Global financial transmission of monetary policy shocks

Michael Ehrmann (ECB) and Marcel Fratzscher (ECB)

Workshop “The Architecture of Financial System

Stability”Capri, 24-25 May, 2006

The usual disclaimers apply.

Introduction

• Understanding global financial integration

– How integrated are global financial markets

– Linking asset price transmission with financial integration / quantities:

– Is the shock transmission to asset prices related to international portfolio holdings?

– Implications for risk-sharing and diversification

• Problem of identification

– Asset prices determined simultaneously, both domestically and internationally

– Origin of shock - direction of causality ?

Introduction

• Strategy of paper

– Identification of US monetary policy shocks: change in Fed funds futures in 30-minute window around FOMC decisions

• Important driver of global markets in recent years

• …in particular for equity markets

– Transmission to 50 foreign equity markets

– Sector equity returns across 10 sectors

– Strength, channels and determinants of transmission

Overview of results

1. Strength of transmission

– Foreign equity markets drop, on average, by 3.8% in response to 100 bp tightening in US monetary policy (US market: 6.5%)

– High degree of cross-country heterogeneity:

• No effect: China, India or Mexico

• ~10% or more: Indonesia, Korea and Turkey

• Strongest among advanced economies: Australia, Canada, Finland and Sweden

– Substantial sector heterogeneity:

• effects range from 1.6% for utilities to 7.4% for information technology

Overview of results

2. Channels of transmission

– Via US asset prices:

• reaction of US short-term interest rates important: transmission 2½ times larger when high US short-term interest rate response

– Via foreign asset prices:

• reaction of foreign short-term interest rates and exchange rates key: transmission up to 2 times larger when high response of these asset prices

Overview of results

3. Determinants of heterogeneity in transmission

– Openness:

• Countries with closed markets (equity market, domestic financial sector) do not react to US monetary policy shocks

– Exchange rate policy:

• De jure regime does not matter, but de facto exchange rate volatility does – fear of floating?

– Others:

• Business cycle correlation with US matters …

• … indebtedness and information asymmetries do not

Overview of results

3. Determinants of heterogeneity in transmission

– Real integration:

• Transmission ~2 times stronger if relatively high trade with rest of world (ROW)

– Financial integration:

• Transmission 2 – 3 times stronger if relatively high financial integration with ROW

– Key finding: it is the degree of global integration, i.e. with ROW, but not with the United States, that matters for the financial transmission process

Literature & hypotheses

• Paper is on intersection between and links three strands of the literature:

– Literature on domestic financial market effects of monetary policy

– Literature on international asset market linkages

– Literature explaining the linkages

Literature

• Domestic transmission of monetary policy shocks to equity markets

– Identification problem addressed in various ways

– Structural VAR – low frequency (Thorbecke 1997, Patelis 1997): US monetary policy affects US equity markets

– Identification through heteroskedasticity of asset price movements (Rigobon & Sack 2003, 2004)

– Monetary policy shocks derived from Fed funds futures (Kuttner 2001, Gürkaynak et al. 2005) or polls (Ehrmann & Fratzscher 2004)

– Empirical findings (Bernanke & Kuttner 2005, Ehrmann & Fratzscher 2004, Rigobon & Sack 2004): 5.3–6.2%

Literature

• International transmission – focus mainly on individual asset classes

– Equity markets (Hamao, Masulis & Ng 1990; Lin, Engle & Ito 1994; King, Sentana & Wadhwani 1994)

– FX markets (Andersen et al. 2003)

– Spillover of macroeconomic news (Andersen et al 2005; Becker, Finnerty & Friedman 1995; Faust et al. 2003)

– Cross-market spillovers between US and euro area (Ehrmann, Fratzscher & Rigobon 2005)

• Explaning linkages

– Trade (Eichengreen & Rose 1999, Forbes & Chinn 2004), industry composition (Griffin & Stulz 2001), capital controls (Miniane & Rogers 2003)

