Banque nationale de Belgique SA · EURONEXT:BNB
Banque nationale de Belgique SA — Technical Presentation (November 2018)
46 slides from Banque nationale de Belgique SA's technical presentation, Nov 19, 2018
Slide 1: Central Bank Communication and the Yield Curve
CENTRAL BANK COMMUNICATION AND THE YIELD CURVE Banque Nationale de Belgique | 19 November 2018 Matteo Leombroni1 Andrea Vedolin2 Gyuri Venter3 Paul Whelan3 1 Stanford University 2 Boston University & CEPR 3 Copenhagen Business School
Slide 2: Motivation
MOTIVATION
Slide 3: Monetary policy action moves asset prices
Monetary policy action moves asset prices ... 10-May-2001 4.65 2y swap rate 4.6 4.55 Target Rates Announcement 4.5 09:00 12:00 15:00 18:00 03-Nov-2011 1.5 2y swap rate 1.45 1.4 Target Rates Announcement 1.35 09:00 12:00 15:00 18:00 • 10/05/2001: “The Governing Council conducted its regular examinati
Slide 4: But prices can move without any action taken
.. but prices can move without any action taken 06-Apr-2006 3.6 Press Conference Starts 2y swap rate Target Rates Announcement 3.55 Press Conference Ends 3.5 3.45 09:00 12:00 15:00 18:00 05-Jun-2008 5.3 Press Conference Starts Target Rates Announcement 2y swap rate 5.2 5.1 Press Conference Ends 5 09
Slide 5: Bond Yield Changes on ECB Days
Bond Yield Changes on ECB Days 90 peripheral core yield spread 60 in bps 30 0 -30 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 date • Core bond yields move one-for-one with peripheral bond yields on ECB announcement days. 3
Slide 6: Bond Yield Changes on ECB Days
Bond Yield Changes on ECB Days 90 peripheral core yield spread 60 in bps 30 0 -30 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 date • Core bond yields move one-for-one with peripheral bond yields on ECB announcement days. • Since late 2009: Yield spread increases dramatically. 3
Slide 7: This Paper
This Paper • Euro-area is a unique laboratory to study the effect of central bank communication on asset prices for the following reasons: 1. Disentangle target from communication 2. Rich cross-section of asset prices: Direct evidence for risk premia 4
Slide 8: This Paper
This Paper • Euro-area is a unique laboratory to study the effect of central bank communication on asset prices for the following reasons: 1. Disentangle target from communication 2. Rich cross-section of asset prices: Direct evidence for risk premia • In this paper, we argue that central bank commu
Slide 9: Main Finding
Main Finding • Most of the variation of bond yields on monetary policy announcement days is driven by communication shocks (not target shocks). • Dovish monetary policy lowered core yields but peripheral yields are insensitive: Yield spread increases! • We argue that this is due to an emergence of a
Slide 10: Literature review
Literature review Information Effect of Monetary Policy: Campbell, Evans, Fisher, and Justiniano (2012), Campbell, Fisher, Justiniano, and Melosi (2017), Melosi (2017), Nakamura and Steinsson (2018), Paul (2018). −→ Risk premium channel. Effect of ECB announcements on asset prices: Altavilla, Gianno
Slide 11: Model
MODEL
Slide 12: Setup
Setup • World economy with two countries (core and periphery) in a currency union. • Two-period OLG model (t = 0, 1, 2). Agents can invest into four assets: ¬ global risk-free asset (short-term), +® local sovereign bonds (long-term), ¯ world equity index • Exogenous monetary policy: rt+1 = rt + κr
Slide 13: Credit and Breakup Risk
Credit and Breakup Risk • At t = 1, credit event can happen: default of the peripheral country or breakup of currency union. • Event is triggered by random variable Zb,1 that takes value of 1 with probability π0 and is zero otherwise. • Probability of credit event evolves as follows: π0 = π̄ + Zπ,0
Slide 14: Credit and Equity
Credit and Equity • If credit event, bonds’ terminal payoff drops from 1 to e −γi . • E.g., after sovereign default, bond price determined at a CDS auction (Du & Zhu (2017)). • Search/transaction costs, lower liquidity, or different monetary policy. • Interpretation: • Post-financial crisis: γp > γc
Slide 15: Equilibrium Asset Prices
Equilibrium Asset Prices Investors’ maximization problem: X α X max xi,0 (E0 [Ri,1 ] − r0 ) − Var0 xi,0 Ri,1 {xi,0 }i=c,p,s 2 i=c,p,s i=c,p,s Theorem Date-0 equilibrium bond yields are given by 1 1 1 yi,0 = [(2 − κr ) r0 + κr θ0 ] + ασr2 (Sc + Sp ) + γi λπ π0 , |2 {z } |2 {z 2 } expectation
Slide 16: Model Predictions
Model Predictions We run the following type of regressions: ∆yi,0 = αi + βi,r Zr ,0 + βi,θ Zθ,0 + εi,0 , (2) Regression coefficients 1 βi,r = [(2 − κr ) − ηr γi λπ ] , 2 1 βi,θ = [κr − ηθ γi λπ ] , 2 with λπ ≡ 1 + α (γc Sc + γp Sp + γs Ss ) > 0. Proposition The impact of communication shocks in regr
Slide 17: Bond Yields and Risk Premia
Bond Yields and Risk Premia Monetary policy affects bond yields via two different channels: ¬ Expectation channel: Uniform across countries because short-rate is the same. → Negative communication shock implies that future target rates are lower than expected, this will decrease all yields. ⇓ Risk
