Article 1

STOCKS, BONDS, MONEY MARKETS AND EXCHANGE RATES: MEASURING
INTERNATIONAL FINANCIAL TRANSMISSION

Author(s): MICHAEL EHRMANN, MARCEL FRATZSCHER and ROBERTO RIGOBON
Source: Journal of Applied Econometrics, Vol. 26, No. 6, Themes in Macroeconomics and
Finance (September-October 2011), pp. 948-974
Published by: Wiley
Stable URL: http://www.jstor.org/stable/23018258
Accessed: 10-04-2017 01:18 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
Wiley is collaborating with JSTOR to digitize, preserve and extend access to Journal of Applied
Econometrics
This content downloaded from 111.68.99.249 on Mon, 10 Apr 2017 01:18:30 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF APPLIED ECONOMETRICS
J. Appl. Econ. 26: 948-974 (2011)
Published online 30 March 2010 in Wiley Online Library
(wileyonlinelibrary.com) DOI: 10.1002/jae.l 173
STOCKS, BONDS, MONEY MARKETS AND EXCHANGE

RATES: MEASURING INTERNATIONAL FINANCIAL
TRANSMISSION
MICHAEL EHRMANN,3* MARCEL FRATZSCHERa AND ROBERTO RIGOBONb

a European Central Bank, Frankfurt, Germany
b Massachusetts Institute of Technology, Cambridge, MA, USA
SUMMARY
Understanding the complexity of the financial transmission process across various assetsdomestically
well as within and across asset classesrequires the simultaneous modeling of the various transmis
channels in a single, comprehensive empirical framework. The paper estimates the financial transmis
between money, bond and equity markets and exchange rates within and between the USA and the e
area. We find that asset prices react strongest to other domestic asset price shocks, but that there are
substantial international spillovers, both within and across asset classes. The results underline the domina
of US markets as the main driver of global financial markets: US financial markets explain, on aver
around 30% of movements in euro area financial markets, whereas euro area markets account only for ab
6% of US asset price changes. Moreover, the methodology allows us to identify indirect spillovers throu
other asset prices, which are found to increase substantially the international transmission of shocks wit
asset classes. Copyright 2010 John Wiley & Sons, Ltd.
Received 11 April 2008; Revised 15 November 2009
1. INTRODUCTION
Financial markets have become increasingly integrated, both domestically and internatio
The nature of this integration and the transmission channels through which shocks dis
are, however, still not well understood. One strand of the literature focuses exclusive
spillovers across different domestic asset prices, whereas another strand concentrates on
national spillovers only for individual asset prices. However, understanding the increas
close domestic and international linkages of asset prices requires a complete and comp
sive modeling of all transmission channels that are at play. Policymakers and practitione
well aware of the existence of these linkages, but very little is known about their strength a
scope, which remains an important open question for researchers and policymakers (Bern
2007).1
The main limitation the literature has faced in measuring these propagation channels has been
the endogeneity of asset prices, even at daily frequencies. Clearly, macroeconomic shocks such
* Correspondence to: Michael Ehrmann, European Central Bank, Kaiserstrasse 29, D-60311 Frankfurt, Germany.
E-mail: Michael.Ehrmann@ecb.int
' Among the few exceptions are Andersen et al. (2007), who study the transmission among stock markets for each country,
and then across countries for each type of asset market separately, as well as Dungey and Martin (2007), who also study
the propagation of shocks across countries and markets. We discuss below in which dimensions our approach differs from
these two papers.
Copyright 2010 John Wiley & Sons, Ltd.
INTERNATIONAL FINANCIAL TRANSMISSION 949
as shocks to productivity, monetary policy, inflation expectations and risk premia have an eff
on all asset prices; and therefore, estimating the impact of one innovation on the others requir
identifying shocks that are unobservable at these frequencies. In this paper, we estimate the p
agation of shocks by modeling each asset price with a multifactor model, and then using
heteroskedasticity that exists in the data for identification and to estimate the contemporaneo
financial transmission coefficients.2
We use this framework to analyze the nature of financial integration and the transmiss
channels within as well as between the two largest economies in the world: the USA and
euro area. The empirical model concentrates on two-daily returns over a 20-year period
1989-2008 for seven asset prices: short-term interest rates, bond yields and equity market retu
in both economies, as well as the exchange rate. In doing so, the paper allows studying
transmission channels that so far have received little attention in the literature, namely the ro
of international cross-market spillovers and the indirect transmission of financial shocks thro
third markets.
The results of the paper underline the importance of international spillovers, both wit
asset classes as well as across financial markets. Although the strongest international t
mission of shocks takes place within asset classes, we find evidence that international c
market spillovers are significant. For instance, shocks to US equity markets exert a sub
stantial influence on euro area short-term interest rates and bond yields. But the tran
sion of shocks also runs in the opposite direction as, for example, short-term interest ra
of the euro area have a significant impact on US bond markets. Overall, US financial m
kets explain on average around 30% of euro area financial market movements in the pe
1989-2008, whereas euro area markets account for about 6% of the variance of US asset
prices.
A second key result of the paper is that the direct transmission of financial shocks within
asset classes is often magnified substantially through indirect spillovers through other asset prices.
In particular, for the case of bond yields, this increase is of the order of 50%. For instance,
the coefficient for the direct effect of shocks to US bond yields on euro area bond markets is
0.49indicating that on average nearly 50% of the shock to US bond yields is transmitted to
euro area yieldsbut it rises to 0.70 when allowing for indirect spillovers of these shocks via
other US and euro area asset prices.
These two results underline that a better understanding of financial linkages requires the mod
eling of international cross-market financial linkages, which so far has been mostly missing in the
literature. We confirm some familiar results of the literature as, in particular, we find that financial
markets are mostly driven by country-specific and market-specific factors. However, we detect
a rich interaction between asset prices domestically and our methodology allows us to quantify
domestic financial market transmissions much more accurately by controlling for foreign and other
types of shocks.
The paper is organized in the following way. Section 2 briefly reviews the literature on
domestic and on international financial linkages and integration. The methodology based on
identification through heteroskedasticity is summarized in Section 3. Section 4 outlines the
empirical findings, focusing on international asset market spillovers and the resulting variance
decomposition. Section 5 summarizes and concludes with some policy implications arising from
the findings.
2 See Wright (1928), Sentana and Fiorentini (2001) and Rigobon (2003) for the theory of the methodology.
Copyright 2010 John Wiley & Sons, Ltd. J. Appl. Econ. 26: 948-974 (2011)
DOI: 10.1002/jae
950 M. EHRMANN, M. FRATZSCHER AND R. RIGOBON
2. RELATED LITERATURE
The literature on financial linkages has evolved along two separate strands in recent
of these strands has been focusing on the domestic transmission of asset price sho
determinants. Another direction of the literature has been to analyze international linkage
the focus, however, has been mostly on individual asset prices in isolation.
Linkages across domestic financial markets are increasingly well understood. Earlie
the spillovers across different domestic asset prices often finds a positive correlation betw
returns and bond yields, such as Shiller and Beltratti (1992) and to some extent Barsky (
Campbell and Ammer (1993) for the USA, though the analysis of those studies is m
on low-frequency data. More recently, there has been evidence for negative stock-b
correlations (e.g., Baele et al., 2007). Looking at the relation between short-term int
and stock markets, it has been documented that equity prices react strongly to monetar
shocks in the USA (Bernanke and Kuttner, 2005; Ehrmann and Fratzscher, 2004) At
time, monetary policy has been shown to respond to equity markets (Rigobon and S
In an analysis of bond prices, short-term interest rates and equity markets, Rigobo
(2003b) find that the causality of the transmission process may run in several directions
instance, the correlation between US short-term interest rates and equity prices may ch
positive to negative depending on which of the asset prices is dominant in particular per
A closely related literature focuses on explaining the price discovery process in domest
prices through economic fundamentals. Several papers concentrate thereby on the impor
announcements and news of selected macroeconomic variables. Fleming and Remolo
1999), Balduzzi et al. (2001), and Bollerslev et al. (2000) show that macroeconomic ne
USA is an important driving force behind US bond markets. Fleming and Remolona (1999
hump-shaped effect of macroeconomic news along the yield curve in that the largest effe
news usually occurs at intermediate maturities. For equity markets, Flannery and Protop
(2002) and Boyd et al. (2005) also reveal a strong response of US equity markets to macro
news, while the latter paper as well as David and Veronesi (2008) show that the r
between economic fundamentals and equity returns may in some cases be dependent on e
conditions or the type of news.
There have also been various attempts to analyze international spillovers, though the f
this literature has so far concentrated only on individual asset prices in isolation, mostly
markets. For instance, the empirical work by Hamao et al. (1990), King et al. (199
et al. (1994), based on reduced-form GARCH models, detects some spillovers from t
the Japanese and UK equity markets, both for returns and in particular for conditional
Also Becker et al. (1995) find spillovers between the US and UK stock markets and
this is in part due to US news and information, although more recent work by Co
Wang (2003) argues that such macroeconomic news can explain only a small share of
market spillovers between mature economies. Bekaert et al. (2009) document return
across 23 countries, with no evidence for an upward trend in return correlations overall
upward trend for the European stock markets. Diebold and Yilmaz (2009) develop
index based on VAR models, and show that the evolution of return and volatility spillov
19 stock markets is strikingly different. For foreign exchange markets, the semina
Engle et al. (1990) and Andersen and Bollerslev (1998) find strong spillovers in foreig
markets, both in conditional first and second moments. Finally, Dungey and Martin (200
DOT: 10.1002/jae
contagion across different countries and financial markets, albeit analyzing mainly the transmission
of volatility across markets.
Another strand on international financial co-movements attempts to explain the evolution of
financial spillovers through real and financial linkages of the underlying economies. Heston and
Rouwenhorst (1994) and Griffin and Karolyi (1998) argue that mainly country-specific shocks, and
to a lesser extent industry-specific and global shocks, can explain international equity returns. In
addition, several papers emphasize the importance of linkages through trade and capital flows for
explaining financial market spillovers. One direction of the literature has been to focus on contagion
in international markets, marked by the work by Bae et al. (2003) and Forbes and Rigobon (2002).
Hartmann et al. (2004) show that exchange rate linkages strengthen during financial crises for a
broad set of emerging markets. Focusing on mature economies, Forbes and Chinn (2004) find that
the country-specific factors have become somewhat less important and bilateral trade and financial
linkages are nowadays more important factors for explaining international spillovers across equity
and bond markets.
A related literature focuses on the effects of macroeconomic announcements on various asset
prices. Andersen et al. (2003) show that, in particular, US macroeconomic news has a significant
effect on the US dollar-euro exchange rate. For bond markets Goldberg and Leonard (2003) and
Ehrmann and Fratzscher (2005) find that not only are macroeconomic news an important driving
force behind changes in bond yields, but also that there are significant international bond market
linkages between the USA and the euro area. The results of Ehrmann and Fratzscher (2005)
indicate that spillovers are stronger from the US to the euro area market, but that spillovers in
the opposite direction are present since the introduction of the euro in 1999. Finally, Andersen
et al. (2007) and Faust et al. (2005) look at the effect of macro announcements on high-frequency
asset returns across several asset prices, such as exchange rates and the yield curve, confirming
the importance of news and in some cases finding a significant response of risk premia or an
overshooting of exchange rates in the short run.
The paper by Andersen et al. (2007) is probably the closest to ours. As we will do in our paper,
they first control for the response asset prices to macroeconomic news, and subsequently measure
cross-market linkages using the heteroskedasticity of asset prices to identify their system. There
are important differences, though. First, Andersen et al. use high-frequency data, which obviously
limits data availability. Their sample covers a relatively short period, ranging from 1998 to 2002,
whereas we are able to cover 20 years using the lower-frequency data. Second, although Andersen
et al. cover US, German and British stock, bond and foreign exchange markets, this part of their
analysis estimates subsystems rather than the entire model. They study first the contemporaneous
relationship between US bond and stock markets and the euro-dollar exchange rate, and secondly
in a separate analysis the spillovers between US, UK and German stock markets. Hence, in line with
much of the literature, the distinction between domestic transmission across markets on the one
hand, and international transmission within individual asset markets on the other, is kept. Related
to their approach, the objective of the present paper is to provide a framework for analyzing the
interaction of the domestic and international transmission of financial market shocks. However,
we go a step further by testing for and showing that an important element of the transmission
occurs indirectly across market and across countries, e.g., that shocks to US short-term interest
rates are transmitted to not only to money markets in the euro area but also to equity and bond
markets in the euro area. In doing so, it is possible to link the distinct strands of the literature and
to gain a better understanding of the underlying nature of the transmission channels of financial
shocks.
DOI: 10.1002/jae
3. ESTIMATION METHODOLOGY AND DATA
This section explains our estimation methodology and the data we use for our empirical anal
3.1. Identification through Heteroskedasticity
Starting from the observation that asset prices are determined simultaneously, we ass
behavioral model of the following structural form:
A y, = & + n (L) yt-\ + V (L) zt + Mr (1)
where yt is a vector y, = (r,us, b,1JS, .vrus, rtEA, b,EA, s,EA, e,) of the seven endogenous asse
namely short-term interest rates (rf), long-term bond yields (b,) and stock market returns (s
each of the two economies, and the exchange rate return (<?,). n(L) captures the lagged
of the endogenous variables yt and ^(L) the lagged and contemporaneous effects of a
exogenous variables and common shocks zt. We will return below to explaining in mor
how Zt is constructed and what it includes. The 7x7 matrix A is of main interest to u
off-diagonal elements capture the contemporaneous interactions across asset markets.
Because A is not diagonal the true coefficients cannot be recovered from the estimation u
the standard techniques. More precisely, the starting point for identification is to estimate
reduced-form-or factormodel of equation (1) via ordinary least squares (OLS):
y, =A-li} + A~ln (L) yr_! +A~lV (L) z,+A~^t

