Mainstream Model of U.S. Inflation Dynamics

2/12/20081The Mainstream Model of U. S. Inflation Dynamics: Explaining the Past and Addressing the FutureJonathan W. Eller and Robert J. GordonMeeting of Economic AdvisorsChicago Fed, June 16, 20032/12/20082The Phillips Curve, discovered in 1958, has been Taking a Beating Since 1968•The first generation 1958-68 Phillips Curve was attacked by the Friedman-Phelps NRH•The second generation 1968-75 Phillips Curve incorporated NRH–but was attacked by Lucas-Sargent, who left Philips Curve (and by association all of Keynesian economics) “lying in wreckage” –because U-p correlation in mid-1970s was positive not negative2/12/20083Phillips Curve Survived•Positive vs. Negative Correlation: Gordon-Phelps Model showing Macro Externalities of Supply Shocks. –Macro is about Demand and Supply, not either alone–Textbooks embedded this as basic macro in 1978•Mainstream model since 1980: Key elements the “triangle” of inertia, demand, and supply•Staiger-Stock-Watson (1997) and Gordon (1997) merged time-varying NAIRU with Gordon’s 1982 “mainstream” model of U. S. inflation dynamics2/12/20084Meanwhile in the past five years, the NKPC has dominated macro journals and conferences•Attempt to implement specific theoretical models of sticky prices (Calvo, 1983)•Central component: Key Driving Force is “Expected Future Inflation”•Second variable: Two Versions, output gap vs. marginal cost•Not discussed further today–Rejected empirically, triple the SSR–Rejected, 3X to 10X post-sample simulation errors2/12/20085From Theory to Practice: The “Missing Link” in Macroeconometrics•In Applications Where the Driving Force is Lagged Dependent Variable, can’t test by Goodness of Fit (Oh, boy, my model shows inflation is a random walk)•Instead: Post-Sample Dynamic Simulations–No Data from Simulation Period used to Estimate Coefficients–No Information on Actual Behavior of LDV is Used to Calculate Predicted Inflation During Simulation Period2/12/20086Basic Data Show Negative and Positive U,p CorrelationFigure 1. Unemployment Rate and Four Quarter Rate of Change of PCE Deflator, 1960:Q1-2002:Q4 024681012196019631966196919721975197819811984198719901993199619992002PCE DeflatorUnemployment2/12/20087Quick Introduction to the Triangle Model•“Mainstream” vs. “Triangle” •The Triangle model in equation (1). Absorb this notation. –Long lags on past inflation–D(t) defined relative to zero, z(t) relative to zero•The SSW and Gordon 1997 version: Estimating the TV-NAIRU while using overall framework equations (2) and (3)2/12/20088We will look at a graph for the Unemployment Gap Later: Here are the Supply Shocks•Change in Relative Price of Imports, zero when no effect•Food-Energy Effect, change in headline PCE inflation minus core PCE inflation•Medical Care Effect, defined same•NOTICE BIG DIFFERENCES IN VERTICAL AXIS!2/12/20089Supply Shock Variables, all equal to zero when no SS Effect: Start with Relative Price of Non-oil ImportsRelative Price of Imports-15-10-505101520251960196319661969197219751978198119841987199019931996199920022/12/200810Food-Energy Effect: Headline PCE Deflator minus Core PCEFood-Energy Effect-2-1012341960196319661969197219751978198119841987199019931996199920022/12/200811Medical Care Effect defined same as FaE: Notice Crash in 1993-96Medical Care Effect-0.2-0.100.10.20.30.40.50.61960196319661969197219751978198119841987199019931996199920022/12/200812Productivity: Impose Real Wage Aspirations or (later) introduce directlyHP Trend of Nonfarm Private Business Productivity Growth minus a Six-year Moving Average of That Trend00.511.522.533.54196019651970197519801985199019952000HP Trend6-Year MA ofTrend2/12/200813Basic Triangle Model Estimates: Table 2•All Coefficients or Sums of Coefficients Significant at 1 percent level, correct signs•Post-sample simulations 1993-2002–No help from: coefficients or lagged p data–Look at errors for PCE deflator•In comparing with other models or other results, remember SEE of 0.69, SSR of 64.42/12/200814Estimated TV-NAIRUs: Crucial Role of SS VariablesFigure 5The Actual Unemployment Rate and TV-NAIRUs for the GDP and PCE Deflators and for the CPI-RS, 1961:Q1-2002:Q434567891011196019651970197519801985199019952000PercentGDP NAIRUPCE NAIRUCPI NAIRUUnemployment2/12/200815Tests of Robustness & Stability•Significance of LDV lags 13-24•Stability Over Sample Split–1962-81, 1982-2002•Rejection of Shift toward zero of PC coefficient in 1990s2/12/200816Possible Criticism: Productivity Variable (Figure 4 bottom) is ad hoc•Add real wage feedback directly•Sims (1987): wage equation must contain all the same variables as the price equation•See equation (9)without price feedback and (10)with price feedback2/12/200817How to Get Rid of Productivity Variable, Enter Real Wage Feedback Directly?