Hypotheses

• Macroeconomic repercussions of a US monetary policy shock (economy level):

– transmission strongest to those foreign markets with economies that have a high business cycle correlation, close trade links or high integration

• Microeconomic repercussions (firm level, sector level):

– Transmission via effect on financing costs of foreign firms

– Role of direct firm-level linkages with the US (e.g. FDI or portfolio investment)

– Importance of portfolio rebalancing for financial investors

Data: US monetary policy shock– Change in Fed funds future rates in 30-minute

window around FOMC decisions – usually at 14.15 EST (Gürkaynak, Sack & Swanson 2005)

– Period February 1994 – 2005: 93 FOMC meetings (excluding 11 Sept. 2001)

– Daily frequency: hence monetary policy shocks zero for most days

US monetary policy shock-4

0-2

00

20

01jan1994 01jan1996 01jan1998 01jan2000 01jan2002 01jan2004

Financial market data– Equity returns: daily log changes of Datastream

price indices, in national currencies, of 10 sector returns across 50 countries

– Construction of “unweighted” return index based on unweighted average of sector returns in order to control for sector heterogeneity

– Short-term interest rates: mostly 3-month money market rates

– Bond yields for US: 10-year T-bills

– Exchange rates: daily spot rates

– All based on closing quotes of local markets

Integration and macroeconomic determinants– Openness:

• capital account – 0-1 dummy; IMF AREAER

• equity market and domestic financial sector – 0-1 dummies; Kaminsky & Schmukler (2003), Bussiere & Fratzscher (2004)

• Equity market depth: stock market capitalisation / GDP

– Exchange rate policy:

• de jure regime: IMF AREAER

• de facto regime: Reinhart & Rogoff (2004)

• de facto exchange rate volatility: 12-month standard deviation of changes in daily (nominal) exchange rate vis-à-vis US dollar

Integration and macroeconomic determinants– Others:

• Business cycle correlation with US: correlation in annual GDP growth 1980-2003

• indebtedness of country

• information asymmetries: proxied by geographic distance (gravity literature)

• additional variables: domestic inflation, investment etc.

Integration and macroeconomic determinants– Real integration:

• Trade = exports + imports to GDP with rest of world (ROW) and with United States: IMF DOTS

– Financial integration: Unique data set of bilateral capital stocks

• FDI: UNCTAD

• Portfolio equity & portfolio debt: IMF CPIS

• Loans: BIS ILB

– Important advantage: all integration variables are available multilaterally vis-à-vis ROW as well as bilaterally vis-à-vis United States

– Most variables vary across countries & over time !

Benchmark model

ri,t equity return in country i

St US monetary policy shock

rus “cleaned” US equity market return from:

)1(, itn

tinrttiit ZSr US

)2(,,USr

tn

tUSntUS

tUS ZSr

transmission parameter - causality

general linkage parameter - no causality

definition ensures is orthogonal to

Benchmark model

• as robustness check: if all shocks always emanate from US then US

• Robust OLS estimator with panel-corrected standard errors (PCSE): corrects for heteroskedasticity and cross-correlation of observations

Benchmark model

Benchmark effects std error std error

(1) United States, weighted index -0.079 *** 0.017 -- -- --

(2) United States, unweighted index -0.065 *** 0.013 -- -- --

(3) International, weighted indices -0.045 *** 0.007 0.348 *** 0.010

(4) International, unweighted indices -0.038 *** 0.006 0.304 *** 0.009

Parameter estimates

100 bp tightening in US monetary policy reduces foreign equity returns on average by 3.8% (unweighted index)

General linkage ~ 0.30 : 30% of an equity market shock is transmitted internationally

