Financial Crisis, US Unconventional Monetary Policy and International Spillovers; by Qianying Chen, Andrew Filardo, Dong He, and Feng Zhu; IMF Working Paper No. 15/85; April 1, 2015

WP/15/85

Financial Crisis, US Unconventional Monetary Policy and

International Spillovers

by Qianying Chen, Andrew Filardo, Dong He, and Feng Zhu

IMF Working Paper

European Department

Financial Crisis, US Unconventional Monetary Policy and International Spillovers

Prepared by Qianying Chen, Andrew Filardo, Dong He, and Feng Zhu

Authorized for distribution by Bas Bakker

April 2015

Abstract

We study the impact of the US quantitative easing (QE) on both the emerging and advanced

economies, estimating a global vector error-correction model (GVECM) and conducting

counterfactual analyses. We focus on the effects of reductions in the US term and corporate

spreads. First, US QE measures reducing the US corporate spread appear to be more

important than lowering the US term spread. Second, US QE measures might have prevented

episodes of prolonged recession and deflation in the advanced economies. Third, the

estimated effects on the emerging economies have been diverse but often larger than those

recorded in the US and other advanced economies. The heterogeneous effects from US QE

measures indicate unevenly distributed benefits and costs.

The views expressed are those of the authors and do not necessarily represent the views of the Bank for International

Settlements or the International Monetary Fund. We very much appreciate the comments by the journal referee, and the

comments on the drafts of this and earlier versions from Jonathan Batten, Menzie Chinn, Dietrich Domanski, Alex Heath,

Roong Mallikamas, Richhild Moessner, Shinobu Nakagawa, Patrizio Pagano, Eswar Prasad, John Taylor, Bernhard Winkler

and the participants at the 9th Annual Hong Kong Institute for Monetary Research Summer Workshop, the 2011 Joint

Workshop on Emerging Markets of the European Central Bank and the Deutsche Bundesbank, the 2012 EEA-ESEM

meeting, the Bank of Canada 2013 conference on “International macroeconomic policy cooperation: challenges and

prospects”, the Bank of Korea Seminar on “Macro-financial linkages and macro-prudential policies” and the ECB-IMF

conference on “International dimensions of conventional and unconventional monetary policy”.

Qianying Chen, International Monetary Fund (IMF), [email protected]; Andrew Filardo, Bank for International Settlements

(BIS), [email protected]; Dong He, IMF, [email protected]; Feng Zhu, BIS, [email protected]. Dong He was a staff

member of the Hong Kong Monetary Authority (HKMA) when he worked on this paper.

IMF Working Papers describe research in progress by the author(s) and are published to

elicit comments and to encourage debate. The views expressed in IMF Working Papers are

those of the author(s) and do not necessarily represent the views of the IMF, its Executive Board,

or IMF management.

JEL Classification Numbers: E44, E52, E65, F42, F47

Keywords: emerging economies; financial crisis; global VAR; international monetary policy

spillovers; quantitative easing; unconventional monetary policy

Author’s E-Mail Address: [email protected] (Q. Chen), [email protected] (A. Filardo),

[email protected] (D. He), [email protected] (F. Zhu)

Content

I. Introduction ______________________________________________________________4

II. Estimating the Effects of US Unconventional Policies ____________________________6

A. GVECM Analysis: Model and Variables ____________________________________6

B. GVECM Analysis: Impulse Responses ______________________________________9

C. Cross-Border Monetary Policy Spillovers ___________________________________11

III. GVECM-based counterfactual analysis ______________________________________16

A. Spillovers From Reductions in the US Corporate Spread _______________________19

IV. Conclusion ____________________________________________________________22

Appendices

I: Structure of the GVECM Model _____________________________________________24

II: Foreign Exchange Pressure Index ___________________________________________26

III: Time-Varying Weights for Foreign Variables _________________________________27

IV: Data __________________________________________________________________28

References ________________________________________________________________29

Figures

1. Impulse Responses to US Term and Corporate Spread Shocks: United States _________11

2. Maximum Impulse Responses to a US Corporate Spread Shock ____________________14

3. Distribution of Maximum Impulse Responses to US Term and Corporate

Spread Shocks __________________________________________________________15

4. Counterfactual Analysis of Domestic US QE Impact: US Corporate Spread __________18

5. Counterfactual Analysis of Euro Area Impact: US Corporate Spread ________________20

6. Counterfactual Analysis of Impact in Brazil: US Corporate Spread _________________21

7. Counterfactual Analysis of Impact in China: US Corporate Spread _________________22

Table

1. The Federal Reserve’s Large-Scale Asset Purchase (LSAP) Programs ________________5

I. INTRODUCTION

1. The 2007–2009 US subprime mortgage crisis and the Great Recession have had a

major impact on the design and implementation of monetary policy. Following the crisis, the

Federal Reserve lowered the federal funds rate target rapidly to near zero, and has taken

additional measures considered “unconventional” (Table 1).

2. The unconventional policy actions taken by central banks in a number of major

economies have led to a burgeoning literature on their effectiveness. Most work has focused

on their domestic effects and relied on event studies analysing the announcement effects of

quantitative easing (QE) on asset prices: some studies have also employed regression

analysis. Among others, D’Amico and King (2010), Doh (2010), Gagnon, Raskin, Remache

and Sack (2010, 2011), Joyce, Lasaosa, Stevens and Tong (2011), Krishnamurthy and

Vissing-Jorgensen (2011) and Meaning and Zhu (2011, 2012) provide estimates for the

Federal Reserve’s and the Bank of England’s large-scale asset purchase programs.

3. A better understanding of the monetary policy spillovers associated with QE

measures may help policymakers to cope with the challenges posed by such policies and to

assess the need for international policy coordination. Yet we know very little about the

impact of the unconventional policies on real activity, and so far there has been little research

on their cross-border spillovers, especially on emerging economies.

4. Several studies examine the cross-border financial market impact of QE policies.

Relying on event studies of US asset purchases, Neely (2010) finds that US QE lowered bond

rates in the other advanced economies by 20-80 basis points and depreciated the US dollar by

4-11 percent. Glick and Leduc (2012) show that commodity prices on average fell upon the

announcements of US asset purchases, despite a decline in long-term interest rates and US

dollar depreciation. Chen, Filardo, He and Zhu (2012, 2014a) and Rogers, Scotti and Wright

(2014) provide evidence on the international spillovers of the unconventional measures

implemented by the Bank of England, the European Central Bank, the Federal Reserve and

the Bank of Japan. Fratzscher, Lo Duca and Straub (2013) find that earlier US QE measures

were highly effective in lowering sovereign yields and raising equity prices. But since 2010

such measures have had a muted impact on yields across countries. Chen, Filardo, He and

Zhu (2014b) introduce estimated shadow federal funds rates in a global VAR to assess the

domestic and global impact of US unconventional monetary policy. They find that US QE

might not have only prevented US recessions but also had substantial global spillovers. IMF

(2013a, b) finds that unconventional monetary policies have successfully restored market

functioning and intermediation in the early phase of the global financial crisis, but their

continuation carries risks.

5. There are two major views on the spillovers of the unconventional monetary policies

implemented in the major advanced economies. The first view considers that such policies

To assess the macroeconomic effects of QE measures, Chen, Filardo, He and Zhu (2012, 2014a, b) estimate a global VAR

model and Gambacorta, Hofmann and Peersman (2012) employ a panel VAR model. Hofmann and Zhu (2013) study the

effects on inflation expectations of Federal Reserve asset purchases and find these were well-anchored and such purchases

had little impact.

are designed for domestic contingencies; any spillovers are unintended and primarily an issue

for other policymakers to address. This echoes the Obstfeld-Rogoff (2002) proposition that

there are only small gains from policy coordination once individual central banks implement

policies optimised to achieve domestic macro stability. Moreover, Ostry and Ghosh (2013)

consider uncertainties and disagreement about the cross-border effects of QE policies a major

obstacle to policy coordination.