yt = Cq + Bq (L) yt-\ + B\ (L) Zt + t (2)
where the reduced-form residuals st are related to the structural residuals fi, exactly t
matrix A.
Two commonly employed approaches to identification are Cholesky decompositions an
restrictions. However, the imposition of 21 zero restrictions in a model with financial
variables is simply implausible, and sign restrictions lead to an extremely large adm
parameter space. In this paper, therefore, we use an alternative methodology for identificat
known as identification through heteroskedasticity (IH). This methodology exploits the fact
financial variables are generally heteroskedastic to solve the problem of identification.
theoretical derivation of the methodology is well known and has been provided in Rigobon (
we only discuss its implications in this subsection and provide a technical appendix illus
the identification problem at the end of the paper.
By imposing two assumptions, the existence of heteroskedastic regimes solves the identifi
problem at hand.3 First, we assume that the structural shocks are uncorrelated, a co
assumption in the literature. Second, we assume that matrix A is stable across time (wh
will call regimes), which is the usual assumption imposed on models of the ARCH or G
type. This means that each additional heteroskedastic regime adds more equations than unkn
As a matter of fact, the identification problem is solved in the presence of two regimes. Fi
3 The precise form of heteroskedasticity is of no particular interest to us. It could be described as a GARCH model
(Rigobon and Sack, 2003b) or a regime-switching model. For convenience, we therefore assume the latter. As shown in
Rigobon (2003), the estimates of the coefficients are consistent, regardless of how the heteroskedasticity is modeled.
DOI: 10.1002/jae
it should be obvious that heteroskedasticity can be combined with other assumptions

exclusion restrictions), in which case the system of equations will become overidentified. T
the point discussed in the following subsection.
3.2. Overidentifying Restrictions
Identification through heteroskedasticity is a powerful method to identify systems of equ

a situation where it is difficult to impose credible exclusion restrictionsas is the cas
paper. Although the system is identified by the heteroskedasticity in the data, this is only
up to a rotation of matrix A.4 We therefore need to impose some additional restrictions to
that we pick the 'correct' rotation, which represents the underlying economic relationship
restrictions can be either exclusion restrictions or, most notably, sign restrictions.
In order to impose sensible restrictions, we start by discussing the meaning of each
equations in the system. For the purpose of illustration, we can write matrix A of the structu
model as follows:
/ 1 0l\2 ai3 /$i4 ^15 yn\

