The Forward Guidance Puzzle

Transcription

1 The Forward Guidance Puzzle Marco Del Negro, Marc Giannoni Federal Reserve Bank of New York Christina Patterson MIT Penn State University, February 24, 216 Disclaimer: The views expressed here do not necessarily reflect those of the Federal Reserve Bank of New York or the Federal Reserve System

2 Introduction Forward Guidance Announcements about future interest rate changes: key instrument of monetary policy at least since 28 (also before see Campbell et al. 212) 1 What are the effects of forward guidance? On financial markets On expectations Evidence from Blue Chip surveys 2 Can its effects be captured by standard medium-scale DSGE models? No! Forward Guidance Puzzle: Excessive response of output and inflation The farther into the future is the change in FFR, the stronger the economy s response 3 A proposed resolution to the FG puzzle Accounting for finite life: Blanchard-Yaari s perpetual youth in a medium-scale DSGE model Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 2

3 Introduction Pre-Great Recession Transmission of Monetary Policy Key instrument of policy: short-term interest rate Monetary transmission well understood (extensively studied using both VAR models and DSGE models) Post-Great Recession New policy tools: Forward guidance (FG), LSAPs Goal at ZLB: lower long-term bond yields stimulate aggregate expenditures But... effects not well understood; harder to quantify using existing empirical tools (e.g. VARs) Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 3

4 Introduction Analyzing the Effects of Forward Guidance The Challenge Announcement by CB that will maintain FFR at ZLB for longer can have two effects (Campbell, Evans, Fisher, Justiniano 212; Woodford 212): 1 More monetary stimulus (Odyssean/Commitment a la Eggertsson and Woodford 23) stimulates economic activity, higher inflation 2 Reveals bad news about state of economy (Delphic) lower projected activity, lower inflation Interpretation by market depends in very subtle ways on FOMC communication Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 4

5 Introduction Analyzing the Effects of Forward Guidance The Challenge Like Odysseus, central bank commits to keeping FFR low despite temptation to raise FFR once the economy is recovering Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 4

6 Introduction Analyzing the Effects of Forward Guidance The Challenge Announcement by CB that will maintain FFR at ZLB for longer can have two effects (Campbell, Evans, Fisher, Justiniano 212; Woodford 212): 1 More monetary stimulus (Odyssean/Commitment a la Eggertsson and Woodford 23) stimulates economic activity, higher inflation 2 Reveals bad news about state of economy (Delphic) lower projected activity, lower inflation Interpretation by market depends in very subtle ways on FOMC communication Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 4

7 Introduction Analyzing the Effects of Forward Guidance The Challenge Like the Oracle of Delphi, central bank announces a low forecast for FFR, given its forecast of weak economic conditions Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 4

8 Introduction Analyzing the Effects of Forward Guidance The Challenge Announcement by CB that will maintain FFR at ZLB for longer can have two effects (Campbell, Evans, Fisher, Justiniano 212; Woodford 212): 1 More monetary stimulus (Odyssean/Commitment a la Eggertsson and Woodford 23) stimulates economic activity, higher inflation 2 Reveals bad news about state of economy (Delphic) lower projected activity, lower inflation Interpretation by market depends in very subtle ways on FOMC communication Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 4

9 Introduction DSGE Models Suited to Analyze Forward Guidance? Medium-scale New Keynesian DSGE models fit data well Models are structural in principle well suited to perform counterfactual experiments Problem: Model-implied response to FG much larger than observed Forward Guidance Puzzle! Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 5

10 Introduction DSGE Models Suited to Analyze Forward Guidance? Medium-scale New Keynesian DSGE models fit data well Models are structural in principle well suited to perform counterfactual experiments Problem: Model-implied response to FG much larger than observed Forward Guidance Puzzle! Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 5

11 1 Effects of Forward Guidance: Empirical evidence Evidence from Blue Chip Financial Forecasters Compute change in forecasts in a one-month window around the announcement... controlling for: all macro economic news (surprises) asset price movements (ex event window) Panel regression for variable (k), horizon (h), forecaster (i): f (k, h) t,i = γ + γ 1 Macro news + γ 2 Asset Price Changes +γ 3 i-specific control + β Announcement Dummy + ɛ i,t for t = 28.6,.., Std errors corrected for correlation across i s and heteroskedasticity Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 6