Table 1. The Federal Reserve’s Large-Scale Asset Purchase (LSAP) Programs

Announcement

Termination

Assets purchased

Amount

LSAP1

Nov. 2008

Agency mortgage-backed

securities (MBS) and agency

debt

$600 billion

Mar. 2009

Agency securities

$850 billion

Mar. 2010

Longer-term US Treasury

securities

$300 billion

LSAP2

Nov. 2010

Jun. 2011

Longer-term US Treasury

securities

$600 billion

Maturity

extension

program

(MEP)

Sep. 2011

US Treasury securities with

remaining maturities of six to

30 years

$400 billion

Jun. 2012

Dec. 2012

US Treasury securities with

remaining maturities of six to

30 years

LSAP3

Sep. 2012

Oct. 2014

Agency MBS

$40 billion per

month

Dec. 2012

Oct. 2014

Longer-term US Treasury

securities

$45 billion per

month

Initially announced amount of asset purchases for each program or program expansion.

The purchases were open-ended when they were announced. The Federal Reserve started to taper the asset

purchases in January 2014, and eventually halted the purchases altogether in October 2014.

Source: US Federal Reserve.

6. The second view argues that QE policies are less benign. Amongst other things, they

depreciate domestic currencies and inflate risk-adjusted interest rate differentials vis-à-vis

other economies, leading to potentially large capital inflows and consumer and asset price

inflation pressures abroad. Besides concerns with competitive devaluation, Rajan (2013)

highlights the potential danger of “competitive asset price inflation”. Taylor (2013) points

out that, while the Obstfeld-Rogoff (2002) proposition may be true in normal times, sizeable

cross-border spillovers seen in recent years have changed the cost-benefit analysis. This

would particularly be the case if QE measures represent “deviations from rules-based policy”

which create incentives for other central banks to deviate from rules-based policies. The

cross-border effects of QE may also be perceived as beneficial or harmful by those affected,

depending in large part on the cyclical position they find themselves in at the time when QE

is adopted. There is a general consensus that that during the global financial crisis and the

ensuing recession, QE policies helped to stabilize global financial markets and prevented an

even further collapse in the global economic activity. As recovery languished in the advanced

economies but gathered pace in the emerging economies, QE arguably contributed to

economic overheating and asset market excesses in some jurisdictions owing to the large

currency appreciation and capital inflow pressures.

7. In this paper, we study the macroeconomic effects of QE, both domestic and

international, estimating a global vector error correction model (GVECM) covering

17 advanced and emerging economies, using monthly data spanning 2007–2013. Given the

size of the GVECM and the limited data span, the elevated estimation uncertainty is reflected

in the relatively large confidence bands. Our estimates suggest that the cross-border

spillovers varied across economies and over time. We find that reducing the US corporate

spread, and, to a lesser extent, the US term spread, had sizeable effects on financial

conditions and economic activity both domestically and globally. Taken at face value, our

counterfactual analysis indicates that US QE programs, especially LSAP1, were important

counter-cyclical measures, apparently preventing the US and other advanced economies from

prolonged recession and deflation.

8. The effects of US QE measures on the emerging economies are estimated to be

generally larger and more diverse than those in the advanced economies. In our view, the

strength of the effects depends partly on how each economy reacts to the US policy shocks,

and partly on the distinct economic and financial structures, policy frameworks and exchange

rate arrangements. Our estimates also suggest that US QE measures contributed to

overheating in Brazil, China and some other emerging economies in 2010 and 2011, but

supported recovery in these economies in 2009 and 2012. The diverse cross-border QE

effects imply that the costs and benefits of US QE policies have been unevenly distributed

between the advanced and emerging economies and have varied over time.

9. The paper is organized as follows. Section 2 describes the GVECM and provides

empirical results on the cross-border impact of US QE measures with impulse responses to a

US term or corporate spread shock estimated from a GVECM. Section 3 examines the

domestic and spillover effects of US QE measures on financial and real activities, assessed

with a counterfactual analysis based on the impulse response estimates. Section 4 concludes.

II. ESTIMATING THE EFFECTS OF US UNCONVENTIONAL POLICIES

10. To assess the domestic and foreign effects of US unconventional policies on real and

financial activities, we employ a global vector error correction model (GVECM), developed

by Pesaran, Schuermann and Weiner (2004), which is suited for capturing cross-border

macro-financial linkages. We first estimate impulse responses for each economy using the

GVECM. Based on these, we design counterfactual scenarios in which US QE measures are

assumed to be absent, and evaluate their effects by comparing the “no-QE” projections to

actual data.

A. GVECM Analysis: Model and Variables

11. The model is structured as follows.

For economy i, the model VECM









 can be

written as:

See BIS (2012) and De Nicolò, Dell’Ariccia, Laeven and Valencia (2010).



















,1,10

itstisiti

stiistiiiit

t εdxΓzΨzΠccx

(1)

with

 



iid

,0~ε

'*''

),(

itit

z xx

, and

'*''

),,(

tit

itit

z dxx

(2)

where 



is the observed global factor, e.g. the CBOE Volatility Index (VIX). For every

non-US economy i, we have

)(





ititititititit

empspbcmpy



(3)

and

)(

,,,,,,





 titititititiit

empspbcmpy



(4)

12. Each country VECM consists of six domestic endogenous variables: real GDP

growth ( ⊿y), the CPI inflation rate (π), a monetary policy indicator (mp), credit growth (⊿

bc), equity price (sp) and foreign exchange pressure index (emp). The model is then

augmented with a set of foreign variables which include, e.g. foreign real GDP growth and

the VIX.

Except for the VIX, the foreign variables are constructed as the weighted averages

of the corresponding variables in all other economies, and they are assumed to be weakly

exogenous.

13. For the US bloc, we include the same set of domestic variables as in the other

economies, but only the non-US real GDP growth as a foreign variable. Given the

importance of the United States in the global economy, we do not treat the other foreign

variables, especially the financial variables, as weakly exogenous in the US bloc. Therefore

the VIX is treated as endogenous in the US bloc:

)(

,,,,,,,





ttUStUStUStUStUStUStUS

vixempspbcmpy



(5)

and

tUStUS

, 

x

(6)

14. Blinder (2010) suggests that central banks use unconventional tools to reduce

interest rate spreads such as “term premiums and/or risk premiums”, buying long-term

Treasuries or using QE to target “risk or liquidity spreads”. The rationale is that “since

private borrowing, lending, and spending decisions presumably depend on (risky) non-

Treasury rates, reducing their spreads over (riskless) Treasuries reduces the interest rates that

matter for actual transactions even if riskless rates are unchanged.” We therefore describe the

Federal Reserve’s unconventional measures, especially the large-scale purchases of

We provide further details on the structure of the GVECM and on variables and data in Appendices A-D.

The VIX, a key measure of market expectations of near-term volatility conveyed by S&P 500 stock index option prices,

provides a good measure of financial market developments.

sovereign (e.g. Treasuries) and private (e.g. agency MBS) assets, with two monetary policy

“indicators”: the US term spread between the 10-year and three-month Treasury yields; and

the US corporate spread between the BofA Merrill Lynch US corporate AAA bond yield and

the effective federal funds rate.

