International spillovers of the Fed and ECB monetary policy surprises

Transcription

1 International spillovers of the Fed and ECB monetary policy surprises Marek Jarociński Are narrative monetary policy shock measures contaminated by global shocks? Michele Ca Zorzi, Massimo Ferrari & Georgios Georgiadis XX Discussion by Franck Portier University College London 1 / 23

2 . Introduction The Holy Grail in Monetary Economics : What is the effect of a monetary policy shock? 2 / 23

3 . Introduction Typical reaction of a typical scholar (me) when looking at the literature: 3 / 23

4 . Introduction 3 / 23

5 . Introduction Two useful papers in this quest Using different identification strategies Narrative approach to monetary shocks high frequency (financial) data around CBks announcements Both addressing at some point the question of international spillovers of monetary policy shocks Both extending our set of observations The two papers are also questioning the validity of the identification strategy they start from 4 / 23

6 Roadmap 1. A Short Historical Tour 2. Discussion 5 / 23

7 1. A Short Historical Tour What do we know about the effect of monetary policy shocks? 6 / 23

8 C 1. A Short Historical Tour Friedman-Schwartz [1965]: The Great Contraction CHART 16 Money Stock, Income, Prices, and VeIocity in Reference Cyde Expansions and Contractions, Rl Income prices F i Scale) Money incone (SCSI. 4 Money stock 5(4.1 3 $ Veloily of money icait Implicit price inde scsi. I Whlyl8 price mdc. 4-3 ; Indu8tr,) pro4uctto SCSI, I I' 1914 ' I D / 23

9 could be based on information that dependent variables pick up any elements of cient) that DMt would be 11 perc regression, equation or (2), lagged to form a time-series of serial dependence adjustment that per year higher than when FEDV anticipated money growth have not been captured bydm. theunanticipated other independent 1. A Short Historical Tour e up to date t - 1, and might also al information Barro [1977]applicable to date t. money growth, DMR -DM-DM, then correvariables. the 1947 value of FEDV = -.34 DMt would be 3 percentage point that becomes available only after sponds the of residuals from thispart equation. The to form the systematic of thethe money than otherwise.6 values of DM and DMR from equation (2) are growth equation is xample, if the estimated values from indicated in Table 1.4 Mt should not be based on any in- ession for the period were in DM for 195, then information to 195 would be used to "predict" money growth. Specifically, later (1) DMt = ao + a1dmt_1 + 2DMt-2 4A measure of the contemporaneous or la the federal government deficit relative to II. Analysis of Unemployment + (3FEDVt + a4 UNt_ nificant when added to equation (2). A lagge inflation rate (based on the GNP deflator) or o where Mt is an annual average of M1 (see fn. 18 rate on prime commercial paper is also insign A. Setup of the Equation 5Based on the fit of the money growth below on the money definition), DMt The effects of monetary expansion on unem- maximum likelihood estimate of the adaptati ( is in the interval between.15 and.2, w hese coefficients would then be ployment are measured by the impact of current age money growth rate, FEDVt- log(fed)t and lagged values of unanticipated money totic 95 percent confidence interval of (. 1, he 195 values of the independent growth, DMR DM-DM. The number of lags in 3 over the interval from.15 to.3, I ha which the unemployment rate enters corre(.15) (2.1) monetary in this of model. Emsponds toshocks the form of type the unemployment sector implies that the typical measured valu ns on (DM, FEDV, UN) would be mate the coefficients of the DM rela- obtain DM for 195. However, it ted that the manner in which later affect earlier values of DM is solely ning down the estimates of the co- log (M)t - log (M)t_ 1 measures the annual aver- [log(fed)]t*, as defined above,3 and UNt1- tion about the money growth struc- pirically, itgiven turned out that the current and two DMRt- 4.2DMRt- 4.7MILt equation below. annual lag values of DMR had significant effects (1.9) (1.9) (.8) that conveyed in prior observations on unemployment. n the DM equation. If individuals unemployment results showed little sensitivit log (U/THE ( - U))t-I, where U is the annual aver- reported results 18 AMERICAN ECONOMIC REVIEW to the case of fivar =.2. to introduce was not established from a priori age unemployment rate in the total labor force 6Note, however, that FEDV has not been reasoning, although Lucas (1975) presents a the long-run average value equal to z (which includes military personnel). The form in makethe Durbin-Watson statistic f (4) log(u/(1 - )t= DMRt normal value of government expenditure is g theoretical rationale for persistence effects of +.95MINWt, 3Data on federal government expenditures are from the e, from the experiences of other Aside from monetary variables, the unemeconomic Report of the President, various issues. The on theoretical grounds), then the use figures on the nominal includes federal budget by ployment equation two "real" vari- (.46) were divided the GNP deflator (1958 = 1.). R2=.78, 6-=.13, D.W. = 1.96, (4) of 1.96 absence of fir distributed lagindicates of actual values, secular gr correlation in the residuals. This be positive. It turns out that constant growth prising, given thegenerate autocorrelated tures at rate g would afedv va which equals 4g at : = From 1949 U-series,19 since a.2. lagged depen average annual growth rate g is.5, so th was not included to soak up the s ponding "long-run average" value of FEDV ever, growth the public sector at 5 pe tion.2 In of fact, if log(u/l - U)t not seem to be permanently sustainable. 8 / 23 equation (4), its estimated coeff