12 1 Effects of Forward Guidance: Empirical evidence August exceptionally low levels of the FFR at least through mid-213 Projections for 3-month rates and 1-year yields decline Change in forecasts of financial variables in line with asset response in two-day window Forecasters believe the FOMC announcement.5 3-Month TBill 1 1 Treasury Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 7

13 1 Effects of Forward Guidance: Empirical evidence Evidence from Financial Markets: Forward Rates Pre-FOMC: solid; post-fomc: dashed FF fut. (purple); Eurodol. fut. (blue); Fwd rates from yield curve (red) 5. 8/9/11 1/25/12 9/13/12 at least at least mid-15..after the through mid-13 Yield Curves - 8/9/211 through Yield late-14 Curves - 1/25/212 economy strengthens Yield Curves - 9/13/ Fi>ed Forward Yield (before) 2. Fi>ed Forward Yield (before) Fi>ed Forward Yield (acer) Fi>ed Forward Yield (acer) Yield (%) 2.5 Yield (%) 2. Eurodollar Futures: Swap Basis Adjusted (before) Yield (%) 1.5 Eurodollar Futures: Swap Basis Adjusted (before) Eurodollar Futures: Swap Basis Adjusted (acer) Fed Funds Futures (before) FF Futures Yield (acer) 1. Eurodollar Futures: Swap Basis Adjusted (acer) Fed Funds Futures (before) FF Futures Yield (acer) Year 2- Year 3- Year 4- Year 5- Year 6- Year 7- Year 8- Year. 9- Year 1- Year 1- Year 2- Year 3- Year 4- Year 5- Year 6- Year 7- Year 8- Year. 9- Year 1- Year 1- Year 2- Year 3- Year 4- Year 5- Year 6- Year 7- Year 8- Year 9- Year Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 8

14 1 Effects of Forward Guidance: Empirical evidence Evidence from Financial Markets As in KVJ (11): look at cross-section of financial markets data Maturity (years) Femia et al. 213, Raskin 213, Filardo and Hoffman 214, Moessner 213,... Changes in bond yields in 2-day following FOMC meeting Treasury Yields (constant maturity) Agency Yields (Fannie/Freddie) MBS Yields /9/ /25/ /13/ Notes: All figures are in basis points unless otherwise noted. Bond yields fall in Aug. 211 and Jan. 212; increase in Sept. 212 Fed announcements affect yields: Hard to reconcile with lack of credibility story Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 9

15 1 Effects of Forward Guidance: Empirical evidence August economic growth so far this year has been considerably slower than... expected....the Committee now expects a somewhat slower pace of recovery over coming quarters... economic conditions... are likely to warrant exceptionally low levels of the FFR at least through mid-213 GDP Growth CPI Inflation Possible example of Delphic forward guidance: bad news about the economy Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 1

16 1 Effects of Forward Guidance: Empirical evidence September highly accommodative stance... will remain appropriate for a considerable time after the economic recovery strengthens.... at least through mid-215 GDP Growth CPI Inflation Odyssean: significant increase in forecasts for real activity and inflation Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 11

17 1 Effects of Forward Guidance: Empirical evidence Evidence from Financial Markets TIPS (constant maturity) Implied Vol. SP 5 DJ IA FX USD/EUR Maturity (% change) (% change) (years) 8/9/ /25/ /13/ Notes: All figures are in basis points unless otherwise noted. Real rates fall Stocks prices: modest changes in Aug. 211 and Jan. 212; larger increases in Sept. 212 Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 12

18 1 Effects of Forward Guidance: Empirical evidence Breakevens Inflation Swaps Liquidity Premium Maturity (years) /9/ /25/ /13/ Notes: All figures are in basis points unless otherwise noted. Inflation breakeven and Inflation swaps increase especially in Sept. 212 Little variation in liquidity premium (TIPS-Treasury spread, Fleckenstein et al.) Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 13

19 1 Effects of Forward Guidance: Empirical evidence Corporate Yields Intermediate term Long term Aaa Aa A Baa Ba B Aaa Aa A Baa Ba B 8/9/ /25/ /13/ Notes: All figures are in basis points unless otherwise noted. While high-grade yields decrease in August 211 in line with Treasuries, low-grade corporate yields increase (safety premium ) Low-grade corporate yields fall in Sept 212 (safety premium ) Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 14