15. Admittedly, the term and corporate spreads may reflect information beyond that

captured by US monetary policy, given that these spreads are important barometers of US

financial sector health. But even in normal times, the term spread is considered a useful

indicator, as central banks act to shape expectations of a specific interest rate path well into

the future. At the zero lower bound (ZLB), the funds rate loses its information content;

however, the two spreads continue to reflect the immediate objectives (and impact) of US QE

measures, namely, to reduce longer-term Treasury yields, lower borrowing costs for

corporates and households and restore credit flows. Purchasing Treasuries and agency MBS

are expected to reduce long-term Treasury yields directly and corporate bond yields via

portfolio rebalancing.

16. For the other advanced economies, which have faced the ZLB and implemented

unconventional measures, we use the spread between 10-year and three-month government

bond yields for the United Kingdom and Japan as the monetary policy indicator, and the

spread between the 10-year government bond yield and the main refinancing rate for the euro

area. For the emerging economies, we describe monetary policy with the growth rates in a

broad monetary aggregate, as their central banks tend to use a wide range of policy tools and

a broad monetary aggregate may be the more robust indicator for monetary policy.

17. We measure stress on an economy’s currency by computing an exchange rate

pressure index as a weighted average of changes in the nominal effective exchange rates and

in foreign exchange reserves. The index is a variant of the index proposed by Eichengreen,

Rose and Wyplosz (1995), taking into account different exchange rate regimes as well as

policy interventions by the respective governments.

18. One notable feature is our modelling of cross-country linkages using both the

financial and trade linkages, similar to Chen, Gray, N’Diaye, Oura and Tamirisa (2010) and

Eickmeier and Ng (2011). We gauge the strength of the time-varying financial

interdependence across economies based on the flow data from the Bank for International

Settlements’ (BIS) consolidated bank lending statistics. In the construction of the foreign

variables for an economy, the weights on trade and financial linkages are determined by the

relative importance of trade and financial flows in that economy (see Appendix C). Our

robustness analysis indicates that varying their relative weights does not significantly change

the results.

Chen, Filardo, He and Zhu (2012) use corporate and term spread reductions to study the impact of US QE measures, and

Kapetanios, Mumtaz, Stevens and Theodoridis (2012) and Pesaran and Smith (2012) evaluate the effects of UK QE

approximating it with a 100-basis-point reduction in UK term spreads or in the medium- to long-term government bond

yields.

19. In addition, we use a new series of BIS total credit to the non-financial private

sector.

The BIS series on average has a span of 45 years and is available for 40 advanced

and emerging economies.

The database accounts for credit from all sources, not only that

extended by domestic banks.

B. GVECM Analysis: Impulse Responses

20. To estimate the impulse responses, we identify the monetary policy shocks using a

recursive Cholesky scheme, with the following ordering of the endogenous variables in the

US VECM: real GDP growth, CPI inflation, monetary policy indicator, VIX index, equity

price inflation, credit growth, and foreign exchange pressure. The ordering is consistent with

the existing VAR literature. Having explored a number of alternative orderings, we find our

results largely robust. In addition, we follow Dees, di Mauro, Smith and Pesaran (2007) by

assuming that the US economy affects but does not respond to developments in other

economies contemporaneously. This is equivalent to placing the US model as the first

country bloc in the GVECM.

21. We estimate two different GVECMs, one with the term spread as the monetary

policy indicator for the advanced economies, the other with the corporate spread.

Correspondingly, we have two sets of results, one for the US monetary policy shock in terms

of the term spread and the other in terms of the corporate spread.

22. The GVECMs are estimated for the crisis period starting from the outbreak of the

US subprime mortgage crisis in July 2007 to February 2013,

for four advanced economies:

the United States, the euro area, Japan and the United Kingdom; nine emerging Asian

economies: China, Hong Kong SAR, India, Indonesia, South Korea, Malaysia, the

Philippines, Singapore and Thailand; and four Latin American economies: Argentina, Brazil,

Chile and Mexico.

Domestic effects of US term and corporate spread shocks

23. Figure 1 presents two sets of impulse responses for the US economy. One set refers

to responses to a one-standard-deviation cut in the US term spread of 14.2 basis points, the

other to a one-standard-deviation (20.7 basis points) reduction in the US corporate spread.

The “private non-financial sector” includes non-financial corporations (both private- and public-owned), households and

non-profit institutions serving households as defined in the System of National Accounts 2008. In terms of financial

instruments, credit covers loans and debt securities.

Details of the new BIS credit series can be found at: www.bis.org/statistics/credtopriv.htm. Also see Dembiermont,

Drehmann and Muksakunratana (2013).

We focus on the period following the crisis, when the Federal Reserve implemented unconventional monetary policy

measures. This sample period better captures the more recent domestic and international transmissions, which might have

changed after the crisis. Chen, Filardo, He and Zhu (2012, 2014a) provide estimates for the pre-crisis period from February

1995 to June 2007.

24. Notably, as in many studies based on the global VAR models, the confidence bands

tend to be wide.

This is largely due to the limited degrees of freedom in the estimation with

many variables having relatively short time spans. Our confidence bands are subject to the

same limitation, since we focus on the crisis period where the data sample is very short and

the economic and policy uncertainties are particularly elevated.

To improve accuracy, we

exclude from the estimation of each country model those foreign variables considered less

likely to affect or be affected by the economy.

25. Several interesting results emerge. First, US credit growth begins to have a

statistically significant and persistent positive response to a term spread shock in six months’

time: a credit channel might be present as a 14.2-basis-point cut has sustained credit growth

of over 0.3 percentage points higher thereafter. However, the term spread reduction typically

has small and not statistically significant effects on US output growth, and it lowers CPI

inflation and equity prices initially. It also raises the VIX by over 6 percent initially, with

statistically significant effects in the first three months after the shock. This suggests that a

decline in the US term spread may be perceived negatively by markets, for example as a

harbinger of less encouraging prospects.

26. Second, estimates based on the corporate spread model suggest that different

channels might be at play. Lowering the term spread has less impact on output, and over time

it depreciates the US dollar. In contrast, a 20.7-basis-point reduction in the US corporate

spread has a strong, positive and statistically significant impact on US growth, elevating real

GDP growth by 0.2 percentage points throughout the three-year horizon. A cut in the US

corporate spread consistently boosts equity price and CPI inflation, but it raises credit growth

by less than 0.1 percentage points, and it has little impact on the exchange rate.

27. Consistent with the findings in Blinder (2012),

it apparently pays off to take

actions that target corporate borrowing costs rather than indirectly driving down such costs

by purchasing Treasury securities to lower long-term sovereign yields. This corroborates the

earlier findings in the literature that LSAP1 had a larger impact than later asset purchases,

since the LSAP1 program included an important component of private asset purchases (i.e.

agency debt and agency MBS)

Examples include Pesaran and Smith (2006) and Dees, di Mauro, Smith and Pesaran (2007), where the 90 percent

bootstrapped error bands around the mean estimates of impulse responses are generally large and include zero. Chudik and

Fratzscher (2012) instead use the 25th and 75th percentiles as the range of their error bands.

We compute bootstrap confidence intervals with 5000 iterations and provide 90 percent bootstrapped error bands for the

median impulse response estimates.

The persistent response of real GDP growth (and other variables) to a term- or corporate-spread shock may reflect our

choice of not imposing money neutrality while identifying the monetary policy shock in our GVECM, where the real GDP

growth is an I(1) process in most economies.

