[PPT] - Deconstructing Monetary Policy Surprises - The Role of Information PowerPoint Presentation

SLIDE 1

Deconstructing Monetary Policy Surprises - The Role of Information Shocks

Marek Jaroci´ nski and Peter Karadi

European Central Bank FRBNZ Macro-Finance Conference, December 2018 The views expressed here are solely those of the authors and do not necessarily reflect the views of the ECB

SLIDE 2

What we do?

We study how financial markets react to central bank policy an-

nouncements, within the first half-hour. We separate the effects of – news about monetary policy and – news about the central bank’s outlook on the economy.

We track the response of the economy, in the US and the euro area,

using a vector autoregression (VAR).

1/24

SLIDE 3

What we find

Market reactions reflect both news about monetary policy and news

about the economy. Variance shares: US: 65:35, EA: 55:45.

The responses of the economy are very different.

Surprise interest rates increases are ...

contractionary (output, prices decline) when reflecting news about

monetary policy (monetary policy shock),

expansionary (output, prices increase) when reflecting news about

the economy (central bank information shock).

2/24

SLIDE 4

Implications

1. Private agents learn something about the economy (not just about

monetary policy) from central bank annoucements.

2. These news about the economy attenuate the standard estimates of

monetary policy effects.

3/24

SLIDE 5

Plan of the presentation

Key data
Selected literature
VAR, Identification, IRFs
A structural DSGE interpretation

4/24

SLIDE 6

Key data: “surprises” p - price of a financial asset τ - time of a central bank announcement surprise: m = p(τ + 20min) − p(τ − 10min) We compute m for

interest rate derivatives (US: fed funds futures, Euro area:

Eonia swaps). These instruments include also near term forward guidance.

stock prices (US: S&P500, Euro area: EuroStoxx50)

5/24

SLIDE 7

Motivating example: FOMC announcement on March 20, 2001, 2:15pm For immediate release The Federal Open Market Committee at its meeting today decided to lower its target for the federal funds rate by 50 basis points to 5 percent. (...) [The Committee’s analysis] (...) suggests substantial risks that demand and production could remain soft. → fed funds futures and stock prices both drop between 2:05pm and 2:35pm

6/24

SLIDE 8

Expected effect on stock prices

Textbook asset pricing: A monetary policy surprise → negative co-

movement between interest rates and stock markets

lower fed funds rate:

→ lower credit costs, higher demand → higher future dividends → lower discount rate ⇒ present discounted value of dividends goes up = stock price goes up

7/24

SLIDE 9

An empirical observation Stock prices do not always go up after a surprise interest rate cut.

noise?
information

about the economy in the central bank announcement?

Surprise in the 3m FF futures

0.3
0.2
0.1

0.1 0.2 0.3

Surprise in the S&P500

1

1 2 3 4 I: 17 II: 66 III: 41 IV: 50 3/20/2001

8/24

SLIDE 10

An empirical observation Stock prices do not always go up after a surprise interest rate cut.

noise?
information

about the economy in the central bank announcement?

Surprise in the 3m FF futures

0.3
0.2
0.1

0.1 0.2 0.3

Surprise in the S&P500

1

1 2 3 4 Monetary Policy shock CB Information shock

8/24

SLIDE 11

Further investigation

We separate the shock that makes interest rates and stock prices co-

move negatively (as they should after Mon.Pol. shock) from the shock that makes them co-move positively.

We use a Structural VAR with a mix of high-frequency identification

and sign restrictions.

We study how the economy responds.