Cross-country / cross-sector heterogeneity

rikt equity return in country i and sector k

i country-specific transmission parameter

k sector-specific transmission parameter

)4(

,

iktk

krtkktk

ii

rtiiti

ntinkiikt

DDS

DDSZr

US

US

Country-sector effects of US mon. policy

shocks 0

.2.4

.6.8

11

.2g

ener

al e

qui

ty m

arke

t lin

kage

-.15 -.1 -.05 0 .05monetary policy shocks

Cross-country heterogeneity

Country: i

i Country:

i

i Country: i

i

Argentina -0.039 *** 0.467 *** Hong Kong -0.075 *** 0.482 *** Peru -0.001 0.106 ***Australia -0.060 *** 0.375 *** Hungary -0.042 *** 0.346 *** Philippines -0.037 ** 0.283 ***Austria -0.047 *** 0.175 *** India -0.011 0.131 *** Poland -0.030 *** 0.467 ***Belgium -0.035 *** 0.270 *** Indonesia -0.159 *** 0.421 *** Portugal -0.030 *** 0.131 ***Brazil -0.041 *** 0.502 *** Ireland -0.036 *** 0.280 *** Romania -0.002 0.061 ***Canada -0.059 *** 0.564 *** Israel -0.036 *** 0.314 *** Russia -0.043 ** 0.340 ***Chile -0.012 ** 0.244 *** Italy -0.050 *** 0.221 *** Singapore -0.051 *** 0.351 ***China 0.014 0.030 *** Japan -0.011 ** 0.291 *** South Africa -0.044 *** 0.410 ***Colombia 0.015 0.026 Korea -0.101 *** 0.493 *** Spain -0.059 *** 0.213 ***Cyprus -0.048 *** 0.133 *** Luxemburg -0.053 *** 0.191 ** Sri Lanka -0.001 0.033 ***Czech Republic -0.043 *** 0.079 ** Malaysia -0.029 *** 0.344 *** Sweden -0.086 *** 0.406 ***Denmark -0.042 *** 0.360 *** Mexico -0.011 0.574 *** Switzerland -0.039 *** 0.268 ***Finland -0.073 *** 0.358 *** Netherlands -0.050 *** 0.385 *** Taiwan -0.036 *** 0.302 ***France -0.040 *** 0.318 *** New Zealand -0.041 *** 0.293 *** Thailand -0.059 *** 0.340 ***Germany -0.037 *** 0.279 *** Norway -0.057 *** 0.340 *** Turkey -0.117 *** 0.355 ***Greece -0.046 *** 0.323 *** Pakistan 0.031 0.034 United Kingdom -0.041 *** 0.280 ***

Channels of transmission

Channel 1: Via US asset prices:

– US equity markets: e.g. through cross-listing of firms, or through FDI or portfolio investment holdings of firms in the US

– US short-term interest rates: via financing costs of firms raising funds in US financial markets

– US long-term interest rates: uncertain effect on equity markets – may reflect credibility of central bank or growth expectations

)5(,

4

1

uit

ntin

rt

l

ltl

lttli

uit ZDDSr US

Transmission channel 1: US asset prices

reaction of US short-term interest rates important : transmission 2½ times larger when high US short-term interest rate response

Individual US asset price change: std error coef p-value

US equity market - low US equity market reaction -0.032 *** 0.012 -0.005 0.701 - high US equity market reaction -0.037 *** 0.008 US interest rates - low US interest rate reaction -0.032 *** 0.007 -0.047 *** 0.009 - high US interest rate reaction -0.079 *** 0.016 US bond yields - low US bond yield reaction -0.039 *** 0.007 -0.001 0.924 - high US bond yield reaction -0.040 *** 0.012

Parameter estimates Test for difference

Via foreign interest rates and exchange rates:

– fixer: exchange rate reacts little, interest rate reacts much

– floater: exchange rate reacts much, interest rate reacts little

– dependent country: exchange rate reacts much, interest rate reacts much

– independent country: exchange rate reacts little, interest rate reacts little

)6(,

4

1,,

uit

ntin

rt

m

mtim

mtitmi

uit ZDDSr US

Transmission channel 2: Foreign asset prices

Transmission channel 2: Foreign asset prices

Transmission up to 2 times larger for “dependent countries”, i.e. countries where interest rates and exchange rates react strongly