Blinder (2012) argues that “this particular brand of unconventional monetary policy (purchases of private-sector securities

to reduce risk premiums) appeared to work very well in the cases of CP and MBS. But, of course, the risk spreads were then

at crisis levels. One cannot expect such strong effects under more normal market conditions. That said, every private debt

market is less deep and less liquid than the Treasury markets. So it is reasonable to expect more interest rate ‘bang’ for each

‘buck’ of asset purchases.”

See, for example, Meaning and Zhu (2011) and Goodhart and Ashworth (2013).

Figure 1. Impulse Responses to US Term and Corporate Spread Shocks: United States

1,2

Real GDP growth

CPI inflation

Growth of credit to private

sector

Percentage points

VIX

Equity price

Foreign exchange pressure

Per cent

The estimates correspond to the crisis sample ranging from July 2007 to February 2013.

The US term spread shock is a one-standard-deviation

(i.e. 14.2 basis points) negative innovation to US term spread, and the US corporate spread shock is a one-standard-deviation (i.e. 20.7 basis points)

negative innovation to US corporate spread.

A rise in the foreign exchange pressure index represents stronger appreciation pressure.

Source: Authors’ calculations based on an estimated GVECM.

C. Cross-Border Monetary Policy Spillovers

28. We study the cross-border impact of US QE measures using the weighted regional

average impulse responses to a one-standard-deviation shock to US term (14.2 basis points)

and corporate (20.7 basis points) spreads for the other major advanced economies, emerging

Asia and Latin America.

Figure 2 presents, for each individual economy, the corresponding

The impulse responses in each region are presented as the weighted averages of the median impulse response estimates of

the regional economies, the weights being their real GDP shares in the region, based on each economy’s average real GDP

between 1995 and 2013. The weights are similar to those calculated for 2007-2013. The averaging masks sizeable cross-

economy differences, and the “averaged” confidence bands are no longer valid for the average estimates.

maximum impulse responses to a reduction in the US corporate spread; and Tukey boxplots

in Figure 3 provide information on their dispersion in each region.

29. Depending on whether it is a term or corporate spread shock, the non-US economies’

responses vary in terms of the size and direction. The responses to a US corporate spread

shock are typically much larger. In particular, a cut in the US corporate spread tends to

promote persistently higher real GDP growth and inflation in all three regions, with greater

impact in a number of Latin American and emerging Asian economies. This might have

resulted from stronger responses in equity prices, i.e. the cross-border confidence channels

may work better when monetary policy measures focus on driving down the US corporate

rather than the term spread. In addition, lowering the US corporate spread also leads to

typically larger exchange rate appreciation pressure in Latin America and emerging Asia,

implying a stronger exchange rate channel in the QE spillovers to the emerging economies.

30. The effects of US QE measures have differed across economies and variables, with

substantial cross-region differences in the impulse responses to the US spread shocks,

notably in terms of monetary and exchange rate policies. This may indicate different

transmission and adjustment mechanisms in different economies. While monetary policy

loosens in the advanced economies in response to a US term or corporate spread shock, the

emerging economies respond to different types of US easing in different ways. Notably,

besides Argentina, Malaysia and Singapore, monetary policy in most emerging economies

tend to loosen in response to a cut in the US corporate or term spread. More emerging

economies tend to loosen in response to a cut in the US corporate spread. Currencies in the

advanced economies on average depreciate in response to a US term spread shock. But

appreciation pressures tend to rise in Latin America following a cut in the US term or

corporate spread, which turn out to be larger. But some emerging Asian currencies tend to

depreciate while others tend to appreciate.

31. The estimated impulse responses for each economy generally confirm the results

based on the regional averages, but there are sizeable cross-economy differences. To

illustrate this, we first provide some measures of dispersion, e.g. the range and inter-quartile

range, in the estimated maximum impulse responses over a two-year horizon for the 17

economies; we then describe and differentiate the results for the euro area, Brazil and China,

the largest economies from each of the three groupings.

32. Figure 3 presents Tukey boxplots, which summarise the within-region cross-

economy dispersion in each variable’s estimated impulse responses; the responses are to a

one-standard-deviation reduction in the US term (14.2 basis points) and corporate spread

(20.7 basis points). The bottom and top of the boxes indicate the 25th and 75th percentiles of

the maximum impulse response estimates in each region, the bottom and top whiskers

represent the range of the estimates, and the cross indicates the median.

33. The maximum impulse response estimates deliver generally similar messages. In both

cases, the within-region dispersion is sizeable and the estimates differ both in size and sign;

in most cases, the range of impulse response estimates for the emerging economies includes

zero. Moreover, the median estimates, e.g. for the output growth, credit growth, equity price

and foreign exchange pressure, tend to have the expected sign, especially in the case of a

reduction in the US corporate spread. For equity prices, while the median estimates are all

positive in the case of a cut in the US corporate spread, the median responses to a reduction

in the US term spread are negative. This is due to an initial drop in the equity prices which

turned out to be larger than their later persistent rise.

34. Comparing the three regions, the estimates in the non-US advanced economies

typically have a much smaller dispersion, possibly reflecting more similar economic

structures and a higher degree of economic and financial integration, as well as a smaller

number of economies in the group (Figure 2). In contrast, impulse response estimates for the

emerging economies tend to have larger dispersions. Second, the dispersion is generally

greater for the estimated impulse responses to a shock to the US corporate rather than term

spread, except for equity price and foreign exchange pressure. We focus our discussions on

the impulse response estimates for three economies: the euro area, Brazil and China.

35. A US easing raises euro area inflation. Following a 14.2-basis-point cut in the US

term spread, the euro area term spread falls significantly and stays lower by over 10 basis

points during most of the three-year horizon. The almost one-to-one response shows a tight

relationship between the two economies. A 20.7-basis-point cut in the US corporate spread

also lowers the euro area term spread. It drives up euro area credit and output growth by

about 0.1 and 0.2 percentage points, respectively. Euro area equity price inflation rises by

over 1 percentage point in four months. Reducing the US corporate spread depreciates the

euro by about 0.5 percentage points, but lowering the US term spread has little effects on the

euro exchange rate.

36. In Brazil, while money growth rises in response to a US term spread shock, it falls in

response to a US corporate spread shock. The Brazilian equity price rises slightly and then

stays almost unchanged after an initial decline of up to 2.4 percent upon a US term spread

shock, but it rises consistently at 1.2 percent or more four months after a cut in the US

corporate spread. Credit and output growth generally accelerate and currency appreciation

pressure rises following a US easing, with a stronger eventual impact from the US corporate

spread shock.

The impulse response estimates for all 17 economies, with the respective confidence bands, are available upon request.

They tend to be large in many cases, but often not significantly different from zero due to the estimation difficulties with

large-scale GVECMs and the small crisis sample.

Figure 2. Maximum Impulse Responses to a US Corporate Spread Shock

Real GDP growth

CPI inflation

Percentage points

Credit growth to the private sector

Monetary policy indicator

Percentage points

Equity price

Foreign exchange pressure

Per cent

AR = Argentina; BR = Brazil; CL = Chile; CN = China; GB = United Kingdom; HK = Hong Kong SAR; ID = Indonesia; IN = India; JP = Japan; KR =

Korea; MX = Mexico; MY = Malaysia; PH = Philippines; SG = Singapore; TH = Thailand; US = United States; XM = Euro area.

The US corporate spread shock is a one-standard-deviation (i.e. 20.7 basis points) negative innovation to the corporate spread.

For monetary policy

indicators, we use corporate or term spreads for the advanced economies, and growth rates of a broad monetary aggregate for emerging economies.

rise in the foreign exchange pressure index represents stronger appreciation pressure.