•Equation (12)–Feedback entered through TULC–Can run dynamic simulations with endogenous wage via equation (11)•Results in Table 6–wage feedback replaces productivity trend variable–So we didn’t need the ad hocproductivity variable –wage feedback is explicit and does better2/12/200818Dramatic Improvement for GDP Deflator, less for PCE DeflatorGDP Deflator00.511.522.533.544.51987198819891990199119921993199419951996199719981999200020012002ActualTriangle Simulation Without FeedbackTriangle Simulation With FeedbackSimulatedPredicted2/12/200819Concluding Numbers for the Paper: 1993-2002 Dynamic Simulation for PCE Deflator•Actual 4-quarter inflation 2002:Q4 = 1.80•Simulated fixed NAIRU no SS = 4.87•Difference 3.07–2.38 Contribution SS Variables–0.71 Decline in TV-NAIRU•Residual 0.072/12/200820Triangle Model is Validated•Stability of model and coefficients over 164 quarters•Ability to track inflation 10 years after end of sample period with no help from–Coefficients estimated for 1993-2002–Help from actual LDV over 1993-2002•Unified Framework in which Wage-Price-Wage Feedback is Handled Symmetrically2/12/200821Baseline Forecasts of Future Inflation•Look only at Performance for PCE Deflator–Why? Alan Greenspan’s Favorite–Why? It Does the Best in our 1993-2002 Simulations•Start with Sensitivity to Future Unemployment Rates–Assume TV-NAIRU continues forever at 5.33–CRUCIAL: Assume NO supply shocks at all!2/12/200822How Long Does It Take Model to Generate Deflation?Simulations begin in 2002:4, different effects of GAP in absence of supply shocks-1.5-1-0.500.511.522.533.519931994199519961997199819992000200120022003200420052006200720082009201020112012actualunemployment = 5.33unemployment = 6.00unemployment = 6.67actual unemployment 2003:1 and 2003:2,slowly decreasing so that 2005:1 = 5.33Fitted valuesSimulations2/12/200823Sensitivity to Long Lags on Past Inflation•Table 1: Including lags 1-24, 1-12, 1-4•Intuitively, the longer are the lags, the more inertial is the model in generating deflation or accelerating inflation•Sensitivity to shorter lags . . . .2/12/200824Three deflation scenarios with no supply shocksEffect of shortening LDV, sample period 1982-2002-5-4-3-2-101234199319951997199920012003200520072009201124 lags LDV, U = 5.3324 lags LDV, U = 6.6712 lags LDV, U = 6.674 lags LDV, U = 6.672/12/200825Back to Basic Model, What Happens if Oil Prices Retreat•Say Oil Prices go from $31 to $19 over next six quarters•Two unemployment scenarios–U = 6.0 percent forever–U = 6.67 percent forever2/12/200826Here’s What Oil Prices Do to the Food-Energy EffectFood and Energy effects-1.5-1-0.500.511.519871989199119931995199719992001200320052007200920112/12/200827Deflation Comes Sooner!Simulations begin in 2002:4, Unemployment 6.00 unless otherwise indicated, non-zero supply shocks-2-1.5-1-0.500.511.522.533.519931994199519961997199819992000200120022003200420052006200720082009201020112012actualunemployment = 6.00Food and Energy effects reflecting largedrop in oil pricesSame Food and Energy Effect, U = 6.67Fitted valuesSimulations2/12/200828Opposite: What About a pro-Inflation Scenario?•U returns from 6.1 to 5.33 by 2005:Q1•Import prices reverse all of benefit of 1995-2002 dollar appreciation over the next seven years•RelativeInflation in Medical Care Prices Continues at rate of 2002:Q42/12/200829Here’s What Import Prices DoRelative Import Prices-10-8-6-4-2024619871989199119931995199719992001200320052007200920112/12/200830Here’s What Medical Care Prices DoMedical Care 00.10.20.30.40.50.619871989199119931995199719992001200320052007200920112/12/200831High Inflation Scenario, U returns to 5.33, Import Price and Medical as shown, no Oil Price RetreatSimulations begin 2002:4, uses non-zero supply shocks, except for Food and Energy effects,Unemployment gradually declines from 2003:2 level to 5.33 in 2005:100.511.522.533.519931994199519961997199819992000200120022003200420052006200720082009201020112012actualEffect of Projected U Rate"Pro Inflation" ScenarioFitted valuesSimulations2/12/200832What Does Import Price vs. Exchange Rate Relationship Look Like?C4 Import Prices and Exchange rate-15-10-5051015202519741974.81975.51976.319771977.81978.51979.319801980.81981.51982.319831983.81984.51985.319861986.81987.51988.31989199.81990.51991.319921992.81993.51994.319951995.81996.51997.319981998.81999.52000.320012001.82002.5C4 ImportsC4 Exchange2/12/200833Conclusion: Deflation is Remote•Why?–Unprecedented stimulus will drive U back to NAIRU–Decline in dollar has already happened–Real Medical Care inflation is not going away–Our leader recognizes that the natural gas part of the energy equation is dire–Sabotage to Iraqi oil fields and the absence of a postwar plan: “OK, we win, what then?!”