9/24

SLIDE 12

Selected literature

Interest rate surprise ≈ monetary policy shock: Kuttner, 2001; G¨

urkaynak, Sack, Swanson, 2005; Barakchian, Crowe, 2013; Gertler, Karadi, 2015 → we add the central bank information shock

Measurement of CB information shocks: Campbell et.al., 2016; Miranda-

Agrippino, 2016; Lakdawala, Schaffer, 2016 – use Fed private info; Hansen and McMahon, 2016 – textual analysis of statements → we use the markets as Andrade, Ferroni, 2016, Cie´ slak, Shrimpf, 2018, Kerssenfischer, 2018

Models of the information channel of monetary policy: Nakamura and

Steinsson, 2018 QJE; Melosi, 2017 REStud → we add communication policy, use VAR for estimation

10/24

SLIDE 13

Monthly VAR - data (US)

7 × 1 vector
mt

yt

where

– mt (2 × 1): interest rate and stock price surprises that occurred in month t source: updated G¨ urkaynak, Sack and Swanson (2005) dataset – yt (5 × 1): standard macroeconomic and financial variables in month t → government bond yields, S&P500, real GDP and GDP deflator (interpolated using Kalman filter, Stock and Watson 2010), Ex- cess bond premium (Gilchrist and Zakrajsek, 2012)

11/24

SLIDE 14

VAR with surprises – restriction mt - surprises (monthly), yt - macroeconomic variables (monthly) mt are i.i.d.:

mt

yt

=

P

p=1
Bp

Y M

Bp

Y Y

mt−p yt−p

+
cy
+
um

t

uy

t

Bayesian estimation with a Minnesota prior on the nonzero parameters
Within the Gibbs sampler we also draw the missing observations on

mt

12/24

SLIDE 15

Our identification (Restrictions on the impact responses)

shock variable Monetary policy CB information

ther

(negative co-movement) (positive co-movement) Interest rate surprise (m1

t)

+ + Stock price surprise (m2

t)

– + All other variables (yt)

Two surprises: interest surprise and stock price surprise
Zero restrictions: other shocks are unlikely to occur systematically in

the narrow time window around the announcements

Sign restriction on the co-movement of surprises
No restrictions on yt

13/24

SLIDE 16

Sign restrictions

mt is decomposed into two orthogonal shocks (MP,CBI)
Set identification: uniform prior on rotations, provided that they are

consistent with sign restrictions. Robustness check: ‘poor man’s sign restrictions’

Quadrants II and IV: MP shocks; quadrants I and III: CBI shocks

14/24

SLIDE 17

For comparison: Standard high-frequency identification (HFI)

shock variable Interest rate surprise

ther

Interest rate surprise (m1

t)

+ All other variables (yt)

Single surprise: interest rate
Zero restrictions: other shocks are unlikely to occur systematically in

the narrow time window around the announcements

Interest rate surprise ≈ monetary policy shock.

15/24

SLIDE 18

United States impulse responses

A. Standard HFI

Interest rate surprise

10 20 30 0.05

surprise in 3m ff futures

10 20 30

0.1

0.1 0.2

1y gov bond yield (%)

10 20 30

2
1

1

S&P500 (100 x log)

10 20 30

0.2

0.2

Real GDP (100 x log)

10 20 30

0.1

0.1

GDP deflator (100 x log)

10 20 30

0.05

0.05

EBP (%)

months

SLIDE 19

United States impulse responses

A. Standard HFI
B. Sign restrictions

Interest rate surprise

Monetary Policy CB information (negative correlation) (positive correlation)

10 20 30 0.05

surprise in 3m ff futures

10 20 30

0.1

0.1 0.2

1y gov bond yield (%)

10 20 30

2
1

1

S&P500 (100 x log)

10 20 30

0.2

0.2

Real GDP (100 x log)

10 20 30

0.1

0.1

GDP deflator (100 x log)

10 20 30

0.05

0.05

EBP (%)

10 20 30 0.05

surprise in 3m ff futures

10 20 30 0.05 10 20 30

0.4
0.2

0.2 0.4

surprise in S&P500

10 20 30

0.4
0.2

0.2 0.4 10 20 30

0.1

0.1 0.2

1y gov bond yield (%)

10 20 30

0.1

0.1 0.2 10 20 30

2
1

1

S&P500 (100 x log)