Source: Authors’ calculations based on an estimated GVECM.

Figure 3. Distribution of Maximum Impulse Responses to US Term

and Corporate Spread Shocks

1,2

Real GDP growth

CPI inflation

Credit to the private sector

growth

Percentage points

Monetary policy indicator

Equity price

Foreign exchange pressure

Percentage points

Per cent

The US term spread shock is a one-standard-deviation (i.e. 14.2 basis points) negative innovation to the US term spread, and the US corporate spread

shock is a one-standard-deviation (i.e. 20.7 basis points) negative innovation to the US corporate spread.

In the Tukey boxplots the bottom and top of

the boxes are the first and third quartiles of the cumulative impulse responses of the region; the cross indicates the median; and the bottom and top

whiskers represent the range of the responses.

Euro area, Japan and the United Kingdom.

China, Hong Kong SAR, India, Indonesia, Korea, Malaysia,

the Philippines, Singapore and Thailand.

Argentina, Brazil, Chile and Mexico.

For monetary policy indicators, we use term and corporate spreads for

the advanced economies, and the growth rates of a broad monetary aggregate for the emerging economies.

A rise in the foreign exchange pressure index

represents stronger appreciation pressure.

Source: Authors’ calculations based on an estimated GVECM.

37. China’s estimated policy responses to US stimulus differ depending on the nature of

the US shock. Following a cut in the US term spread, China’s money and credit growth rates

drop by 0.2 and 0.3 percentage points in the second month after the shock, they then turn

slightly positive in a few months before a persistent decline. In response to a drop in the US

corporate spread, however, the money and credit growth rates rise modestly for about six

months before falling persistently thereafter. For both shocks, the Chinese yuan faces

persistent depreciation pressures due to its close association with the US dollar, but the

pressure is greater during the first 20 months following a cut in the US corporate spread,

being significant and reaching 0.23 percentage points in one month.

38. The evidence suggests that reducing the US corporate spread is more

accommodative overall for the Chinese economy, even though China tends to lean against it

with monetary and credit policies. Real GDP growth increases significantly and rapidly by

0.13 percentage points following a 20.7-basis-point cut in the US corporate spread, despite an

initial drop, and eventually, inflation rises persistently by about 0.16 percentage points. The

output response to a US term spread shock is smaller, and the inflation response is mostly

negative. Equity price rises following a US corporate spread shock, and it generally falls after

a US term spread shock.

39. One interesting finding is that US QE measures turn out to have a greater impact on

economic and financial variables in many emerging economies than on the US economy.

This is consistent with previous work. For example, Mackowiak (2007) finds that US

monetary policy shocks in the pre-crisis period quickly and strongly affect interest and

exchange rates in a typical emerging economy, and price and real output there respond more

than the US counterparts. This evidence supports the view that cross-border monetary policy

spillovers cannot be dismissed as insignificant by-products of little consequence for the

global economy. As Rajan (2013) puts it, “even if the unconventional monetary policies that

focus on lowering interest rates across the term structure have limited effects on interest rates

in the large, liquid, sending country Treasury markets, the volume of flows they generate

could swamp the more illiquid receiving country markets, thus creating large price and

volume effects”.

Robustness check

40. The results of impulse response analyses are robust to different model specifications

and variable definitions, including the use of base money growth instead of broad money

growth, the use of the federal funds rate instead of the three-month US Treasury bill rate to

calculate the US term spread, and the use of the three-month US Treasury bill rate instead of

the federal funds rate to calculate the US corporate spread. The results are also robust to

alternative orderings of the variables in the identification scheme for the shocks to the US

term and corporate spreads.

III. GVECM-BASED COUNTERFACTUAL ANALYSIS

41. We conduct counterfactual analyses to evaluate the domestic and global impact of

US QE measures. We construct two counterfactual scenarios based on US corporate spread

developments.

We then make conditional forecasts for model variables based on the

assumption that the US corporate spread follows a predetermined counterfactual path. By

comparing these projections to the actual data, we can assess the overall impact on the global

We present the results for the counterfactual analysis based on US corporate spread developments, given that their

economic effects are larger. Details on the counterfactual analysis based on US term spread developments can be provided

upon request.

economy of the US QE measures, and the separate impacts of the LSAP1, LSAP2, MEP and

LSAP3 programs.

42. The counterfactual analysis is based on the estimated GVECM model and one-step-

ahead projections. Specifically, equation (8) in Appendix A illustrates how an endogenous

variable 



can be expressed as the sum of the lagged explanatory variables (both domestic

and foreign) multiplied by the corresponding parameter estimates, plus the estimated

residuals. Given the values of all model variables up to time t, and conditional on the time-t

counterfactual value of the corporate spread, we obtain the one-step-ahead forecasts for the

endogenous variables (





). In the next step, we use the forecasts 





and the time t+1

counterfactual value of the corporate spread to obtain the time t+2 forecasts (





), and so

on. The forecasts of each endogenous variable therefore depend on the past forecasts of the

other variables and the specified US monetary policy path.

43. We design two different counterfactual scenarios:

1. Constant scenario: we assume that the US corporate spread remains constant within

each period of the QE program, at the level seen immediately before the

implementation of each US asset purchase program, namely LSAP1, LSAP2, MEP

and LSAP3;

2. Jump scenario: we assume the US corporate spread jumps by 200 basis points at the

start of each QE program, thereafter it stays above the actual values during the entire

program.

44. Figure 4 shows both the actual and the two counterfactual paths for the US corporate

spread.

45. Our counterfactual analyses suggest that US QE measures had sizeable domestic

effects, and such effects varied substantially depending on whether the measures led to a

substantial fall in the US corporate spread. In cases where the Federal Reserve asset

purchases kept the US corporate spread at a low level relative to the baseline, such actions

appeared to have contributed to stronger US economic recovery.

In doing this exercise, we need to bear in mind that the actual data could also reflect many other factors affecting the

global economy; these may include supply-side shocks such as euro area sovereign debt crisis, large fiscal stimulus in China,

and commodity price fluctuations.

We also examine an “increasing scenario”, in which the US corporate spread is assumed to rise by 10 basis points, in each

and every month during each QE program. As the results are similar, we only present the results associated with the

“constant” and “jump” scenarios.

Figure 4. Counterfactual Analysis of Domestic US QE Impact:

US Corporate Spread

Real GDP growth

CPI inflation

Credit growth to the private

sector

Percentage points

Corporate spread

Equity price

Foreign exchange pressure

Percentage points

Natural logarithm

Per cent

The grey areas indicate the periods of implementation of LSAP1, LSAP2 and MEP. The black lines are actual values. The red lines are the values

associated with the jump scenario where the US corporate spread jumps by 200 basis points and stays 200 basis points above the actual levels

throughout the respective QE program, and the blue lines depict the scenario where the US corporate spread stays equal to the actual level observed just

before the QE program.

A rise in the foreign exchange pressure index represents stronger appreciation pressure.

Source: Authors’ calculations based on an estimated GVECM.

46. Figure 4 presents the dynamics of US economic and financial variables under the

“constant” and “jump” scenarios for the US corporate spread.

The counterfactual exercise,

at face value, indicate that without QE, especially the purchases of agency MBS and agency

debt which helped to lower the US corporate spread, the US economy would have remained

mired in a recession with deflation. The “jump” scenario suggests that asset purchases may

have supported higher real GDP growth by about 8 percentage points, and inflation by over

Notice that during LSAP1, the US corporate spread actually drifted back up midway through the program to levels higher

than when LSAP1 began, and then kept climbing during LSAP2 (Figure 4). This can be interpreted to suggest that the

LSAP1 and LSAP2 programs had a transitory impact on the US corporate spread, and would raise questions about whether

the “constant” scenarios are truly “stress” scenarios. Another possible explanation is that other factors such as adverse

supply shocks or further financial sector strains could have diluted the effects of the asset purchases and pushed the US

corporate spread higher than otherwise.