Slide 18: Identifying Signalling
Identifying Signalling How can we identify whether signalling is present or not? Use effect of monetary policy shocks on stock returns • When signalling is absent, i.e. ηr = ηθ = 0, then Cov[ps,0 , Zr ,0 ] < 0 and Cov[ps,0 , Zθ,0 ] = 0 → Bernanke & Kuttner (2005): An increase in policy rates decreas
Slide 19: Empirical Analysis
EMPIRICAL ANALYSIS
Slide 20: Monetary Policy Shocks
Monetary Policy Shocks t1 target rate decision t2 pc starts pc ends t3 13 : 40 13 : 45 14 : 25 14 : 30 15 : 30 16 : 10 target rate window: ∆yr communication window: ∆yθ 14
Slide 21: Monetary Policy Shocks
Monetary Policy Shocks t1 target rate decision t2 pc starts pc ends t3 13 : 40 13 : 45 14 : 25 14 : 30 15 : 30 16 : 10 target rate window: ∆yr communication window: ∆yθ PCA on swap changes to extract two shocks: ? Target shocks (1st PC in target window) ? Communication shocks (1st PC in communicatio
Slide 22: Data
Data • Tick-by-tick data on • OIS rates with maturities ranging between 1 and 12 months, and • swap rates (written on 6-month Euribor) with a 2-year maturity. • Sample period is Jan 2001 to Dec 2014 • 177 announcements, from which we exclude 14 that were not followed by a press conference and two wh
Slide 23: Monetary Policy Shocks
Monetary Policy Shocks • Calculate PCs in overall, target, and communication windows: PC1 PC2 PC3 Monetary Policy 87.68% 6.56% 2.48% Target 86.36% 5.66% 1.71% Communication 89.86% 4.15% 2.90% • PCA over whole MP window concludes that 2 PCs explain the vast majority of event day variation: • 2 shocks
Slide 24: Shocks Time-Series
Shocks Time-Series Target shocks 10 5 0 basis points -5 6 5 -10 -15 1 -20 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 date Communication shocks 20 2 10 3 basis points 0 -10 4 -20 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 date 1. May 10, 2001: surprise 25bp cut after bad German IP numbers. 2. Jun
Slide 25: Shocks Time-Series (cont'd)
Shocks Time-Series (cont’d) Target shocks 10 5 0 basis points -5 6 5 -10 -15 1 -20 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 date Communication shocks 20 2 10 3 basis points 0 -10 4 -20 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 date 3. March 3, 2011: Trichet announces interest rate hike at nex
Slide 26: Bond yields sensitivity in the cross-section and over time
Bond yields sensitivity in the cross-section and over time ... 1. Bond yield curves for Core (France and Germany) and Periphery (Italy and Spain) 19
Slide 27: Bond yields sensitivity in the cross-section and over time
Bond yields sensitivity in the cross-section and over time ... 1. Bond yield curves for Core (France and Germany) and Periphery (Italy and Spain) 2. Regress zero-coupon yields on our monetary policy shocks: τ τ τ ∆yi,t = βi,r Zr ,t +βi,θ Zθ,t +τi,t , |{z} |{z} target shock comm shock τ where ∆yi,t i
Slide 28: Bond yields sensitivity in the cross-section and over time
Bond yields sensitivity in the cross-section and over time ... 1. Bond yield curves for Core (France and Germany) and Periphery (Italy and Spain) 2. Regress zero-coupon yields on our monetary policy shocks: τ τ τ ∆yi,t = βi,r Zr ,t +βi,θ Zθ,t +τi,t , |{z} |{z} target shock comm shock τ where ∆yi,t i
Slide 29: Estimated coefficient (Communication): pre-crisis
While pre-crisis sensitivities were identical ... τ τ τ ∆yi,t = βi,r Zr ,t + βi,θ Zθ,t +τi,t , |{z} |{z} target shock comm shock Communication: Pre-2009 1.5 Estimated coefficient (Communication) Core Peripheral 1 0.5 0 3m 6m 1y 2y 3y 4y 5y 6y 7y 8y 9y y 10 Maturity 20
Slide 30: Estimated coefficient (Communication): post-crisis
... sensitivities dramatically diverge post-crisis τ τ τ ∆yi,t = βi,r Zr ,t + βi,θ Zθ,t +τi,t , |{z} |{z} target shock comm shock Communication: Post-2009 1.5 Estimated coefficient (Communication) Core Peripheral 1 0.5 0 -0.5 3m 6m 1y 2y 3y 4y 5y 6y 7y 8y 9y y 10 Maturity 21
Slide 31: Regression Coefficients
Regression Coefficients • Rolling regression: Coefficient on the 2y Bond yield Rolling Regression: Communication 2 Estimated coefficient (Communication) 1.5 1 0.5 0 Peripheral Core -0.5 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 date 22
Slide 32: Effect of MP shocks over time (cont'd)
Effect of MP shocks over time (cont’d) Rolling Regression R2 0.8 0.6 0.4 core 0.2 peripheral 0 -0.2 06 07 08 09 10 11 12 13 14 15 date ⇒ Total breakdown in R 2 for peripheral bonds. 23
Slide 33: Effect of MP shocks over time (cont'd)
Effect of MP shocks over time (cont’d) Rolling Regression R2 0.8 0.6 0.4 core 0.2 peripheral core comm only peripheral comm only 0 -0.2 06 07 08 09 10 11 12 13 14 15 date ⇒ All driven by communication shocks. 24
Slide 34: Cumulative Monetary Policy Shocks
Cumulative Monetary Policy Shocks Target Cumulative Shocks Yield Spread due to Communication 15 50 10 basis points 5 40 0 -5 30 -10 -15 basis points 20 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 date Communication Cumulative Shocks 10 50 40 30 basis points 0 20 10 -10 0 -10 -20 -20 01 02 03 04 05
Slide 35: What happened?