Countries classified as: std error (1) (2) (3)

fixer - low exchange rate & high interest rate reaction -0.036 *** 0.005 floater - high exchange rate & low interest rate reaction -0.029 *** 0.006 0.48 dependent country - high exchange rate & high interest rate reaction -0.052 *** 0.005 0.02 0.00 independent country - low exchange rate & low interest rate reaction -0.023 *** 0.007 0.48 0.12 0.00

Parameter estimates Test for difference vs.

Determinants of financial transmission

uit

ntin

highit

midit

lowit

highit

rtt

midit

rtt

lowit

rtti

uit

Z

XXX

XS

XS

XSr

US

US

US

,

321

33

22

11

Determinants of financial transmission

Openness: Countries with closed markets (equity market, domestic financial sector) do not react to US monetary policy shocks; exception: capital account

X std error (2) (3)

A. OPENNESS

Capital account (1) closed -0.031 *** 0.008 0.200

(2) open -0.041 *** 0.007

Equity market (1) closed -0.001 0.012 0.001

(2) open -0.041 *** 0.007

Domestic financial (1) closed -0.010 0.012 0.012

sector (2) open -0.040 *** 0.006

difference

Determinants of financial transmission

Exchange rate policy: De jure regime does not matter, but de facto exchange rate volatility does – fear of floating?

X std error (2) (3)

B. EXCHANGE RATE

Volatility of effective (1) low -0.027 *** 0.007 0.284 0.001

exchange rate (2) medium -0.036 *** 0.008 0.020

(3) high -0.056 *** 0.008

Volatility vis-à-vis (1) low -0.027 *** 0.007 0.261 0.001

US dollar (2) medium -0.036 *** 0.008 0.019

(3) high -0.055 *** 0.008

Regime - de jure (1) fix -0.031 *** 0.007 0.333

(2) float -0.038 *** 0.007

difference

Determinants of financial transmission

– Others: Business cycle correlation with US matters

– Information asymmetries do not

X std error (2) (3)

C. OTHER

GDP correlation (1) low -0.026 *** 0.008 0.193 0.012

with US (2) medium -0.036 *** 0.007 0.098

(3) high -0.048 *** 0.008

Geographic distance (1) low -0.037 *** 0.007 0.389 0.679

(2) medium -0.032 *** 0.008 0.307

(3) high -0.040 *** 0.007

difference

Real and financial integration with ROW

X std error (2) (3)

A. WITH THE WORLD - ASSETS & LIABILITIES, IN- & OUTFLOWS

Total trade flows (1) low -0.024 *** 0.008 0.084 0.018

(2) medium -0.037 *** 0.007 0.191

(3) high -0.047 *** 0.008

Total capital stocks (1) low -0.018 ** 0.008 0.000 0.010

(2) medium -0.048 *** 0.008 0.518

(3) high -0.043 *** 0.008

FDI (1) low -0.019 ** 0.008 0.001 0.003

(2) medium -0.045 *** 0.008 0.976

(3) high -0.045 *** 0.007

Portfolio equity (1) low -0.024 *** 0.007 0.002 0.015

(2) medium -0.047 *** 0.007 0.490

(3) high -0.043 *** 0.007

Portfolio debt (1) low -0.019 *** 0.007 0.000 0.013

(2) medium -0.051 *** 0.008 0.242

(3) high -0.042 *** 0.008

Other investment (1) low -0.015 0.009 0.004 0.007

loans (2) medium -0.048 *** 0.008 0.841

(3) high -0.046 *** 0.008

difference

parameter estimates

Real and financial integration with US

X std error (2) (3)