5 percentage points. Equity prices would have been much lower without the QE measures.

The “constant” scenario indicates even greater effects. In addition, while the “jump” scenario

points to appreciation pressures on the US dollar during much of the implementation of each

program, cumulatively the programs exerted sizeable depreciation pressures. Moreover, the

counterfactual exercise indicates that the programs actually slowed credit growth.

47. In sum, the counterfactual exercises suggest that the domestic effects of different US

QE measures were diverse. In the model, QE programs which reduced the US corporate

spread delivered a sizeable stimulus to US output growth and equity markets and led to

substantial currency depreciation. If the counterfactual exercise is seen as a reasonable

approximation to what would have happened, the findings suggest that QE programs helped

to prevent the US economy from sliding into a prolonged recession with severe deflation.

The results suggest that if policymakers aim to lower private-sector borrowing costs and

stimulate growth, it pays to design programms to influence the corporate spread.

A. Spillovers from Reductions in the US Corporate Spread

48. The results in this section show that the cross-border spillover effects from US QE

policies that reduce the US corporate spread are sizeable. We present the results on the euro

area, Brazil and China.

49. In the counterfactual analysis, US unconventional policies are estimated to have had

an important impact on the euro area (Graph 3.2.1): the lower US corporate spread is

estimated to have supported euro area credit and output growth, with the impact ranging from

3 and 8 percentage points (jump scenario) to 7 and 16 percentage points (constant scenario),

respectively, significantly boosting equity prices.

50. The analysis also suggests that US QE measures had even greater spillover effects

on the emerging economies, again much through the reduction in the US corporate spread.

The estimated impact tended to be diverse across economies and across variables, which may

reflect diverse policy responses, exchange rate regimes and economic structures.

51. The evidence from the counterfactual exercise also suggests that lowering the US

corporate spread stimulated Brazil’s output growth while having little impact on inflation.

Arguably, this evidence suggests that LSAP1 helped the Brazilian economy recover rapidly

from the 2009 recession, and that MEP and LSAP3 might have helped Brazil avoid a

possible recession in 2012. But LSAP2 began when Brazil’s output growth reached a peak of

almost 8 percent, and therefore might be interpreted as having contributed to Brazil’s

overheating at the time. These Brazilian episodes highlight that the perception of monetary

policy spillovers may be influenced by the receiving country’s cyclical position.

Figure 5. Counterfactual Analysis of Euro Area Impact: US Corporate Spread

Real GDP growth

CPI inflation

Credit growth to the private

sector

Percentage points

Term spread

Equity price

Foreign exchange pressure

Percentage points

Natural logarithm

Per cent

The grey areas indicate the periods of implementation of LSAP1, LSAP2 and MEP. The black lines are actual values. The red lines are the values

associated with the jump scenario where the US corporate spread jumps by 200 basis points and stays 200 basis points above the actual levels throughout

the respective QE program, and the blue lines depict the scenario where the US corporate spread stays equal to the actual level observed just before the

QE program.

A rise in the foreign exchange pressure index represents stronger appreciation pressure.

Source: Authors’ calculations based on an estimated GVECM.

Figure 6. Counterfactual Analysis of Impact in Brazil: US Corporate Spread

Real GDP growth

CPI inflation

Credit growth to the private

sector

Percentage points

Money growth

Equity price

Foreign exchange pressure

Percentage points

Natural logarithm

Per cent

The grey areas indicate the periods of implementation of LSAP1, LSAP2 and MEP. The black lines are actual values. The red lines are the values

associated with the jump scenario where the US corporate spread jumps by 200 basis points and stays 200 basis points above the actual levels throughout

the respective QE program, and the blue lines depict the scenario where the US corporate spread stays equal to the actual level observed just before the

QE program.

A rise in the foreign exchange pressure index represents stronger appreciation pressure.

Source: Authors’ calculations based on an estimated GVECM.

52. The counterfactual exercise provides evidence that US QE programs had an

expansionary spillover to the Chinese economy, but lower US corporate spreads were less

expansionary than in the case of Brazil, with China’s real GDP growth being boosted by 2.5

(jump) to 5.5 (constant) percentage points by the end of 2012 (Figure 7). One possible reason

for the weaker impact is the apparently tighter Chinese monetary and credit conditions that

accompanied the lowering of the US corporate spread: cumulatively, money and credit

growth were estimated to be lower by up to 8 and 15 percentage points, respectively,

compared to the jump and constant counterfactuals. As well, currency appreciation pressures

rose moderately relative to the counterfactuals since mid-2010. Taken together, the evidence

suggests that the responses in money growth, credit growth and exchange rate pressure

tended to offset the accommodative spillover effects from the US monetary stimulus.

However, as in the case of Brazil, the timing of the estimated spillover from US monetary

policy suggests that it contributed to China’s overheating, i.e. when China’s output growth

exceeded 9 percent in 2010–11, and inflation was over 5 percent in 2011.

Figure 7. Counterfactual Analysis of Impact in China: US Corporate Spread

Real GDP growth

CPI inflation

Credit growth to the private

sector

Percentage points

Money growth

Equity price

Foreign exchange pressure

Percentage points

Natural logarithm

Per cent

The grey areas indicate the periods of implementation of LSAP1, LSAP2 and MEP. The black lines are actual values. The red lines are the values

associated with the jump scenario where the US corporate spread jumps by 200 basis points and stays 200 basis points above the actual levels throughout

the respective QE program, and the blue lines depict the scenario where the US corporate spread stays equal to the actual level observed just before the

QE program.

A rise in the foreign exchange pressure index represents stronger appreciation pressure.

Source: Authors’ calculations based on an estimated GVECM.

IV. CONCLUSION

53. We examine the domestic and cross-border effects, both real and financial, of the

Federal Reserve’s unconventional monetary policies using an estimated GVECM. First, we

find that QE measures which lower the US corporate spread have had sizeable effects, which

There are at least two key reasons for the Chinese economy being less affected by US QE than Brazil. First, the Chinese

economy was far bigger and more diversified, capable of absorbing large external shocks. Second, our results suggest that

China responded to the US stimulus with tighter monetary and credit policies and cushioned the impact of the stimulus on

the yuan exchange rate.

vary significantly across regions and individual economies. This is consistent with Blinder

(2012) that purchasing US Treasuries to lower the term spread may be a weak tool, and

reducing risk premia by acquiring private-sector assets is much more potent. Second,

monetary policy and exchange rate responses have been diverse in the emerging economies,

which may partly explain the important cross-economy differences in the responses of

output, inflation and credit. Third, US QE measures have had sizeable and widespread effects

on global equity prices, and the confidence channel may be important. Fourth, such measures

tend to have a greater impact on many emerging economies than on the US economy.

54. Our counterfactual analyses suggest that, first, US QE measures have supported the

advanced economies, countering forces which might have led to a prolonged recession and

deflation. Second, the cross-border impact of such measures appear to have helped support

the recovery in the emerging economies in 2009 and 2012, as well as contributing to their

overheating in 2010 and 2011. Third, some of the differential cross-border impacts might

depend on policy responses to the US policy actions. For example, there is evidence that

China tightened the monetary and credit conditions in response to lower US corporate

spreads, which tended to partially offset their expansionary impact on its output growth. In

Brazil, the evidence suggests that a sizeable credit expansion led to more expansionary

overall economic and financial conditions.