20 1 Effects of Forward Guidance: Empirical evidence Evidence from Financial Markets August 211: Bond yields and real rates fall; little change in stocks prices Inflation breakeven and inflation swaps increase slightly January 212: Financial market response similar to that of August 211, but more modest September 212: Different response Real yields fall But bond yields rise with inflation breakeven and inflation swaps; stock market rises Sept. 212: Could be consistent with Odyssean forward guidance: monetary policy more accommodative than expected and provides more stimulus... highly accommodative stance of monetary policy will remain appropriate for a considerable time after the economic recovery strengthens. What happened to output forecasts? Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 15

21 1 Effects of Forward Guidance: Empirical evidence Effect of Different Aspects of the FOMC Statement Add dummies for announcements of: Forward guidance episode QE Continuation of QE Output conditions Inflation conditions Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 16

22 1 Effects of Forward Guidance: Empirical evidence Effect of Different Aspects of the FOMC Statement Forward Guidance QE Announcement Bad Output Language GDP Growth Month TBill Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 17

23 2 Forward Guidance in DSGE Models The Forward Guidance Puzzle Medium-scale DSGE Good forecasting performance In principle well suited for counterfactual experiments 212Q2 experiment : FFR kept at ZLB through 215Q2 See also Carlstrom et al. 212 Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 18

24 2 Forward Guidance in DSGE Models The Two Legs of the Forward Guidance Puzzle 1: Consumption depends on the expected future short-term real rates: ĉ t = IE t [ˆR t ˆπ t+1 + ĉ t+1 ] = ĉ t = IE t [ˆR t+j ˆπ t+1+j ] }{{} j= ˆr t+j Contemporaneous shock: ˆr t ĉ t, ĉ t+1 =,... Anticipated shock: ˆr t+h ĉ t, ĉ t+1,..., ĉ t+h The farther the rate drop, the longer does consumption boom last (McKay, Nakamura, Steinsson, 215) 2: Now let π move. NK Phllips curve implies ˆπ t = κ β j IE t [ĉ t+j ] Anticipated shock: more prolonged consumption boom = ˆπ t, ˆπ t+1,... rises more = real rate drops even more today = consumption increase amplified j= Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 19

25 2 Forward Guidance in DSGE Models The Two Legs of the Forward Guidance Puzzle 1: Consumption depends on the expected future short-term real rates: ĉ t = IE t [ˆR t ˆπ t+1 + ĉ t+1 ] = ĉ t = IE t [ˆR t+j ˆπ t+1+j ] }{{} j= ˆr t+j Contemporaneous shock: ˆr t ĉ t, ĉ t+1 =,... Anticipated shock: ˆr t+h ĉ t, ĉ t+1,..., ĉ t+h The farther the rate drop, the longer does consumption boom last (McKay, Nakamura, Steinsson, 215) 2: Now let π move. NK Phllips curve implies ˆπ t = κ β j IE t [ĉ t+j ] Anticipated shock: more prolonged consumption boom = ˆπ t, ˆπ t+1,... rises more = real rate drops even more today = consumption increase amplified j= Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 19

26 2 Forward Guidance in DSGE Models Possible Resolutions 1 The Euler equation? McKay, Nakamura, Steinsson (215), Caballero and Fahri (214) Here: Discounting in the Euler equation coming from overlapping generations Werning (215) 2 The NKPC? Kiley et al. 214, Carlstrom et al Lack of credibility? At odds with surveys and financial markets responses 4 Deviations from rational expectations? Gabaix (215), Garcia-Schmidt, Woodford (215) Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 2

27 3 A Proposed Resolution A Proposed Resolution: Finite Life (Blanchard-Yaari) Agents face probability p of dying (β(1 p)) s log(c j,t+s ) s= Life-insurance companies offer an annuity contract individual wealth accumulates at R/(1 p) S j,t+1 = Individual EE for each cohort j: where H t = s= Aggregate EE: R t 1 p (S j,t + Y t C j,t ) C j,t+1 = βr t C j,t C j,t = (S j,t + H t ) 1 β(1 p) Y t+s s 1 l= (R t+l /(1 p)) C t+1 = R t βc t p(1 β(1 p)) S t+1 (1 p) Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 21