21 1 <*23 Pu P25 Pl6 YXl
<*31 <*32 1 ^34 @35 P36 737
A = @41 Pa2 Pa3 1 <*45 <*46 747

@5\ p52 P53 <*54 1 <*56 V51
P(>\ P62 P63 <*64 65 1 Y67
\ e ! \ Yi\ yi2 Yi3 Yia Yi5 Yi6 1 >

so that the a parameters indicate the spillovers across domestic asset prices within the USA and
within the euro area, the fl parameters the international spillovers and y are the spillovers from
and to the US dollar-euro exchange rate.
Turning to the interpretation of the equations, the equations for the short-term interest rate can
essentially be interpreted as the market's expectations about the course of monetary policy in the
short to medium term. The equation of long-term interest rates may be understood as reflecting
inflation expectations. Hence a fall at the long end of the yield curve may at least in part indicate
that markets anticipate lower inflation rates, conditional on the current short rate. The stock market
equation may be interpreted as a proxy of domestic demand in that a positive demand shock at
home raises domestic equity prices. Alternatively, changes in equity prices may also be explained
by supply shocks, such as productivity changes. Finally, the exchange rate movements may be
understood as reflecting changes in the relative demand across the two economies (see Pavlova and
4 A rotation is the multiplication of the matrix A by another matrix B that is full rank and has determinant equal
to one. This issue can be clarified best with the simple demand and supply problem, where D = aP + e\ (supply)
and P = bD + e2 (demand). In this problem there are always two matrices that produce the same pattern of second
momentsthe truthful one and the symmetric one: D = 1 /b P + 63 and P = \/a D + e^. Both systems of equations
produce the exact same covariance matrices and are therefore indistinguishable empirically. However, the two systems
have very different implications on the slopes of the curves. How can we differentiate these two systems? It is easy to
show that if we impose that demand equations are decreasing, or supply increasing, then there is a unique rotation which
is the truthful one. In other words, if the coefficient on D in the equation that has P on the left is forced to be negative,
then the procedure chooses the first rotation. The sign restrictions we discuss below have that objective.
DOI: 10.1002/jae
Rigobon, 2007). Of course, these interpretations are not clear-cut and may not exclude alternative
interpretations and explanations. When discussing the empirical results, we will go into more detail
about the interpretation of each of the equations and possible caveats.
We impose a first set of identification restrictions on domestic spillovers, as we can use existing
priors about their signs from the literature. Most restrictions are actually imposed on monetary
policy, as this is probably the best understood subsystem in our model. Note that, since matrix A
pre-multiplies the vector of endogenous variables on the left-hand side of equation (1), the sign
of the restriction is opposite to the expected reaction of asset prices. The assumptions are the
following:
1. We would expect that an inflationary shock should trigger market expectations of a monetary
tightening and thus a rise in short-term rates (due to the opposite sign we need to impose on A,
this implies ol\i, < 0).
2. As to the effects of monetary policy, an increase in short-term interest rates raises the discount
value and lowers the demand for goods and services and hence should lead to a decline in
equity prices (Rigobon and Sack, 2004; (*31, a(,4 > 0).
3. Moreover, also a rise in long-term interest rates should lower equity prices (<232, <*65 > 0).
Since we believe that these lines of reasoning should apply both to the direct effects of shocks on
asset prices (as measured by matrix A) as well as the overall effects, including indirect spillovers
(as measured by A-1), we impose the equivalent set of restrictions on A-1.
Turning to the international linkages, our theoretical priors for some of the spillovers are fairly
clear-cut but are less so for others.
4. Domestic and foreign money markets and bond markets should exhibit positive spillovers (ji 14,
/?4i < 0; ^25> /3s2 < 0). This has indeed been found to hold empirically between the USA and
the euro area in Ehrmann and Fratzscher (2005), based on a reduced-form GARCH type of
model, as the openness of financial markets and arbitrage may mean that interest rate shocks
are transmitted across economies.
5. We express all variables in basis points (bp), and therefore can impose the restrictions that
the international spillovers within marketswithin equity markets, within money markets and
bond marketsare less than one. This assumption boils down to assuming that a domes
tic shock should not have an amplified and more than proportional effect on foreign mar
kets (0 < I/614I, |>04i|, I/J25I, \f>52\- 1/^36!> |/tal < ') This assumption is reasonable for developed
economies, whereas it may be incorrect for emerging markets.
These restrictions are imposed on the structural coefficients. As indicated above, significant
empirical evidence on closed-economy monetary models supports the assumptions we are making.
Furthermore, we find in the empirical results that most of these restrictions are not binding, but
they help us further in the process of identification.5
The next issue relates to the international cross-market spillovers. Recall that the parameters in
the structural-form or behavioral model should be interpreted as indicating only the direct linkages
5 The reason is that the sign restrictions limit the space in which parameters have to be searched to minimize the
moment restrictions. This influences the speed of convergence but does not affect precision unless the estimates are on
the boundaries. As we will see later, very few of the coefficients are on the boundaries, suggesting that only a small set
of the restrictions are binding.
DOI: 10.1002/jae
between markets, whereas the parameters of the reduced-form model capture

as indirect linkages across asset prices. By indirect linkages we mean spill
occur via other asset prices. For international cross-market spillovers it is
instance, a rise in short-term interest rates in the USA should have a direct im
equity prices Of course, a rise in US interest rates is likely to affect al
prices, but this effect should be an indirect one in the sense that it is transm
asset prices such as euro area interest rates. In this case, a rise in US inter
increase in euro area rates, which then in turn raises the discount factor
causes a drop in euro area equity prices.
Hence, in addition to the overall sign, we also impose zero restrictions on all
market spillovers in the structural-form model. This assumes that the cross-m
spillovers are zero, but remember that we still allow for indirect spillovers in
model indicated by matrix A-1. These exclusion restrictions have been use
contagion literature (see Dungey and Martin, 2007, for one of the multiple ex
Finally, we restrict some y parameters for the spillovers from and to
exchange rate. We presume that an increase in long rates in the USA leads to a
US assets, leading to an appreciation of the dollar and vice versa (772 > 0,
Overall, our benchmark identification of matrix A looks as follows:
/ 1 a\2 < 0 <*13 1 < pu < 0 0 0 K17\

<*21 'l <*23 0 -1 <p25 <0 0 Y21
<231 > 0 (*32 > 0 1 0 0 -1 < /S36 < 1 Yn
A = 1 < /S41 < 0 0 0 1 <*45 < 0 <*46 K47
0 1 < p52 < 0 0 <*54 1 <*56 Y51
0 0 1 < pa < 1 CU54 > 0 <*65 > 0 1 Y61
V Yn yn > 0 V73 K74 V75 < 0 Yl(y 1/
3.3. Estimation Procedure
The estimation procedure is as follows:
1. Given that equation (2) can be estimated by a simple VAR and the residuals share the exact
same contemporaneous relationship as the one we are interested in, we run a VAR with six lags,
which is sufficient to remove any serial correlation and the constant terms. From this step, we
recover the reduced form residuals, which contain only the contemporaneous effects.
2. From the reduced-form residuals we define the heteroskedastic regimes. We do so by computing
rolling windows variances of 20 2-day observations for each variable. A regime is identified by
having at least 16 observations for which the relative variances of one or several asset returns
exceed their average value plus one standard deviation.
3. We can identify 28 regimes in total, but restrict the analysis to the seven 'single-market' regimes,
i.e., those where one of the seven asset prices exhibits an elevated conditional volatility, whereas
all others remain in the 'tranquil' state and the regime where all are in the 'tranquil' zone.
4. We estimate the parameters of interest by minimizing the following minimum distance
min g'g with g = A'Y,jA Q,
s.t. E,- is diagonal, A restrictions
DOI: 10.1002/jae
where is the variance of the structural shocks we are interested in, and 2 is the variance
covariance matrix we estimated in each regime i. This is analogous to a GMM estimator, the
distribution of which is easily derived. Identification requires the equations to be linearly indepen
dent, which is assured by the fact that the volatility of one of the observed variables is elevated,
while the others are relatively stable. Because the model has overidentifying restrictionswe only
need two regimes and we have 7 + 1 we can test the overidentifying restrictions (OIR) in the
usual way.
5. We block-bootstrap the significance of our parameter estimates. For each of the heteroskedas
ticity regimes, we use the estimated regime-specific covariance matrices to create new data with
the same covariance structure in each bootstrap replication. For each draw, we estimate the
coefficients by minimizing the moments given the restrictions.
6. As our model is fully identified, variance decompositions can be estimated, either including only
endogenous variables and shocks, or also lags and exogenous variables. These are calculated as
var(yf) = [/ -A-In (L)]var(/z,) {[/ - A"'n (L)]r}-1and