10 20 30

2
1

1 10 20 30

0.2

0.2

Real GDP (100 x log)

10 20 30

0.2

0.2 10 20 30

0.1

0.1

GDP deflator (100 x log)

10 20 30

0.1

0.1 10 20 30

0.05

0.05

EBP (%)

10 20 30

0.05

0.05

months months months

SLIDE 20

United States impulse responses - zooming in on yt

A. Standard HFI
B. Sign restrictions

Interest rate surprise

Monetary Policy CB information (negative correlation) (positive correlation)

10 20 30

0.1

0.1 0.2

1y gov bond yield (%)

10 20 30

2
1

1

S&P500 (100 x log)

10 20 30

0.2

0.2

Real GDP (100 x log)

10 20 30

0.1

0.1

GDP deflator (100 x log)

10 20 30

0.05

0.05

EBP (%)

10 20 30

0.1

0.1 0.2

1y gov bond yield (%)

10 20 30

0.1

0.1 0.2 10 20 30

2
1

1

S&P500 (100 x log)

10 20 30

2
1

1 10 20 30

0.2

0.2

Real GDP (100 x log)

10 20 30

0.2

0.2 10 20 30

0.1

0.1

GDP deflator (100 x log)

10 20 30

0.1

0.1 10 20 30

0.05

0.05

EBP (%)

10 20 30

0.05

0.05

months months months

SLIDE 21

United States: shocks over time

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

35
30
25
20
15
10
5

5 10 15

Monetary Policy (sign restrictions) Central Bank Information (sign restrictions)

Every month different mix of policy and information shock
Both shocks occur throughout the sample.

19/24

SLIDE 22

Euro area results We have created a dataset of ECB announcement surprises. 284 ECB policy announce- ments from 1999 to 2016. press release + press confer- ence. VAR results similar to the US

Surprise in the 3m Eonia swaps

0.3
0.2
0.1

0.1 0.2 0.3

Surprise in the Euro Stoxx 50

3
2
1

1 2 I: 61 II: 57 III: 68 IV: 85

20/24

SLIDE 23

Euro area: IRFs

(More information shocks in the mix) Interest rate surprise

10 20 30

0.1

0.1

1y Bund yield (%)

10 20 30

2

2

STOXX50 (100 x log)

10 20 30

0.2

0.2

Real GDP (100 x log)

10 20 30

0.1

0.1

GDP deflator (100 x log)

10 20 30

0.5

0.5

BBB bond spread (%)

months months months

SLIDE 24

Euro area: IRFs

(More information shocks in the mix) Interest rate surprise Monetary Policy CB information (poor man’s s.r.) (poor man’s s.r.)

10 20 30

0.1

0.1

1y Bund yield (%)

10 20 30

2

2

STOXX50 (100 x log)

10 20 30

0.2

0.2

Real GDP (100 x log)

10 20 30

0.1

0.1

GDP deflator (100 x log)

10 20 30

0.5

0.5

BBB bond spread (%)

10 20 30

0.1

0.1

1y Bund yield (%)

10 20 30

0.1

0.1 10 20 30

2

2

STOXX50 (100 x log)

10 20 30

2

2 10 20 30

0.2

0.2

Real GDP (100 x log)

10 20 30

0.2

0.2 10 20 30

0.1

0.1

GDP deflator (100 x log)

10 20 30

0.1

0.1 10 20 30

0.5

0.5

BBB bond spread (%)

10 20 30

0.5

0.5

months months months

SLIDE 25

A structural interpretation

... through the lenses of a standard DSGE model with financial fric-

tions Gertler-Karadi, 2011, 2013). We match impulse responses.

To explain real effects of monetary policy shocks

– Nominal frictions less important – Financial frictions more important in our identification vs. the standard HFI identification.