28 3 A Proposed Resolution A Proposed Resolution: Finite Life (Blanchard-Yaari) Agents face probability p of dying (β(1 p)) s log(c j,t+s ) s= Life-insurance companies offer an annuity contract individual wealth accumulates at R/(1 p) S j,t+1 = Individual EE for each cohort j: where H t = s= Aggregate EE: R t 1 p (S j,t + Y t C j,t ) C j,t+1 = βr t C j,t C j,t = (S j,t + H t ) 1 β(1 p) Y t+s s 1 l= (R t+l /(1 p)) C t+1 = R t βc t p(1 β(1 p)) S t+1 (1 p) Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 21

29 We A Proposed Resolution A Proposed Resolution: Finite Life (Blanchard-Yaari) -.1 Announced future drop in R. Death -.1 probability: p =, p > Counsumption Individual Response Aggregate Response Individuals: consumption, wealth (standard Euler eq) But unborn cohorts cannot react to the announcement In the aggregate, C increases as it gets closer to drop in R (as newborn cohorts react) Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 22

30 3 A Proposed Resolution Smets-Wouters Model with Blanchard-Yaari Households Aggregate consumption Euler equation (simplified): ) ĉ t = (ˆRt IE t [ˆπ t+1 ] + (1 η)ie t [ŝ t+1 ] + ηie t [ĉ t+1 ] where η < 1 when p > Evolution of wealth ŝ t and fiscal policy All other equations are the same as in SW (with β = ηβ), e.g. NK Phillips Curve: π t = E t β j κ mc t+j j= SWBY: Tractable medium scale DSGE Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 23

31 3 A Proposed Resolution Calibration of p: Does it Matter Quantitatively? Average death prob. (Soc. Sec.) per quarter:.4% to.8% In addition, can loosely think of p as the probability of entering/exiting hand-to-mouth status (e.g. bankruptcy,.., from Kaplan, Violante, Wieder 214: 2.3%) Baseline: p = 3%; alternative: p = 6% All other parameters taken from Smets and Wouters Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 24

32 3 A Proposed Resolution Contemporaneous drop in FFR Response to contemporaneous shock similar for p =, 3% or 6% Interest rate (% annualized) p =, η = 1 p =.3, η =.987 p =.6, η =.96 Horizon Inflation (% annualized) Consumption (level, %).14 Output (level, %) Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 25

33 3 A Proposed Resolution Announcement of FFR drop in 8 quarters Interest rate (% annualized) p =, η = 1 p =.3, η =.987 p =.6, η =.96 Horizon Inflation (% annualized) Consumption (level, %) 8 Output (level, %) With p = : FG causes huge changes in output and inflation With p = 3%, response of output and inflation cut by 2/3 Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 26

34 3 A Proposed Resolution Estimated Model Interest rate (% annualized) p =, η = 1 p =.3, η =.99 p =.6, η =.971 p =.145, η =.899 Horizon Inflation (% annualized) Consumption (level, %).4 Output (level, %) r = 1/ β is fixed across simulations Very preliminary results! Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 27

35 Conclusions Conclusions 1 What are the effects of forward guidance? Stimulative, non-trivial, but not huge 2 Can its effects be captured by standard medium-scale DSGE models? No! Estimated DSGE model delivers implausibly large responses to forward guidance 3 A proposed resolution to the forward guidance puzzle Blanchard-Yaari Compositional effects imply discounting in the Euler equation = mitigate aggregate response Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 28

36 Reference Slides Reference Slides Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 29

37 Reference Slides September highly accommodative stance... will remain appropriate for a considerable time after the economic recovery strengthens.... exceptionally low levels for the FFR are likely to be warranted at least through mid Month TBill 1-Year Treasury Long term-rates increase (in line with market reaction) Del Negro, Giannoni, Patterson Forward Guidance Puzzle Penn State 3

38 Fwd Guidance Puzzle and Effects of Changes in the Reaction Function Excessive response of output and inflation as well Note: Nominal rates can in equilibrium following an announcement about the reaction function (consistent with 9/13/12) 6 Federal Funds Rate 4 Ex Ante Real Interest Rate Inertial Taylor 99 Baseline Forecast Output Growth 12 Inertial Taylor 99 Baseline Forecast Inertial Taylor 99 Baseline Forecast Inflation (GDP Deflator, Q/Q Annualized) 7 Inertial Taylor 99 Baseline Forecast