var(_y,) = [/ -A~lTl (!)][ (L)varfe)vf>r (I) + var(/xf)][/ - A"1^)]7",
respectively. As both variance decompositions are similar, we will present results from the second.6
3.4. Data
The empirical analysis focuses on financial linkages between the US and the euro area m
markets, bonds markets, equity markets and foreign exchange markets in the period 1989-200
The data source for all financial market series is Datastream. For the USA, we include
month Treasury bill rate for the short rate, the 10-year Treasury bond rate for the long rate
the S&P 500 index for the stock market. For the euro area, we use the 3-month interbank rat
FIBOR rate before 1999 and the EURIBOR after 1999for the short rate, the German 10-
government bond for the long rate, and the S&P Euro index for the equity market. The excha
rate included is the US dollar-Deutsche Mark before 1999 and the US dollar-euro since 199
Each of the seven variables is expressed in terms of percentage returns, i.e., the level of short
long interest rates, and the log first difference of equity prices and the exchange rate.8 All o
series exhibit the typical characteristics of heteroskedasticity, skewness and excess kurtosis.
A further important issue is that of the data frequency and timing. Trading in the Europea
markets takes place earlier than in the USA, which implies that shocks emanating from
European markets are always incorporated into US asset prices on the same day. By cont
since there is only a limited overlap in trading times between the US and the euro area marke
(especially for the short rates, as the closing quotes for the German and euro area marke
determined at 11:00 Central European Time), some of the US shocks only affect European
6 If we were using instrumental variables to estimate only some of the coefficients, a variance decomposition cou
have been derived. This is one of the advantages of the methodology presented here.
7 The US dollar-Deutsche Mark exchange rate prior to 1999 is multiplied by the Deutsche Mark-euro conversion r
8 An alternative approach, namely to model the changes of interest rates rather than their level, has been pursued
working paper version (Ehrmann et al., 2005). The main findings (spillovers from the USA to the euro area being
than in the other direction; magnification of responses when allowing for indirect effects, and substantial cross-m
linkages) are qualitatively replicated also there.
DOI: 10.1002/jae
prices on the subsequent business day. To reduce this problem of only partial overlap of tradin
times, we change the frequency of the analysis and use 2-day returns for all of the asset retu
series.9
Finally, recall that one of the central conditions to achieve identification is that the structural
form shocks are orthogonal. This condition may not be fulfilled if asset price shocks are driven
by common shocks. Accordingly, we model a number of variables that control for common
factors. These variables are denoted by the vector zt in equation (1). First, we include a set of
macroeconomic news in the USA and the euro area. We use the unexpected or news component of
these variables, measured as the difference between the actual announcement and its expectations
(stemming from survey data conducted by Money Market Services (MMS) International).10
While there is ample evidence for financial market reactions to the news component of US
announcements, it has often been difficult to establish such a result for European announcements.
This could be due to a reduced information content of European announcements, which are often
less timely than their US counterparts, and in several instances provided as an aggregate of earlier
announcements.11 As this might imply that we capture only a fraction of the common shocks,
we furthermore include oil price changes and allow for one common factor in the structural-form
model (1). A further extension to two common factors (not shown for brevity) does not affect our
results, such that we are confident that one common factor is sufficient for our purposes.
One issue deserves mentioning at this stage. The decision to model the two largest economies,
and to trace the feedback mechanisms between seven of their financial markets, constitutes a
substantial enlargement of the models that have to date been typically estimated in the related
literature. Nonetheless, it still neglects other markets that might be of relevance. In particular,
the influence of Asian markets is not modeled in our framework, such that we might in part
misattribute the various transmission channels. While clearly a limitation of the current model,
we are facing issues of tractability that prevent us from further enlarging the estimated system.
However, we will conduct a robustness test in that direction, whereby we eliminate all days during
major international crises from the sample, as for these days we have a prior that shocks might
have been originating in other economies than the USA or the euro area.
4. RESULTS
We now turn to presenting our empirical results, starting with the estimates for the
indirect international transmission channels (Section 4.1), and specifically the ex
9 This cannot eliminate the problem entirely, but it reduces its importance, as the relative share of the
time periods is smaller in a 2-day window. As we will show below, the results are robust to using low
such as weekly data.
10 This approach follows Andersen et al. (2007). The announcements include, for the USA: the NAP
of purchasing managers and consumer confidence; non-farm payroll employment and unemployment f
workweek, GDP, and industrial production; retail sales, trade balance and housing start figures; as well
releases. For the euro area, our set of news includes those for the euro area since 1999 as well as for
back to the early 1990s: Ifo business climate, business and consumer confidence indices; GDP, industrial
manufacturing orders; unemployment, trade balance figures; and M3, PPI and CPI. To match the frequency
returns, we construct the sum of their surprise components over the 2 days.
11 This is in particularly the case for euro area announcements, which are aggregated from the nati
themselves are often announced at earlier dates. However, even for national data this might be the case
German inflation announcements are preceded by announcements by each German state's inflation figur
and Fratzscher, 2005).
DOI: 10.1002/jae
(Section 4.2), before outlining the findings for the variance decomposition (Section 4.3) and
concluding with some robustness tests (Section 4.4).
4.1. International transmission
The primary focus of this paper is on the international transmission mechanism, such that w
concentrate on the discussion of the corresponding results in this section.12 We highlight parameter
that are significant at the 90% level using bold font below. All the benchmark results are shown
in Tables I and II as well as in Figures 1 and 2, which synthesize the results of 500 bootstrap
replications. The significance is tested through the share of parameter values in the distributions
depicted in Figures 1 and 2 that are beyond zero, or the share of replications in which the paramete
restrictions are binding.
4.1.1 Direct Effects

As discussed above, we restrict all parameters that relate to international spillovers across differe
markets to zero in the structural-form model, such that we will only report those parameters th
show international spillover effects across the same markets, as well as those for the exchange rate.
US asset returns respond to their euro area counterparts and the exchange rate in the following
way:
r,us = 0.0008 r,EA - 0.0038 e, + ... (3)
b,vs = 0.4402 b,EA + 0.0678 e, + ... (4)
s,vs = 0.0594 srEA - 0.1324 <?, + ... (5)
whereas the spillovers from the USA to the euro area look as follows:
r,EA = 0.1704 r,us + 0.0043 e, + ... (6)
b,EA = 0.4892 btvs - 0.0596 e, + ... (7)
s,EA = 0.6269 s,vs + 0.7071 e, +... (8)
Although we restrict all the parameters for the international spillovers across the same markets
to be between zero and one, none of the restrictions is actually binding. The spillovers from the
USA to the euro area are substantially larger and generally estimated at much higher levels o
statistical significance than in the other direction. The most extreme difference in this respect
is found for the stock markets, with a spillover of 0.63 from the USA to the euro area, and
statistically insignificant effect of euro area markets on the USA. These estimates seem plausible,
given the generally observed leading role of the US market. Whereas we had imposed the sign on
the spillovers across short- and long-term interest rates, no such restriction had been used for th
case of stock markets. In line with the literature on contagion, we find that these spillovers are
positive, suggesting that a positive shock to domestic equity prices induces a positive spillove
and leads to a rise in foreign equity markets as firms and demand are linked internationally.
12 The results for the domestic transmission are shown in all tables yet not discussed, for brevity.
DOl: 10.1002/jae
Table I. Parameter estimates and bootstrap results of structural form model
Point estimate Bootstrap
Mean SD p-value
Domestic transmission
USA
a 12 -0.2850*** -0.2988 0.0474 0.0000