B. WITH THE UNITED STATES - ASSETS & LIABILITIES, IN- & OUTFLOWS

Total trade flows (1) low -0.031 *** 0.007 0.008 0.611

(2) medium -0.050 *** 0.008 0.015

(3) high -0.035 *** 0.007

Total capital stocks (1) low -0.047 *** 0.009 0.588 0.130

(2) medium -0.051 *** 0.009 0.018

(3) high -0.035 *** 0.006

FDI (1) low -0.043 *** 0.008 0.652 0.281

(2) medium -0.039 *** 0.008 0.467

(3) high -0.035 *** 0.006

Portfolio equity (1) low -0.035 *** 0.008 0.319 0.745

(2) medium -0.042 *** 0.008 0.107

(3) high -0.033 *** 0.006

Portfolio debt (1) low -0.042 *** 0.009 0.204 0.312

(2) medium -0.049 *** 0.009 0.060

(3) high -0.034 *** 0.006

Other investment (1) low -0.025 *** 0.008 0.002 0.034

loans (2) medium -0.049 *** 0.008 0.243

(3) high -0.040 *** 0.007

difference

Real and financial integration

– Real integration:

• Transmission ~2 times stronger if relatively high trade with rest of world (ROW)

– Financial integration:

• Transmission 2 – 3 times stronger if relatively high financial integration with ROW

– Key finding: it is the degree of global integration, i.e. with ROW, but not with the United States, that matters for the financial transmission process

Robustness

• Various extensions underline robustness of results:

– Caveat: significant correlation across determinants

– Split between asset and liabilities, inflows and outflows

– Inclusion of macroeconomic news (GDP, IP, CPI, PPI, consumer and producer confidence, employment etc.)

– Example of shocks to US industrial production…

Robustness: macroeconomic shocks

Parameter estimatesType of US shock: std error

(1) Monetary policy -0.039 *** 0.004(2) Non-farm payroll employment -0.193 *** 0.032(3) NAPM / ISM confidence indicator 0.061 0.043(4) CPI inflation -0.335 *** 0.059(5) PPI inflation -0.084 *** 0.027(6) GDP (advance release) 0.370 *** 0.058(7) Industrial production 0.064 * 0.038(8) Retail sales -0.026 0.044(9) Consumer confidence 0.130 0.092

(10) Trade balance 0.413 *** 0.090(11) Housing starts -0.071 0.051(12) Hours worked per week -0.043 0.063

Real and financial integration with ROWindustrial production shock

X std error (2) (3)

A. WITH THE WORLD - ASSETS & LIABILITIES, IN- & OUTFLOWS

Total trade flows (1) low -0.026 0.095 0.595 0.069

(2) medium 0.043 0.090 0.196

(3) high 0.198 ** 0.079

Total capital stocks (1) low -0.033 0.096 0.384 0.120

(2) medium 0.081 0.090 0.511

(3) high 0.160 ** 0.079

FDI (1) low -0.030 0.095 0.353 0.108

(2) medium 0.091 0.090 0.516

(3) high 0.168 ** 0.079

Portfolio equity (1) low 0.044 0.092 0.879 0.404

(2) medium 0.024 0.088 0.313

(3) high 0.150 * 0.088

Portfolio debt (1) low 0.094 0.092 0.467 0.882

(2) medium 0.002 0.088 0.371

(3) high 0.113 0.088

Other investment (1) low -0.094 0.100 0.085 0.052

loans (2) medium 0.134 0.087 0.864

(3) high 0.154 ** 0.079

difference

Summary & conclusions

• Linking asset price transmission with financial integration / quantities

• Identification problem addressed by focusing on transmission of US monetary policy shocks - important driver of global financial markets

• Strength of transmission

– Foreign equity markets drop, on average, by 3.8% in response to 100 bp tightening in US monetary policy

– High degree of heterogeneity in cross-country (0->10%) and cross-sector (1.6%-7.4%) transmission

Summary & conclusions

• Channels of transmission

– Via US asset prices: mainly via US s-t interest rates

– Via foreign asset prices: both s-t interest rates and exchange rates

• Determinants of heterogeneity in transmission

– Openness and exchange rate policy important

• De jure regime does not matter, but de facto exchange rate volatility does – fear of floating?