22/24

SLIDE 26

Why does the economy respond to the CB information shock? Alternative stories behind the IRFs

Central banks have superior information on fundamentals (Romer and

Romer 2000). This information would have become apparent anyway, central bank merely reveals it a bit earlier. Announcements predict the trajectory of the economy. → Nakamura and Steinsson (2018)

Self-fulfilling announcements (unexplored). Confidence; strategic com-
plementarities. Public signal, even imprecise, affects the equilibrium.

Announcements cause the trajectory of the economy.

23/24

SLIDE 27

Conclusions

We partition interest rate surprises into two components:

– monetary policy shock - an interest rate increase followed by a contraction, – central bank information shock - an interest rate increase followed by an expansion.

Lessons

– Stronger effects of monetary policy on the economy (we purge the attenuation bias from the information effects) – Central bank information is relevant. We don’t know if it causes

r merely predicts the trajectory of the economy.

24/24

SLIDE 28

Additional slides

25/24

SLIDE 29

Refining the information shocks We add daily change in inflation compensation to high-frequency variables

shock variable Monetary Central Bank Central Bank Policy Demand Supply all mt (high frequency) interest rate + + + stock index – + – inflation compensation – + + yt (low frequency) . . .

26/24

SLIDE 30

Refining the information shocks: results

Monetary Policy CB demand CB supply

20

0.1

0.1 0.2

1y govt. bond yield (%)

20

0.1

0.1 0.2 20

0.1

0.1 0.2 20

2
1

1

S&P500 (100 x log)

20

2
1

1 20

2
1

1 20

0.2

0.2

Real GDP (100 x log)

20

0.2

0.2 20

0.2

0.2 20

0.1

0.1

GDP deflator (100 x log)

20

0.1

0.1 20

0.1

0.1 20

0.05

0.05

EBP (%)

20

0.05

0.05 20

0.05

0.05

SLIDE 31

Relaxing the zero restrictions - almost the same with zero restrictions on B without zero restrictions on B

10 20 30 0.05

surprise in 3m ff futures

10 20 30 0.05 10 20 30

0.4
0.2

0.2 0.4

surprise in S&P500

10 20 30

0.4
0.2

0.2 0.4 10 20 30

0.05

0.05 0.1

1y govt. bond yield (%)

10 20 30

0.05

0.05 0.1 10 20 30

2
1

1

S&P500 (100 x log)

10 20 30

2
1

1 10 20 30

0.2
0.1

0.1

Real GDP (100 x log)

10 20 30

0.2
0.1

0.1 10 20 30

0.05

0.05

GDP deflator (100 x log)

10 20 30

0.05

0.05 10 20 30

0.05

0.05

EBP (%)

10 20 30

0.05

0.05 10 20 30 0.02 0.04

surprise in 3m ff futures

10 20 30 0.02 0.04 10 20 30

0.4
0.2

0.2 0.4

surprise in S&P500

10 20 30

0.4
0.2

0.2 0.4 10 20 30

0.05

0.05 0.1

1y govt. bond yield (%)

10 20 30

0.05

0.05 0.1 10 20 30

1

1

S&P500 (100 x log)

10 20 30

1

1 10 20 30

0.2
0.1

0.1

Real GDP (100 x log)

10 20 30

0.2
0.1

0.1 10 20 30

0.08
0.06
0.04
0.02

0.02 0.04

GDP deflator (100 x log)

10 20 30

0.08
0.06
0.04
0.02

0.02 0.04 10 20 30

0.05

0.05

EBP (%)

10 20 30

0.05

0.05

SLIDE 32

US VAR: Robust error bands (Giacomini-Kitagawa, 2015)

Monetary policy CB information (negative co-movement) (positive co-movement)

10 20 30

0.1

0.1 0.2

1y govt. bond yield (%)

10 20 30

0.1

0.1 0.2 10 20 30

2

2

S&P500 (100 x log)

10 20 30

2

2 10 20 30

0.2

0.2

Real GDP (100 x log)

10 20 30

0.2

0.2 10 20 30

0.1

0.1

GDP deflator (100 x log)