<*13 -0.0011 -0.0009 0.0030 0.3533
<*21 -0.3245*** -0.3276 0.0507 0.0000
<*23 -0.0276*** -0.0278 0.0049 0.0000
<*31 0.0000 0.0078 0.0339 1.2056
<*32 2.9113*** 2.9232 0.1588 0.0000
Euro area
<*45 -0.0761*** -0.0802 0.0379 0.0040

<*46 0.0029 0.0018 0.0016 0.1497
<*54 -0.0548** -0.0672 0.0398 0.0140
<*56 0.0034 0.0026 0.0032 0.1996
<*64 0.0994** 0.1122 0.0630 0.0279
<*65 0.2680** 0.2637 0.0987 0.0120
International transmission
USA to euro area

-0.1704*** -0.1284 0.0458 0.0080
P52 -0.4892*** -0.4668 0.0429 0.0000
063 -0.6269*** -0.6376 0.0716 0.0000
Euro area to USA
/14 -0.0008 -0.0078 0.0157 0.1756

052 -0.4402*** -0.4213 0.0595 0.0020
063 -0.0594 -0.0397 0.0860 0.2515
Exchange rate effects

Yll 0.0038 0.0054 0.0044 0.1058
Y21 -0.0678*** -0.0659 0.0075 0.0000
Y31 0.1324 0.1303 0.1053 0.1018

K47 -0.0043* -0.0038 0.0030 0.0978
K57 0.0596*** 0.0596 0.0047 0.0000
Y61 -0.7071*** -0.6758 0.0634 0.0000

Y71 -0.0856* -0.0808 0.0755 0.0818
Yll 3.4602*** 3.4396 0.1046 0.0000

Y13 -0.0277 -0.0208 0.0519 0.3154
YlA 0.3800*** 0.3840 0.0952 0.0100
Yt 5 -4.2461*** -4.1854 0.1435 0.0000

0.0047 -0.0108 0.0391 0.6168
Y16
Note: The table reports the parameter estimates of model (1) obtained in the identification through heteroskedasticity.
Asterisks denote significance at the * 90%, ** 95% and *** 99% level. Significance is judged through the p-value obtained
in a bootstrap. The vector of variables is (rus, bvs, ,vus, rEA, bFA, sEA, e)'.
theoretical justification of this effect is provided in Zapatero (1995), Cass and Pavlova (2004) and
Pavlova and Rigobon (2007).
As to the effects of the exchange rate, statistical significance is much more pronounced for the
effects on the European markets than for the USA, where only bond markets are affected in a
significant fashion. We find that a depreciation of the US dollar leads to an increase in US bond
yields and a reduction in euro area bond yields, which is consistent with our interpretation of
DOI: 10.1002/jae
M. EHRMANN, M. FRATZSCHER AND R. RIGOBON
Table II. Parameter estimates and bootstrap results of reduced-form model

Mean SD p-value
USA
oil 1.1163*** 1.1291 0.0550 0.0000
021 0.4065*** 0.4110 0.0965 0.0000
031 -1.1964*** -1.1914 0.3154 0.0000
an 0.3522*** 0.3748 0.0978 0.0000
022 1.2305*** 1.2213 0.1145 0.0000
032 -3.6041*** -3.4947 0.5180 0.0000
<213 0.0111*** 0.0115 0.0027 0.0000
<323 0.0351*** 0.0348 0.0058 0.0000
033 0.9347*** 0.9166 0.0397 0.0000
Euro area
044 1.0074*** 1.0095 0.0081 0.0000

<*54 0.0768*** 0.0915 0.0404 0.0100
64 -0.2653*** -0.2844 0.0845 0.0020
045 0.1335*** 0.1286 0.0486 0.0000
055 1.1943*** 1.1709 0.0687 0.0000
065 0.0003 -0.0512 0.1367 0.7685
046 -0.0033 -0.0021 0.0018 0.1417
056 -0.0031 -0.0028 0.0040 0.2515
066 1.0322*** 1.0219 0.0472 0.0000
[ASA (o euro area
&41 0.2099*** 0.1654 0.0572 0.0000

bs\ 0.2384*** 0.2334 0.0609 0.0000
b6\ -1.1223*** -1.1295 0.2764 0.0000
b42 0.1176*** 0.1046 0.0377 0.0040
bs2 0.7019*** 0.6770 0.0940 0.0000
b62 -3.4318*** -3.3895 0.3424 0.0000
^43 0.0014** 0.0017 0.0010 0.0439
b53 0.0169*** 0.0165 0.0039 0.0000
^63 0.5629*** 0.5603 0.0526 0.0000
Euro area to USA
b\4 0.0084*** 0.0193 0.0229 0.0040

>24 0.0249** 0.0340 0.0244 0.0399
b}4 -0.0694* -0.0973 0.0987 0.0559
b\5 0.1859*** 0.1868 0.0563 0.0000
b25 0.6949*** 0.6641 0.1002 0.0000
hs -2.3628*** -2.2651 0.2747 0.0000
b 16 0.0000 0.0000 0.0013 0.5269
b26 -0.0005 0.0002 0.0040 0.5309
&36 0.0645 0.0433 0.0813 0.2615

Ol -0.4065*** -0.4350 0.1238 0.0040
C72 -1.3753*** -1.4673 0.2153 0.0000
C73 -0.0260 -0.0247 0.0375 0.2275
C74 -0.1429* -0.1259 0.1206 0.0599
C75 2.5665*** 2.5266 0.1495 0.0000
C76 -0.0133 0.0009 0.0340 0.5110
C17 0.0071** 0.0057 0.0034 0.0299
DOI: 10.1002/jae
Table II. (Continued)

Mean SD p-value
Ctf 0.0356*** 0.0337 0.0057 0.0000

c37 -0.1775** -0.1791 0.0793 0.0200
C47 0.0010 0.0004 0.0019 0.4311
C57 -0.0286*** -0.0304 0.0038 0.0000
c67 0.4258*** 0.4024 0.0811 0.0000
C77 0.7488*** 0.7527 0.0190 0.0000
Note: The table reports the parameter estimates of model (2) obtained in the identification through heteroskedasticity.
Asterisks denote significance at the * 90%, ** 95% and *** 99% level. The significance is judged through the p-value
obtained in a bootstrap. The vector of variables is (rus, bus, iUS, rEA, bEK, sEA, e)'.
long-term interest rates as reflecting inflation expectations.13 Interestingly, the magnitude of these
effects is fairly similar for both economies. The most interesting effect is found for equity markets.
Whereas US equity markets do not respond to exchange rate movements, the euro area markets
rise by a substantial amount following an appreciation of the euro: a 10% appreciation of the euro
is estimated to induce a 7% rise in euro area equity prices. Such a finding can result, for instance,
from a return chasing motive in the spirit of Albuquerque et al. (2007), with the euro area being
financially more open than the USA, and accordingly more strongly affected by such a pattern.
4.1.2 Overall Effects

