2.0 in 2009 Page 21
Real-Time Output Gaps in the 1970s What was U * in the 1970s? Official CEA (Real-Time) Estimates U * = 4.0% from 1962-1976 U * = 4.9% in 1977 CBO (Revised) Estimates U * = 5.9% in 1971 U * = 6.2% in 1975 Difference Between CEA and CBO Estimates Productivity Growth Slowdown Not Recognized at Time Shifts in Composition of Labor Markets Well Understood Page 22
Research on U * in the 1970s Source Natural Rate Task Force on Inflation (1969) 4.5% Hall (1970) - BPEA 4.5% Perry (1970) 5.0% Gordon (1971) 5.2% Gordon (1972) 4.8% Hall (1974) 5.5% Modigliani and Papademos (1975) 5.5% Wachter (1976) 5.5% Gordon (1976) CR 5.5% Page 23
Real-Time Output Gaps in the 1970s Best Estimate of U * for 1975 is 5.5 Percent Hall (1974), Modigliani and Papademos (1975), Wachter (1976), Gordon (1976) Real-Time Output Gap Estimates for 1975:3 y = -3.0 (8.8 5.5) = -9.9 percent (Okun s Law) y = -16.2 percent (Orphanides) y = - 5.9 percent (HP) y = -10.8 percent (Linear) y = -10.4 percent (Quadratic) Page 24
Real-Time Output Gaps in the 1970s Best Estimate of U * for 1971 and 1972 is 5.2 Percent Less evidence than for 1975 Gordon (1971, 1972) Real-Time Output Gap Estimates for 1972:1 y = -3.0 (6.0 5.2) = -2.4 percent (Okun s Law) y = -6.7 percent (Orphanides) y = - 0.0 percent (HP) y = -2.2 percent (Linear) y = -2.9 percent (Quadratic) Page 25
Gordon Real-Time Output Gaps Gordon Textbook (1978) Quarterly Natural Rate Estimates from 1947:1 1977:4 Use Methodology from Gordon (1977) Okun s Law Coefficient of -3.0 Gordon Output Gap Estimates Use Real-Time Unemployment Rates Do Not Have Real-Time Natural Rate Estimates Close to Real-Time for 1970s Page 26
Gordon real-time output gaps -10.5% Page 27
Actual and implied federal funds rate Page 28
Return to Initial Question Which Real-Time Output Gap Reflects Real-Time Research Available to Policymakers? Linear Output Gap Most Used Method in the 1970s Does well on Okun s law Metric Quadratic Output Gap Second Most Used Method Also Does Well on Okun s law Metric Page 29
Estimated Taylor Rules The Taylor (1993) rule: Combining terms: Allow for interest rate smoothing: Leads to the following equation: Page 30
Taylor Rule Estimates Page 31
Taylor Rule Estimates Page 32
Taylor Rule Estimates with Forecasts Page 33
Inflation Forecasts Page 34
Taylor Rule Estimates with Forecasts Page 35
Taylor Rule Estimates with Forecasts Page 36
Inflation Forecasts Page 37
Start-Stop Monetary Policy The Taylor Rule with Shifts in the Inflation Target: D1970Q2 and D1976Q1 are time dummy variables. After redefining the variables: ω 1 =φδ 1 and ω 1 =φδ 1. The dummy variables now represent shifts in the intercept. Page 38
Taylor Rule Estimates with Dummies Page 39
Taylor Rule Estimates with Dummies Page 40
Taylor Rule Estimates Page 41
Taylor Rule Estimates Page 42
Taylor Rule Estimates Page 43
Conclusions and Implications The Fed Did Not Follow a Taylor Rule in the 1970s They Did Not Stabilize Inflation They Responded Too Much to Output Gaps Taylor Rule Not Rescued by Shifting Inflation Target The Fed Stabilized Inflation Forecasts Not Good Policy when Forecasts are Too Optimistic Lesson from the 1970s Large Output Gap Coefficient + Stabilizing Overly Optimistic Inflation Forecasts = Too Stimulative Policy Page 44