55. There is evidence that cross-border monetary policy spillovers can be important

sources of macroeconomic and financial instability. This raises important questions about

whether central banks should do more to take into account the unintended consequences of

their actions on others and how to best promote stability. While having made some progress

in quantifying monetary policy spillovers, we are still some distance from being able to

assess the nature, as well as the costs and benefits, of international policy coordination.

As Cœuré (2014) points out, “the case for formal central bank cooperation remains limited, and practical consideration

makes its implementation difficult,” yet “central banks need to be engaged in a constant dialogue so as to remain ready for

rapid coordinated action in exceptional circumstances.”

APPENDIX I. STRUCTURE OF THE GVECM MODEL

Consider N+1 economies, indexed by i = 0, 1, 2, ..., N, and for the i-th economy, a vector 



of 



domestic variables. By stacking the vectors of country-specific variables, we have

















Ntttt

xxxx ...,,,

(1)

A VAR model in



would contain too many parameters to be estimated if the data’s time

dimension T is large enough relative to N+1, the number of economies.

Instead of regressing 



on 



, where

















 tNtitittti ,,1,110,

,...,,,,...,,, xxxxxx

(2)

the GVECM links

to a 





  vector



, where









iljtlijtlit

klwithxx

,,2,1 



(3)

The weight

lijt



captures the spillover effect of variable l of foreign economy j on variable l

of domestic economy i. Since

lijt



measures the relative importance of economy j to

economy i, the spillover effect of variable l is in proportion to the weight chosen to measure

the relative strength. Hence each economy’s component model of the GVECM is given as a

VECM*



 



:



















,1,10

itstisiti

stiistiiiit

t εdxΓzΨzΠccx

(4)

where 



is the observed common factor (

1q

) and

 



iid

,0~ε

The economy i, the vector 





reflects its interdependence with other economies, and it

serves as a proxy for the unobserved common effects across the economies. The foreign

variables and common factors are assumed to be weakly exogenous, i.e., they are “long-run

forcing” the domestic variables, in the sense that the coefficients on the error-correction

terms are set to zero in the equations for foreign variables. The dynamics of foreign variables

are unaffected by any deviations from the long-run equilibrium path, in contrast to the

dynamics of domestic variables.

The VECMX* can be estimated for each economy with the ordinary least squares (OLS) or

rank-reduced approach if the cross-dependence of the idiosyncratic shock is sufficiently

small:

(5)

for all   , l and s.

From equation (3), it can be seen that

tiit

xWz 

for

Ni ,,2,1 

(6)

where 



 





 







 and 



is a properly defined weight matrix. Stacking (4) across

, the

endogenous variables can be solved for a global system:

 

 





tstsstsiit

1 0

udΨxΦaaGx

(7)

and

 

 







tstsstsiit

1 0

uGdΨGxΦGaGaGx

11111

(8)

where  max







 





,   









and

(9)

Equation (8) is a VAR for the complete set of domestic variables for all economies. The

GVECM model makes it feasible to estimate (8) by explicitly taking into account the cross-

economy interdependence while estimating each economy separately, allowing the inclusion

of a large number of economies. The impulse responses are then estimated based on (8).

We conduct the augmented Dickey-Fuller (ADF) and the weighted-symmetric Dickey-Fuller

(WSDF) unit root tests for all model variables. The two tests produce broadly similar results.

At the 5 percent significance level, we find that in most economies, the domestic variables

are tested to be integrated of order 1, i.e. I(1), with the exception of some variables being

tested to be I(0) or near I(1). These include: based on the WSDF test results, real GDP

growth in Hong Kong, Thailand and the United Kingdom, and CPI inflation in Chile, China

and the Philippines, foreign exchange pressure indices of Argentina and the euro area, equity

price inflation in Hong Kong, Indonesia and South Korea, and monetary aggregates in India;

and based on the ADF test results, the foreign exchange pressure index of India and monetary

aggregates in Singapore. Most foreign variables are tested to be I(1), so is the global factor

VIX. The form in which the model variables are included in the GVECM ensures a stable

global solution with an eigenvalue less than or equal to 1.

 

, / 0,

it sjt

Cov N









,0 0 0,

,1 1 1,

, , .

s N N N t

B W u

G H u

B W u

   



   



   



  

   



   





   

APPENDIX II. FOREIGN EXCHANGE PRESSURE INDEX

The foreign exchange pressure index

emp

measures the pressure of capital inflows. In

economies with flexible exchange rate regimes, strong net capital inflows push up the

demand for domestic currency, which in turn leads to its appreciation. If the authorities

intervene in the foreign exchange market to moderate the currency appreciation, we may not

observe significant changes in the exchange rates, but rather a rise in the foreign reserves. In

a fixed-exchange-rate regime, strong capital inflows are reflected in an increase of foreign

reserves only. We therefore construct the foreign exchange pressure index as follows, similar

to Eichengreen, Rose and Wyplosz (1995):

)(100

,, trevttett

revwewemp 









revtet





, for

reveX ,

where



is the standard deviation of the corresponding variable in the previous five years,

for   . For   , we use the standard deviation based on the data for the first five years.

Moreover,

)ln()ln(

12



ttt

EEe

and

)ln()ln(

12



ttt

RRrev

, where

is the nominal

effective exchange rate and

denotes the foreign reserves.

APPENDIX III. TIME-VARYING WEIGHTS FOR FOREIGN VARIABLES

The weight that an economy-i foreign variable assigns to economy j at year t is

tij

agg

tij

WwWwW

,,,,,



for all

ji 

where

tij

and

tij

are the bilateral trade and financial weight (based on capital flows in the

previous year), respectively.

and

are the relative importance of trade and capital

flows in an economy, respectively. They are computed according to the values of the

respective aggregate trade flow (export and import) and capital flow (both inflow and

outflow) relative to the total value of these two types of flows in the previous year. The

financial weight of economies with no capital flow data in the 1990s is set to zero.

APPENDIX IV. DATA

Data sources include the Bank for International Settlements (BIS), the International

Monetary Fund’s International Financial Statistics, CEIC, Bloomberg and Datastream:

Variable

Description

Source

Notes

Real GDP (y)

IMF IFS,

national data

Real GDP of China is at 1990 prices,

those of other countries at 2005 prices

(billions of domestic currency units). The

monthly time series are interpolated using

the method of Chow and Lin (1971) with

industrial production series as a

reference. Series for HK is interpolated

using compound growth rate due to

unavailability of monthly industrial

production.

CPI inflation

(π)

Year-on-year

change in consumer

price index

CEIC, IMF

IFS, national

data

Credit (bc)

Total credit to the

non-financial private

sector

BIS

Term spread

(mp)

Difference between

10-year US

Treasury bond yield

and 3-month US

Treasury bill rate

CEIC, IMF,

IFS, national

data

For euro area, due to data limitations, the

main refinancing rate is used instead of

3-month government bond yield.

US corporate

spread (mp)

BofA Merrill Lynch

US Corporate AAA

minus the federal

funds rate.

CEIC, IMF,

IFS, national

data

Implied

volatility (VIX)

CBOE Volatility

Index; in natural

logarithm

CBOE

VIX is a key measure of market

expectations of near-term volatility

conveyed by S&P 500 stock index option

prices.

Money growth

(mp)

Year-on-year M2

growth rate

CEIC, IMF

IFS

Equity price

(sp)

Stock price index; in

natural logarithm

Bloomberg

Index of stock prices in each economy.