Focusing on the reduced-form model (2) allows us to understand and analyze the overall spillovers,
including both direct and indirect effects. The following equations show the estimates for the
contemporaneous spillovers from euro area assets and the exchange rate to the three US asset
returns:
r,us = 0.0084 /j** + 0.1859 ^ + 0.0000 ^ + 0.0071 /jeJ + ... (9)

b,vs = 0.0249 + 0.6949 - 0.0005 /if* + 0.0356 + ... (10)
.5;us = -0.0694 ^ - 2.3628 /if* + 0.0645 ^ - 0.1775 ne>t + ... (11)
and for the spillovers from the USA to the euro area, the results are as follows:
r,EA = 0.2099 + 0.1176 + 0.0014 + 0.0010 ne,t + ... (12)

btEA = 0.2384 /tjjf + 0.7019 /tj + 0.0169 /j, - 0.0286 ne,, + ... (13)
stEA = -1.1223 - 3.4318 + 0.5629 /x + 0.4258 ne,, + ... (14)
13 The exchange rate is defined as US dollar in units of euro; i.e., an appreciation of the US dollar is a fall in the exchange
rate.
DOI: 10.1002/jae
Figure1.Distrbuonfstruc aloeficntsofhebncmark odelin50botsrapelictons
Copyright201JohnWiley&Sons,Ltd. J.AplEcon.26:948-7(201)
DOI:10. 2/jae
<
9 5 S 5 ? 8 ? 9 88&8838851 588S88S8
9 2 s g 8 8 ? 8 e
-<
i 3 2 5 8 '
<
5SS&8SS8800 258838 25883! 9 258888 s s E
888 " | 2 2 5 8 8 5 8
~<
<
< 3
2S2S5S85S0 = S?2SS88S8 e2S588S8c
Figure2.Distrbuonfreduc-formcefintsofhebncmark odelin50botsrapelictons
Copyright201JohnWiley&Sons,Ltd. J.AplEcon.26:948-7(201)
DOI:10. 2/jae
The key finding is that not only is the international transmission of shocks significant for the lar
majority of asset prices, but also there are substantial international cross-market linkages. This
underlines and confirms our argument that a more complete understanding of financial linkages
requires the modeling of international cross-market financial linkages, which so far has be
missing in the literature. The importance of the international cross-market transmission manife
itself not only through the significance of the point estimates of the cross-market coefficients
in equations (9)-(14), but also through the changes in the coefficients of the within-marke
coefficientsi.e., the international spillovers within equity markets, money markets and bo
marketswhen comparing the results of the structural-form model (3)-(8) and those of th
reduced-form model (9)-(14).
In general, the results show that US shocks are generally more influential for euro area marke
than euro area shocks for US markets. Moreover, spillovers are largest within the same asset clas
and we find that the estimates have the expected sign and magnitude.
For equations (9) and (12) for short-term interest rates, one would expect that a positive shock
to short rates, long rates or equity prices in the foreign markets should raise short-term interes
rates at home, which we find to be true, although we find that shocks to the equity markets are
not particularly relevant for short rates.
A more interesting finding is present in equations (10) and (13) for US and euro area bon
yields. The estimates suggest that there are significant spillover effects across bond markets of
these two economies, exceeding by far those of the money markets. This points to a large degree o
co-movement of bond rates due to international portfolio allocation. Interestingly, the magnitud
of spillovers is estimated to be very similar regardless of the direction. An interesting differenc
compared to the results obtained in the structural form relates to the effects of short rates on bon
yields. Despite being restricted to zero in the structural form, the coefficients are highly significa
in the reduced form, with a sizable effect being exerted from US short rates on euro area bond
yields.
Coming to the results for the equity market equations in (11) and (14), we find large spillovers
from all asset prices in the USA to the euro area. The effects from equity markets to equity
markets mirror those found for the structural form. On average, a 1% shock to US equities leads
to a corresponding adjustment of euro area equity prices of 0.56%. By contrast, the spillover from
euro area to US stock markets is very small and not statistically significant. This confirms the
centra] role that US equity markets play in world stock markets.
Moreover, euro area equities are found to respond significantly to shocks in short-term rates and
bond yields in the USA. In fact, a 100 bp rise in US short rates is estimated to lower euro area
equity prices by more than 1%, and a 100 bp rise in US long rates leads to a drop of more than
3% in euro area stock markets. Again, these are the total effects of US rates on euro area equities,
suggesting that the channels of transmission can be manifold. First, a rise in US interest rates is
likely to induce a similar movement in euro area short rates and bond yields. Second, there may
be a direct effect in that a rise in US interest rates leads to higher borrowing costs for many euro
area firms, in particular those that are active internationally. And third, the effect of US interest
rate changes may be transmitted to European markets via the exchange rate (we will analyze in
the subsequent section to what extent the exchange rate responds to the various asset prices and
shocks).
By contrast, euro area money and bond markets exhibit a smaller influence on US equity
markets, although the effects are mostly statistically significant. Both of the signs of the estimates
are correct, indicating that higher short-term and long-term interest rates in the euro area lower
DOI: 10.1002/jae
US equity prices. In particular, shocks to euro area bond and money markets have a signif
effect on all three US markets.
4.2. Response of the Exchange Rate
4.2.1 Direct Effects

We will now turn to the effect of asset prices on the exchange rate. The estimates are
ûsd/eur _ Q 0856 r us _ 3 4602 . brus + 0.0277 .v,us
- 0.3800 r,EA + 4.2461 brEA - 0.0047 s,EA + ... (15)
Although we had imposed two restrictions on the exchange rate equation, none of them is actually
binding. The largest estimates are found for the effect of bond yields on the exchange rate: a
100 bp rise in US bond yields leads to a 3.5% appreciation of the US dollar, whereas a 100 bp
increase in euro area long-term interest rates induces a euro appreciation of 4.2%. The effects of
short-term rates are substantially smaller, and have the opposite sign, suggesting that an increase
in short rates depreciates the respective currency.
4.2.2 Overall Effects

While this latter finding might be somewhat surprising, it is interesting to note that once all
transmission channels are taken into account, the US dollar does appreciate in response to an
increase in US short rates. Also, the magnitude of the overall effect is substantially larger than
for the direct effect. For the euro area, the pattern remains unchanged, although the indirect
channels tend to attenuate the euro depreciation in response to higher short-term rates, given that
the parameter estimates changes from 0.38 to 0.14. At the same time, the overall exchange
rate reaction to bond yield changes is smaller than the direct effect, while it remains significant,
large and with the predicted sign.
ûsd/eur = _Q 4Q65 . Mus _ j 3753 . ûs _ o 0260 . ûs
- 0.1429 nEA + 2.5665 ^ - 0.0133 nEA + ... (16)
4.3. Variance Decomposition
Having identified and analyzed the domestic and international transmission of shocks, we now
turn to assessing the relative, overall importance of each of the financial markets in the system. In
particular, how much of developments in domestic financial markets are explained by shocks in
foreign markets and how much is due to domestic factors? Moreover, what is the role of common
shocks and the exchange rate?
In order to answer these questions in this section, Table III shows the variance decomposition
for the reduced-form model (2) over the whole sample period 1989-2008. Each cell indicates the
share of the total variance of each financial market that is explained by the respective shocks to the
seven asset prices fiitt as well as the common shock iic l. The numbers in italics are standard errors
to gauge the statistical significance of the estimated parameters. As expected, and as is usually
found, by far the largest share of the respective variances is explained by the own idiosyncratic
shocks, ranging between 55% and 88%.
DOI: 10.1002/jae
Table III. Variance decomposition of benchmark model
r,m b,vs r us
>( brEA
. EA
>/ e,
/. us
Mr,t 88.39%*** 8.57%*** 2.02%* 13.70%*** 444%*** 1.33%** 0.61%
5.63% 2.59% 1.06% 4.46% 1.65% 0.66% 0.42%
Mr US
Mb,t 8.04%** 71.69%*** 16.76%*** 3.93%** 35.13%*** 11.38%*** 6.38%'
3.17% 2.44% 3.86% 1.91% 2.91% 1.70% 2.44%
// US
Hs,t 0.53%** 3.88%*** 75.08%*** 0.04% 1.36%*** 20.40%*** 0.15%
0.25% 1.36% 4.67% 0.07% 0.52% 5.76% 0.63%
EA
Mr,/ 0.00% 0.01% 0.00% 79.40%*** 0.12% 0.02% 0.02%
0.04% 0.04% 0.02% 6.84% 0.52% 0.02% 0.06%
ii, EA 1.20%** 12.29%*** 3.87%*** 2.72% 54.68%*** 0.00% 11.95%'
Mb,t
0.55% 1.88% 0.80% 2.55% 5.55% 0.06% 1.58%
// EA
r~S,t 0.00% 0.00% 0.29% 0.17% 0.04% 56.44%*** 0.03%
0.01% 0.06% 0.77% 0.14% 0.12% 5.88% 0.31%
lê,t 0.14% 2.54%** 1.72% 0.01% 2.47%*** 7.43%*** 80.25%'