10 20 30

0.1

0.1 10 20 30

0.05

0.05

EBP (%)

10 20 30

0.05

0.05

months months

SLIDE 33

US VAR: 2-year rate as instrument

A. Standard HFI
B. Sign restrictions

2-year interest rate surprise Monetary policy CB information (negative co-movement) (positive co-movement)

10 20 30

0.05

0.05 0.1

1y govt. bond yield (%)

10 20 30

1
0.5

0.5

S&P500 (100 x log)

10 20 30

0.2
0.1

Real GDP (100 x log)

10 20 30

0.1
0.05

GDP deflator (100 x log)

10 20 30 0.02 0.04

EBP (%)

10 20 30

0.1

0.1

1y govt. bond yield (%)

10 20 30

0.1

0.1 10 20 30

2
1

1

S&P500 (100 x log)

10 20 30

2
1

1 10 20 30

0.2
0.1

0.1

Real GDP (100 x log)

10 20 30

0.2
0.1

0.1 10 20 30

0.1
0.05

0.05

GDP deflator (100 x log)

10 20 30

0.1
0.05

0.05 10 20 30

0.05

0.05

EBP (%)

10 20 30

0.05

0.05

months months months

SLIDE 34

US VAR: sample until December 2006

A. Standard HFI
B. Sign restrictions

interest rate surprise Monetary policy CB information (negative co-movement) (positive co-movement)

10 20 30

0.1

0.1 0.2

1y gov bond yield (%)

10 20 30

2
1

S&P500 (100 x log)

10 20 30

0.2
0.1

0.1

Real GDP (100 x log)

10 20 30

0.1
0.05

GDP deflator (100 x log)

10 20 30

0.02

0.02 0.04 0.06

EBP (%)

10 20 30

0.1

0.1 0.2

1y govt. bond yield (%)

10 20 30

0.1

0.1 0.2 10 20 30

2
1

S&P500 (100 x log)

10 20 30

2
1

10 20 30

0.2
0.1

0.1

Real GDP (100 x log)

10 20 30

0.2
0.1

0.1 10 20 30

0.1
0.05

0.05

GDP deflator (100 x log)

10 20 30

0.1
0.05

0.05 10 20 30

0.05

0.05

EBP (%)

10 20 30

0.05

0.05

months months months

SLIDE 35

US: the GSS decomposition is different from ours

A. Standard HFI
C. GSS(2005)

Interest rate surprise

Target factor Path factor

10 20 30 0.1 0.2

1y govt. bond yield (%)

10 20 30

1.5
1
0.5

S&P500 (100 x log)

10 20 30

0.2

0.2

Real GDP (100 x log)

10 20 30

0.1

0.1

GDP deflator (100 x log)

10 20 30 0.02 0.04

EBP (%)

10 20 30 0.1 0.2

1y govt. bond yield (%)

10 20 30 0.1 0.2 10 20 30

1.5
1
0.5

0.5

S&P500 (100 x log)

10 20 30

1.5
1
0.5

0.5 10 20 30

0.2
0.1

0.1

Real GDP (100 x log)

10 20 30

0.2
0.1

0.1 10 20 30

0.1
0.05

GDP deflator (100 x log)

10 20 30

0.1
0.05

10 20 30

0.02

0.02 0.04

EBP (%)

10 20 30

0.02

0.02 0.04

months months months

SLIDE 36

US VAR with IP and CPI standard HFI sign restrictions

10 20 30

0.1

0.1 0.2

1y govt. bond yield (%)

10 20 30

2
1

1

S&P500 (100 x log)

10 20 30

0.6
0.4
0.2
Ind. Prod.

(100 x log)

10 20 30

0.1

0.1

CPI (100 x log)

10 20 30

0.05

0.05

EBP (%)

10 20 30 0.1 0.2

1y govt. bond yield (%)

10 20 30 0.1 0.2 10 20 30

2
1

1

S&P500 (100 x log)

10 20 30

2
1

1 10 20 30

0.6
0.4
0.2
Ind. Prod.