What happened? According to ECB President Mario Draghi (July 2012, London): “The short-term challenges relate mostly to .... 26
Slide 36: What happened?
What happened? According to ECB President Mario Draghi (July 2012, London): “The short-term challenges relate mostly to .... 1. “ ... financial fragmentation. Investors retreated within their national boundaries.” 26
Slide 37: What happened?
What happened? According to ECB President Mario Draghi (July 2012, London): “The short-term challenges relate mostly to .... 1. “ ... financial fragmentation. Investors retreated within their national boundaries.” 2. “ ... the premia on sovereign states borrowings. These premia have to do with defau
Slide 38: What happened?
What happened? According to ECB President Mario Draghi (July 2012, London): “The short-term challenges relate mostly to .... 1. “ ... financial fragmentation. Investors retreated within their national boundaries.” 2. “ ... the premia on sovereign states borrowings. These premia have to do with defau
Slide 39: The Information Channel of Central Bank Communication
THE INFORMATION CHANNEL OF CENTRAL BANK COMMUNICATION
Slide 40: Monetary Policy and Equity Returns
Monetary Policy and Equity Returns • Recall that communication shocks can induce a risk premium in the presence of signalling. • We can disentangle signalling from non-signalling days by observing the correlation between communication shocks and equity returns around the announcement. • We construct
Slide 41: Risk Premium Channel Bond Yields
Risk Premium Channel Bond Yields 2y ∆yi,t = αi + βi Zθ,t + γi Dummyt × Zθ,t + δi Dummyt + i,t , • γi measures the effect of communication on signalling days vs non-signalling days. • δi measures the average change in bond yields on signalling vs non-signalling days. Full Pre Post Core Per Spread Cor
Slide 42: Bond Yield Spread on Signalling vs Non-Signalling Days
Bond Yield Spread on Signalling vs Non-Signalling Days Effect of Communication Shocks: All Days Signalling vs Non-Signalling Days 0.5 1 0 Estimated coefficient Estimated coefficient 0 -0.5 -1 -1 -2 -1.5 non-signalling signalling -2 -3 8 0 3 6 12 24 36 48 60 72 84 96 8 0 3 6 12 24 36 48 60 72 84 96 1
Slide 43: Break-Even Inflation
Break-Even Inflation Germany Italy Spread 60 120 60 120 60 120 Zθ 0.47 0.13 -0.86 -0.43 -1.33 -0.57 t-stat (3.22) (0.86) (-0.59) (-0.60) (-0.91) (-0.91) interaction -0.09 -0.15 1.57 -0.22 1.66 -0.07 t-stat (-0.45) (-0.39) (0.63) (-0.13) (0.64) (-0.05) dummy 0.02 -0.01 -0.00 0.11 -0.02 0.12 t-stat (2
Slide 44: Risk Premia on Signalling Days
Risk Premia on Signalling Days Illiquidity CDS CDS quantos Core Peripheral Spread Core Peripheral Spread Core Peripheral Spread Zθ -0.33 0.37 0.71 -0.09 0.44 0.53 0.07 0.09 0.03 t-stat (-4.03) (1.50) (3.55) (-1.47) (0.91) (1.22) (1.76) (2.64) (0.40) interaction -0.70 -0.13 0.57 -1.43 -6.76 -5.33 -0.
Slide 45: Conclusion
CONCLUSION
Slide 46: Conclusion / Thank you very much
Conclusion • Words speak louder than actions... • ... but dovish monetary policy can signal “bad news” to the market. • Core yields ↓ but peripheral yields showed no reaction because accommodating monetary policy (↓) was reversed by increase in risk premium (↑). Thank you very much 32
Banque Nationale De Belgique — Technical Presentation (November 2018) on 6ix