0.12% 1.08% 1.73% 0.07% 0.86% 2.58% 2.09%
^c,t 1.70%*** 1.01%*** 0.25%** 0.04% 1.76%*** 3.00%*** 0.61%'

0.15% 0.12% 0.12% 0.04% 0.19% 0.25% 0.18%
Note: The table reports the share of the variance of each series that is explained by the various structural shocks. Numbers
in italics are standard errors. Asterisks denote significance at the * 90%, ** 95% and *** 99% level.
The key result is that a significant and relatively large share of the behavior of financial
markets is explained by foreign asset prices. US short rates are mainly determined nationally,
with euro area bond yields explaining 1.2% of their variance. At the same time, however,
12.3% of the variance of US bond yields and even 3.9% of the variance of equity prices go
back to euro area bond yields. Taking an average of all euro area effects on US financial
markets, we find these to be responsible for about 6% of the variance. By contrast, a much
larger share of euro area financial market movements are driven by US financial markets: on
average, around 30% of the variances of euro area financial markets are explained by US financial
market shocks. The spillovers to bond yields and equity markets are particularly strong: shocks
to US long-term rates account for 35% of the variance of euro area long rates, and 32% of
the variance of euro area equity markets is explained by US equity markets and US bond
yields.
These results stress that over the whole sample period 1989-2008 financial markets in both
regions were to a large extent driven by US shocks.
A final note concerns the role of exchange rate and common shocks, shown in the bottom two
rows of the table. Exchange rates have a similar importance for US and euro area bond yields,
accounting for around 2.5% of movements in each market. When it comes to equity markets, the
euro area is substantially more affected by the exchange rate (which explains 7.4% of its variance)
than the USA, where no significant effect is found. This is in line with the earlier discussion on
the exchange rate effects on euro area stock markets.
The common factor explains up to 3% of the variance of the different financial markets. These
estimates are small. It is, however, important to note that the contribution of the common factor
constitutes a lower bound of the actual importance of common shocks; the response of financial
markets to macroeconomic announcements and oil prices has already been accounted for in the
DOI: 10.1002/jae
VAR estimate of the first step, such that all results presented here are net of the effects of t
common shocks.
In summary, the key finding is that US and euro area financial markets are closely linked, not
only within asset classes but also across financial assets. On average, 30% of the variance of the
three euro area financial assets is accounted for by US developments, whereas a still sizeable albeit
smaller share of, on average, around 6% of US financial market movements are due to euro area
developments.
4.4. Robustness
In what follows, we will subject our estimates to a number of robustness analyses. These ca
clustered into three different groups. First, we test whether different approaches to defining the
area data have a relevant impact on the thrust of results. Second, we look into possible differen
depending on the position in the business cycle or the incidence of major international financia
crises. Third, we will test whether changes in a number of technical modeling assumptions affe
our results. Given the large number of robustness tests, we need to focus our discussion on som
of the estimated parameters. Table IV shows the parameter estimates of the structural form m
(1), which constitutes the basis of our estimation approach. In the discussion, we will once more
restrict ourselves to the international transmission parameters.
One possible concern about the interpretation of our results relates to our approach of model
the euro area pre 1999. We had used German interest rates, and the US dollar-Deutsche M
exchange rate to proxy for the euro area. Other approaches have been discussed in the literatur
e.g. by Anderson et al. (2007), Beyer and Juselius (2008) and Briiggemann et al. (2008). To
for robustness, we have constructed an alternative dataset prior to 1999: based on as many natio
data series as were available to us, we constructed their first principal component, with the int
to create a data series that captures the common movement of euro area countries.14
Results of re-estimating the model with these data are provided in the second column
Table IV; for ease of comparison, the earlier benchmark results are replicated in the first colum
There are mainly two changes that are of interest, both relating to the short-term interest rat
The effect of US short rates on those in the euro area construct is reduced, while those in
other direction are magnified. Interestingly, the parameters are now estimated at similar order
of magnitude. A larger effect on the USA can be rationalized, given that the short rates do now
represent a broader European coverage than before; a reduced effect from the USA seem
suggest that the link of German rates to those in the USA had indeed been somewhat stro
than for the euro area as a whole.
A second possibility to cleanly identify the euro area is to re-estimate the model over a shorter
sample period, when proper euro area data are available. The results of this exercise are reported
in column 3 of Table IV. Compared to the benchmark specification, there seems to have been
an increasing linkage of financial markets, with a number of parameter estimates increasing
substantially in magnitude. While this is the case for US as well as euro area parameters, the
relative differences remain, with the euro area being substantially more strongly affected by the
USA than vice versa.
14 In more detail, we had data on exchange rates for Austria, Belgium, Finland, France, Germany, Ireland, Italy, the
Netherlands, Portugal and Spain; on short-term interest rates for Belgium, Finland, France, Germany, Ireland, Italy,
Luxembourg and the Netherlands; on bond yields for Austria, France, Germany, Ireland and the Netherlands.
DOI: 10.1002/jae
Table IV. Robustness tests: parameter estimates of the structural form
Benchmark Common Shorter Excluding Excluding 5-day 10 2-day All variance

model factors sample post recessions financial returns rolling regimes
1999 crises window
USA
a 12 -0.2850*** -0.1869*** -0.2742*** -0.1693*** -0.2106*** -0.2483*** -0.2905*** -0.1239***
"13 -0.0011 -0.0067*** -0.0030 -0.0046** -0.0022 -0.0324** -0.0202*** -0.0025
<*21 -0.3245*** -0.2024*** -0.0571** -0.6081*** -0.5208*** 0.0556 -0.1112*** -0.2644***
<*23 -0.0276*** -0.0264*** -0.0071*** -0.0333*** -0.0217*** -0.0382 -0.0187*** 0.0062**
<*31 0.0000 0.0000 0.1229 0.0000 0.0138 0.1545 1.6813*** 0.1067**
<*32 2.9113*** 2.9947*** 0.2500** 3.5444*** 2.7315*** 0.4008*** 1.5857*** 0.2803***
Euro area
<*45 -0.0761*** -0.2656*** -0.1654*** -0.0579** -0.0675*** 0.0000 -0.0190*** 0.0000

<*46 0.0029 0.0035* 0.0023 0.0021 -0.0012 -0.0301*** 0.0014 -0.0010
<*54 -0.0548** -0.0674*** -0.0011 -0.3550*** -0.6981*** -0.0681** -0.3989*** -0.1388***
<*56 0.0034 0.0044*** 0.0008 0.0014 -0.0081 0.0076 -0.0001 -0.0004
<*64 0.0994** 0.0000 0.0000 0.0474 0.0619*** 1.2509*** 0.2712*** 0.7877***
<*65 0.2680** 0.0000 0.0000 0.0162 0.4613*** 0.0299 0.1549** 0.1780***
USA to euro area

Pa\ -0.1704*** -0.0317*** -0.1633* -0.0858*** -0.1193*** -0.1457*** -0.1427*** -0.2837***
P52 -0.4892*** -0.4766*** -0.7501*** -0.5051*** -0.3067*** -0.6335*** -0.4203*** -0.4727***
fe -0.6269*** -0.6392*** -0.7404*** -0.6865*** -0.6530*** -0.6819*** -0.5915*** -0.5266***
Euro area to USA
$14 -0.0008 -0.0216** -0.0128 -0.0455** -0.0444*** -0.1433** -0.0941*** -0.0046*

@25 -0.4402*** -0.4331*** -0.7748*** -0.4292*** -0.1757*** -0.2498*** -0.3323*** -0.4742***
-0.0594 -0.0563 -0.1421* 0.0051 -0.0341 -0.2650*** -0.1008 -0.1273*