10 1. A Short Historical Tour VARs 4 Federal Reserve Bank of Minneapolis Quarterly Review Summer 1979 A Way to Improve Economic Forecasting tp. n Help for the Regional Economic Forecaster: Vector Autoregression <P. 2) Estimating Vector Autoregressions Using Methods Not Based on Explicit Economic Theories (P. 8) District Conditions (p. 16) 9 / 23

11 MP Shock => Pcom 1. A.15 Short Historical Tour.1 Christiano-Eichenbaum-Evans [1998] MP Shock => Pcom FFR response MP Shock => FF Fed Funds Output MP Model Shock response => with FFM1 MP Shock => Y MP Shock => NBR MP MP Shock Shock => Price => NBR / 23

12 1. A Short Historical Tour -.2 Romer-Romer[24] Figure 3.2B Proxy FFR response Monetary (Ramey) SVAR, Romer, 1983m1 27m12 Output (9% confidence response intervals) Federal Funds Rate Unemployment VOL. 94 NO ROMER AND ROMER: A NEW MEASURE OF Industrial Production FIGURE 2. THE EFFECT OF CPI MONETARY POLICY ON OUTPUT variable is troubling. Closer inspection of the standard about 1 Secon 48 rathe This cha estimate Thereaf increase addition 48 is Third findings gression any of t Section estimate 11 / 23

13 1. A Short Historical Tour Gertkler-Karadi [215] NO.1 Percent nt GERTLER AND KARADI: MONETARY POLICY SURPRISES Federal FFR response funds rate Two-year rate Percent Percent rcent.2 Percent nt CPI.2 Output response One-year rate IP Excess bond Five-year premium rate Percent Percent rcent / 23

14 2. Discussion Marek Monetary shocks derived from high frequency post data around CBk announcements 13 / 23

15 te 2. surprise Discussion Figure 1: Euro area Monetary impulsepolicy responses shock to Fed s shocks, CB information baseline VAR. shock HFI) Marek (Sign restrictions) 1y gov. bond yield (%) Interest rate surprise.1 Monetary policy shock.1 CB information shock (Standard HFI).5 (Sign.5 restrictions).4 Figure 1: US.1 Shocks - Response of Stock Prices Interest rate surprise Monetary 1 policy 2 shock 3 CB information 1 2 shock 3 Stock index (1 x log) orporate bond pread (%) y gov. bond yield (%) Stock index (1 x log) porate bond ead (%) 1y gov. bond yield (%) Stock index (1 x log) orporate bond pread (%) / 23

16 Discussion Marek Figure 2: US.2Shocks.4- Response of GDP Interest rate surprise Monetary policy shock -.4 CB-.2 2 information shock y y Stock index (1 x log) Corporate bond spread (%) Real GDP (1 x log) GDP deflator (1 x log) Stock in (1 x Corporate bond spread (%) Real GDP (1 x log) DP deflator x log) Stock index (1 x log) Corporate bond spread (%) Real GDP (1 x log) 1y y Stock index (1 x log) Corporate bond spread (%) Real GDP (1 x log) DP deflator 1 x log) GDP deflator (1 x log) / 23

17 Discussion Marek Figure 3: EZ Shocks - Response of GDP Interest rate surprise Monetary-.2 policy shock CB information shock CBk information shock :.5 seems.5 to explain.5 a lot of the.5 effect.5 of the interest rate.4.2surprise the monetary policy shock is not so well identified What is the CBk information shock? High order expectations? S&P5 (1 x log) Corporate bond spread (%) Real GDP (1 x log) GDP deflator (1 x log) S& (1 Corporate bond spread (%) Real GDP (1 x log) GDP deflator (1 x log) S&P (1 Corporate bond spread (%) Real GDP (1 x log) GDP deflator (1 x log) S&P5 (1 x log) Corporate bond spread (%) Real GDP (1 x log) GDP deflator (1 x log) 16 / 23

18 2. Discussion Giorgios Monetary shocks derived from the Narrative Approach Seem to include global factors 17 / 23

19 2. Discussion Giorgios B Figures Figure 4: Monetary FigureShocks 1: Narrative Derives monetary From policy theshocks. Narrative Approach 18 / 23

20 2. Discussion Giorgios Figure 3: Response of IP in the EA, US, Canada and the UK to 1% monetary policy shock. Figure 5: Response to Monetary Policy Shock Notes: mean response are estimated with local projection methods as shown in equation (2). Control variables include 6 lags of domestic and foreign IP. 19 / 23

21 2. Discussion Giorgios USA EAR GBR CAN Sample start 1969m1 1999m5 1975m1 1973m1 Sample end 28m6 216m9 27m12 215m1 Table 2: Correlation between narrative shocks USA 1 USA EAR GBR CAN EAR,1 1,29 GBR,,19 1,95,5 CAN,3,1,6 1,52,94,2 No correlated? (add a few leads and lags) At this stage, puzzling... 2 / 23

22 2. Discussion Giorgios Industrial Production is likely to be much more correlated across countries than GDP (because services are less tradable Why not looking at GDPs? 21 / 23

23 3. Conclusion Food for thought I cannot resist wondering if monetary policy shock effects on the economy are not too small to be easily measured? Why do we study monetary policy shocks? CBks should be interested in monetary policy rules... but not much variability Macroeconomists are interested if it is some easy way to measure demand shocks... but it seems not to be that easy / 23

24 23 / 23