Foreign

exchange

pressure

(emp)

Nominal effective

exchange rate

BIS

Period average; 2005 = 100.

Foreign reserves

IMF IFS

Total reserves minus gold, in billions of

USD. Euro area data starting from

Jan 1999 are official reserves as

published by ECB; data before 1999 are

either an estimate or the aggregate

reserves of 11 EU Member States

participating in the euro area in 1999.

Oil price

Spot oil price

IMF IFS

Brent crude oil, US dollars per barrel;

period end data.

Export/import

IMF IFS

Cross-border

bank lending

BIS consolidated

bank lending

statistics

BIS

Capital inflow

and outflow

IMF IFS

REFERENCES

Bank for International Settlements, 2012. BIS Annual Report.

Blinder, A., 2010. Quantitative easing: entrance and exit strategies. Federal Reserve Bank of

St. Louis Review, November/December.

Blinder, A., 2012. Revisiting monetary policy in a low inflation and low utilization

environment. J. Money Credit Bank. Supplement, 44(1), 141-6.

Chen, Q., Filardo, A., He, D., Zhu F., 2012. International spillovers of central bank balance

sheet policies. BIS Papers No. 66, Bank for International Settlements.

Chen, Q., Filardo, A., He, D., Zhu F., 2014a. Financial crisis, unconventional monetary

policy and international spillovers. HKIMR Working Paper No.23/2014, Hong Kong

Institute for Monetary Research.

Chen, Q, A Filardo, D He, F Zhu., 2014b. Global impact of US monetary policy at the zero

lower bound. manuscript, Bank for International Settlements.

Chen, Q., Gray, D., N'Diaye, P., Oura, H., Tamirisa, N., 2010. International Transmission of

Bank and Corporate Distress. IMF Working Paper No. 10/124, International

Monetary Fund.

Chow, G, Lin, A., 1971. Best linear unbiased interpolation, distribution, and extrapolation of

time series by related series. Rev. Econ. Stat. 53(4), 372-375.

Chudik, A., Fratzscher, M., 2012. Liquidity, risk and the global transmission of the 2007-08

financial crisis and the 2010-2011 sovereign debt crisis. ECB Working Paper Series

1416, European Central Bank.

Cœuré, B., 2014. The internationalisation of monetary policy. Keynote address at the ECB-

IMF conference on the “International dimension of conventional and unconventional

monetary policy”. Frankfurt, 30 April.

D’Amico, S., King, T., 2010. Flow and stock effects of large-scale treasury purchases.

Finance and Economics Discussion Series, no 2010-52, Board of Governors of the

Federal Reserve System.

Dembiermont, C., Drehmann, M., Muksakunratana, S., 2013. How much does the private

sector really borrow - a new database for total credit to the private non-financial

sector. BIS Quarterly Review, March, 65–81.

De Nicolò, G., Dell’Ariccia, G., Laeven, L., Valencia, F., 2010. Monetary policy and bank

risk taking. IMF Staff Position Note, no SPN/10/09.

Dees, S., di Mauro, F., Smith, V., Pesaran, H., 2007. Exploring the international linkages of

the euro area: a global VAR analysis. J. Appl. Econ. 22(1), 1–38.

Doh, T., 2010. The efficacy of large-scale asset purchases at the zero lower bound. Federal

Reserve Bank of Kansas City Economic Review.

Eichengreen, B., Rose, A., Wyplosz, C., 1995. Exchange market mayhem: The antecedents

and aftermath of speculative attacks. Econ. Policy 21, 249-312.

Eickmeier, S., Ng T., 2011. How do credit supply shocks propagate internationally? A

GVAR approach. CEPR Discussion Paper, 8720.

Fratzscher, M., Lo Duca, M, Straub, R., 2013. On the international spillovers of US

quantitative easing. DIW Discussion Papers, no 1304.

Gagnon, J., Raskin, M., Remache, J., Sack, B., 2010. Large-scale asset purchases by the

Federal Reserve: did they work? FRB New York Staff Reports.

Gagnon, J., Raskin, M., Remache, J., Sack, B., 2011. The financial market effects of the

Federal Reserve’s large-scale asset purchases. Int. J. Central Bank. 3-43.

Gambacorta, L., Hofmann, B., Peersman, G., 2012. The effectiveness of unconventional

monetary policy at the zero lower bound: a cross-country analysis. BIS Working

Papers, no 268.

Glick, R., Leduc, S., 2012. Central bank announcements of asset purchases and the impact on

global financial and commodity markets. J. Int. Money Finance 31(8), 2078-2101.

Goodhart, C., Ashworth, J., 2013. QE: a successful start may be running into diminishing

returns. Oxf. Rev. Econ. Policy, 28(4), 640-70.

Hofmann, B., Zhu, F., 2013. Central bank asset purchases and inflation expectations. BIS

Quarterly Review, March, 23-35.

International Monetary Fund, 2013a. 2013 spillover report – analytical underpinnings and

other background. IMF Policy Paper.

International Monetary Fund, 2013b. Global impact and challenges of unconventional

monetary policies. IMF Policy Paper, October.

Joyce, M., Lasaosa, A., Stevens, I., Tong, M., 2011. The financial market impact of

quantitative easing. Int. J. Central Bank. 113-161.

Kapetanios, G., Mumtaz, H., Stevens, I., Theodoridis, K., 2012. Assessing the economy-wide

effects of quantitative easing. Econ. J. 122(564), 316-47.

Krishnamurthy, A., Vissing-Jorgensen, A., 2011. The effects of quantitative easing on

interest rates. Brookings Pap. Econ. Act. 43(2), 215-265.

Mackowiak, B., 2007. External shocks, U.S. monetary policy and macroeconomic

fluctuations in emerging markets,” J. Monet. Econ. 54(8), 2512-20.

Meaning, J., Zhu, F., 2011. The impact of recent central bank asset purchase programmes.

BIS Quarterly Review, December, 73-83.

Meaning, J., Zhu, F., 2012. The impact of Federal Reserve asset purchase programmes:

another twist. BIS Quarterly Review, March, 23-30.

Neely, C., 2010. The large scale asset purchases had large international effects. Federal

Reserve Bank of St. Louis Working Paper, no 2010-018, July.

Obstfeld, M., Rogoff, K., 2002. Global implications of self-oriented national monetary rules.

Q. J. of Econ. 117(2), 503–36.

Ostry, J., Ghosh, A., 2013. Obstacles to international policy coordination, and how to

overcome them. IMF Staff Discussion Notes 13/11, International Monetary Fund.

Pesaran, M.H., Schuermann, T., Weiner, S., 2004. Modeling regional interdependencies

using a global error-correcting macroeconometric model. J. Bus. Econ. Stat. April.

Pesaran, M.H., Smith, R., 2006. Macroeconometric modelling with a global perspective.

Manchester School, 74(s1), 24-49.

Pesaran, M.H., Smith, R., 2012. Counterfactual analysis in macroeconometrics: an empirical

investigation into the effects of quantitative easing. CESifo Working Paper Series

3879.

Rajan, R., 2013. A step in the dark: unconventional monetary policy after the crisis“, Andrew

Crockett Memorial Lecture delivered at the BIS, 23 June.

Rogers, J, Scotti, C., Wright, J., 2014. Evaluating Asset-Market Effects of Unconventional

Monetary Policy: A Cross-Country Comparison. International Finance Discussion

Papers No 1101, Board of Governors of the Federal Reserve System.

Taylor, J., 2013. International monetary policy coordination: past, present and future. BIS

Working Papers 437, Bank for International Settlements.