(100 x log)

10 20 30

0.6
0.4
0.2

10 20 30

0.1

0.1

CPI (100 x log)

10 20 30

0.1

0.1 10 20 30

0.05

0.05

EBP (%)

10 20 30

0.05

0.05

SLIDE 37

US VAR: Responses of expectations

A. Standard HFI
B. Sign restrictions

Monetary Policy

Monetary Policy CB information (negative correlation) (positive correlation)

10 20 30

Expected GDP growth (%)

0.1

0.1 10 20 30

Expected GDP growth (%)

0.1

0.1 10 20 30

0.1

0.1 10 20 30

Expected inflation (%)

0.1

0.1 10 20 30

Expected inflation (%)

0.1

0.1 10 20 30

0.1

0.1 10 20 30

5y break-even inflation rate (%)

0.1
0.05

0.05 10 20 30

5y break-even inflation rate (%)

0.1
0.05

0.05 10 20 30

0.1
0.05

0.05

months months months

Positive impact of GDP-growth expectations in standard HFI

is due to information channel (as in Nakamura, Steinsson, 2018)

Delayed adjustment (Coibion, Gorodnichenko, 2012)

SLIDE 38

Euro area VAR: IRFs

(More information shocks in the mix) Interest rate surprise Monetary Policy CB information (sign restr.) (sign restr.)

10 20 30

0.1

0.1

1y Bund yield (%)

10 20 30

2

2

STOXX50 (100 x log)

10 20 30

0.2

0.2

Real GDP (100 x log)

10 20 30

0.1

0.1

GDP deflator (100 x log)

10 20 30

0.5

0.5

BBB bond spread (%)

10 20 30

0.1

0.1

1y Bund yield (%)

10 20 30

0.1

0.1 10 20 30

2

2

STOXX50 (100 x log)

10 20 30

2

2 10 20 30

0.2

0.2

Real GDP (100 x log)

10 20 30

0.2

0.2 10 20 30

0.1

0.1

GDP deflator (100 x log)

10 20 30

0.1

0.1 10 20 30

0.5

0.5

BBB bond spread (%)

10 20 30

0.5

0.5

months months months

SLIDE 39

Euro area VAR with IP and HICP

Interest rate surprise Monetary Policy CB information (sign restr.) (sign restr.)

10 20 30

0.05

0.05

1y govt. bond yield DE (%)

10 20 30

1

1 2

STOXX50 (100 x log)

10 20 30

0.2

0.2

IP (100 x log)

10 20 30

0.15
0.1
0.05

HICP

10 20 30

0.4
0.2

0.2

BBB bond spread (%)

10 20 30

0.05

0.05 0.1

1y govt. bond yield DE (%)

10 20 30

0.05

0.05 0.1 10 20 30

1

1 2

STOXX50 (100 x log)

10 20 30

1

1 2 10 20 30

0.2

0.2 0.4

IP (100 x log)

10 20 30

0.2

0.2 0.4 10 20 30

0.1

0.1

HICP

10 20 30

0.1

0.1 10 20 30

0.4
0.2

0.2

BBB bond spread (%)

10 20 30

0.4
0.2

0.2

months months months

SLIDE 40

Euro area: shocks over time

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

30
25
20
15
10
5

5 10 15

Monetary Policy (sign restrictions) Central Bank Information (sign restrictions)

August 2011: no IR change; “particularly high” uncertainty
July 2012: IR cut, because “downside risks have materialized”
October 1999: Millenium bug

37/24

SLIDE 41

Construction of high-frequency surprises for the 3-month Eonia swap

4.5 4.55 4.6 4.65 4.7 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 Bid Price Ask Price midquote

EUREON3M 2001-05-10

.8 .85 .9 .95 1 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 Bid Price Ask Price midquote

EUREON3M 2011-03-03

38/24