– Others:

• Business cycle correlation with US matters

Summary & conclusions

• Determinants of heterogeneity in transmission

– Real integration and in particular financial integration matter: transmission 2 – 3 times stronger if relatively high financial integration with rest of world

– Key finding: it is the degree of global integration, i.e. with ROW, but not with the United States, that matters for the financial transmission process

• International financial transmission of monetary policy similar to that of other types of shocks

• Implications for risk sharing / diversification

Annex

Country-specific effects of US mon. policy

shocks 0

.2.4

.6.8

1g

ener

al e

qui

ty m

arke

t lin

kage

-.15 -.1 -.05 0 .05monetary policy shocks

Sector-specific effects of US mon. policy

shocks .2

.3.4

.5g

ener

al e

qui

ty m

arke

t lin

kage

-.07 -.06 -.05 -.04 -.03 -.02monetary policy shocks

Sector-specific effects of US mon. policy

shocks 0

.2.4

.6.8

11

.2g

ener

al e

qui

ty m

arke

t lin

kage

-.5 0 .5 1 1.5monetary policy shocks

Cross-sector heterogeneity

Sector: k

k

Financial -0.044 *** 0.318 ***Information technology -0.074 *** 0.518 ***Utilities -0.016 *** 0.191 ***Non-cyclical services -0.051 *** 0.416 ***Cyclical services -0.040 *** 0.303 ***Cyclical consumer goods -0.040 *** 0.265 ***Industrial -0.045 *** 0.319 ***Basic industries -0.034 *** 0.276 ***Resources -0.047 *** 0.245 ***Non-cyclical consumer goods -0.026 *** 0.232 ***

Parameter estimates

Time asymmetries: US policy shocks

No evidence for (time) asymmetric effects stemming from the US monetary policy shock itself

Difference

Monetary policy shock 1 std error

2 std error p-value

Unweighted indices

Monetary policy: no change vs. change -0.033 ** 0.016 -0.039 *** 0.007 0.713Directional change: no vs. yes -0.039 *** 0.007 -0.037 *** 0.013 0.902Monetary policy shock: negative vs. positive -0.039 *** 0.007 -0.034 ** 0.015 0.491Monetary policy shock: small vs. large -0.039 *** 0.007 -0.036 ** 0.015 0.882Volatility: small vs. large -0.039 *** 0.007 -0.039 *** 0.012 0.998

Parameter estimates

Transmission channel 1: US asset prices

Result confirmed when using finer categorisation for equity market and short-term interest rate reactions

Combined US asset price change: b std error (1) (2) (3)

US equity market channel - low US interest rate & high US equity reaction -0.030 *** 0.004 US interest rate channel - high US interest rate & low US equity reaction -0.073 *** 0.011 0.00 US interest rate & equity channels - high US interest rate & high US equity reaction -0.077 *** 0.012 0.00 0.81 No US asset price channel - low US interest rate & low US equity reaction -0.030 *** 0.007 0.96 0.00 0.00

Parameter estimates Test for difference vs.

Transmission channel 2: Foreign asset prices

Higher reaction of foreign equity markets to US monetary policy shocks if large change in short-term interest rate or in exchange rate

Foreign asset price change: std error coef p-value

exchange rate - low exchange rate reaction -0.032 *** 0.004 -0.009 * 0.075 - high exchange rate reaction -0.041 *** 0.004 interest rate - low interest rate reaction -0.028 *** 0.005 -0.010 * 0.076 - high interest rate reaction -0.038 *** 0.003

Parameter estimates Test for difference