Xl7 0.0038 0.0171*** 0.0237** -0.0018 -0.0140 0.0457 -0.0233*** -0.0107**
Y21 -0.0678*** -0.0448*** -0.0204* -0.0333*** -0.0762*** -0.2396*** 0.0176*** -0.0440***
Yil 0.1324 0.0960 0.2861* 0.0661 0.1916* 0.3433*** 0.1089 0.1708***
YV1 -0.0043* -0.0072** -0.0036 -0.0003 0.0223*** 0.0313** 0.0042 00121***
K57 0.0596*** 0.0003 0.0123** 0.0073** 0.0711*** 0.0195 0.0167*** 00312***
K67 -0.7071*** -0.7903*** -0.9644*** -0.7542*** -0.4679*** -0.6189*** -0.6316*** -0.6613***
K71 -0.0856* -0.6736*** -0.4365** -0.1574 0.1375*** -0.0211 1.0924*** 0.2631***
Y12 3.4602*** 2.1906*** 2.2345*** 1.3508*** 4.2644*** 1.1127*** 0.0000 2.4475***
yn -0.0277 -0.0474 -0.0989* -0.0520 0.0343 0.0152 0.0462 0.0323*
Yl\ 0.3800*** 0.0881 0.4226*** -0.8735*** -0.8407*** -0.7959*** -0.8013*** -1.3796***
Y75 -4.2461*** 0.0000 -0.1558** 0.0000 -6.1514*** -0.0147* -0.9225*** -1.7175***
Yl 6 0.0047 0.0315 0.0556 0.0210 -0.1401 -0.0916** -0.0278 -0.0292*
Note-. The table reports the parameter estimates of model (1), using a number of model variations. Asterisks denote
significance at the * 90%, ** 95% and *** 99% level, respectively. The significance is judged through the p-value obtained
in a bootstrap. The vector of variables is (rus, bvs, sv , rEA, b , iEA, e)'.
The robustness test in column 4 of Table IV relates to possible differences over the business
cycle. Andersen et al. (2007) had identified different linkage patterns depending on the stage of the
business cycle, with stronger cross-country linkages during contraction periods. To test for these
effects in our framework, we exclude all data that are observed in periods of US recessions, where
the latter are identified based on the NBER classification. While most parameters are relatively
robust, we find interesting patterns for short-term interest rates, with a reduced linkage from the
DOI: 10.1002/jae
USA to the euro area, and an intensified linkage in the other direction. This suggests that duri
US recessions, US short-term interest rates decouple from those in Europe, whereas Europ
more dependent on the developments in the USA.
Another possibility is that international linkages differ during international financial crises.
Especially if these are not emanating in the USA or the euro area, it is possible that sh
originate outside our system, and that we misattribute the various transmission channels. We h
therefore excluded all such crises days from our analysis, and re-estimated the model (see colum
5 of Table IV).15 As with the exclusion of recessions, we tend to find a somewhat reduced impa
of the USA on the euro area, along with an increasing feedback from euro area short rate
US short rates. In analogy to the recession case, this suggests that the US decouples from Europ
during financial crises, whereas developments in Europe are following those in the US m
closely. Note, however, that the main resultthat US developments are relatively more import
than euro area developmentsgoes through also in this case.
The last three robustness tests check whether the choice of 2-day returns, of 202-day rolling
windows to estimate the variances that underlie our regime identification, and the identificati
based on the seven regimes where only one individual market shows elevated conditional volatil
affect our results. While some parameter estimates change somewhat in magnitude, it is impor
to note that the main result of US dominance goes through in all these robustness tests.
Finally, none of the other key results, namely the magnification of responses when allowing
for indirect effects and the existence of substantial cross-market linkages, is affected in any of
robustness tests. Our earlier conclusion about the direction and importance of the transmission
shocks is therefore largely robust.
5. CONCLUSIONS
The objective of the paper has been to propose an empirical framework for meas
analyzing the transmission of international financial shocks across markets as wel
countries. This framework models each financial asset with a multifactor model and then u
heteroskedasticity of the asset return to identify the financial transmission between m
and equity markets and exchange rates within and between countries. For a long time, a
policymakers and practitioners have known of the complexity of these linkages. Howeve
research has been devoted to measuring them or to understanding the extent to which the
co-movement across asset prices around the world. The literature mostly concentrated e
different markets within one country, or on one market across different countries. In th
have shown that indeed the transmission of shocks is highly complex, and that most tr
channels are significant and economically relevant. These findings are important for pol
in understanding the exposure to foreign financial market shocks, as well as for risk m
and international portfolio diversification.
We have applied this framework in this paper to the transmission of financial market
between the USA and the euro area over the 20-year period of 1989-2008. The re
the importance of domestic as well as international spillovers, both within asset cla
as across financial assets. Although the strongest international transmission of shocks ta
15 These relate to the crises in Mexico in 1994/1995, Asia in 1997, Russia in 1998, Brazil in 1999, Argentina
and Turkey in 2002. Their dates are specified following Bekaert and Harvey (2004). Excluding furthermore t
crisis and the 2007-2009 financial crisis does not alter results.
DOI: 10.1002/jae
within asset classes, we find evidence that international cross-market spillovers are significant,
both statistically as well as economically. Such an analysis of international cross-market spillovers
is fairly novel in the literature on financial linkages. Overall, US financial markets explain on
average around 30% of euro area financial market movements in the period 1989-2004, whereas
euro area markets account for about 6% of the variance of US asset prices.
A second key result of the paper is that the direct transmission of financial shocks within asset
classes is often magnified substantially through indirect spillovers through other asset prices. In
particular for the case of bond yields, this increase is of the order of 50%.
Many open questions and avenues for future research remain. We have made several assumptions
with the objective to solve the problem of identification. Interpreting each asset pricing equation
as a structural equation is one such assumption, although we have argued that this interpretation
is not required for the precise identification of the financial transmission process but primarily
serves to give an economic meaning to the financial linkages in the data. Relaxing some of
these assumptions, expanding the current model and using high-frequency data may certainly be
fruitful avenues for future research. This paper hopes to contribute to a better understanding of
the complexity of financial market integration and interdependence.
ACKNOWLEDGEMENTS
We are grateful to Terhi Jokipii for excellent research assistance. We would like to
anonymous referees, Steven Durlauf (the editor), as well as Steve Cecchetti, Jon Faust, D
Malliaropulos, Pierre Siklos, Mark Spiegel, Cedric Tille and the participants of th
conference on 'Global financial integration, stability and business cycles', the New
conference on 'Financial globalization', the European Finance Association Annual M
the American Economic Association Annual Meetings, the ISK Vienna Symposi
Management and seminars at Trinity College Dublin, the Central Bank of Norway,
University and Bilkent University for comments and suggestions. This paper presents th
personal views and does not necessarily reflect those of the European Central Bank.
REFERENCES
Albuquerque R, Bauer G, Schneider M. 2007. International equity flows and returns: a quantitativ
rium approach. Review of Economic Studies 74: 1-30.
Andersen TG, Bollerslev T. 1998. 'Deutsche Mark-dollar volatility: intraday activity patterns, m
nomic announcements, and longer run dependencies. Journal of Finance 53: 219-265.
Andersen TG, Bollerslev T, Diebold FX, Vega C. 2003. Micro effects of macro announcements: re
price discovery in foreign exchange. American Economic Review 93: 38-62.
Andersen T, Bollerslev T, Diebold F, Vega C. 2007. Real-time price discovery in stock, bond and f
exchange markets. Journal of International Economics 73: 251-277.
Anderson H, Dungey M, Osborn D, Vahid F. 2007. Constructing historical euro area data. CAMA W
Paper 18-2007.
Bae K-H, Karolyi A, Stulz R. 2003. A new approach to measuring financial contagion. Review of F
Studies 16: 717-763.
Baele L, Bekaert G, Inghelbrecht K. 2007. The determinants of stock and bond return comovements. National
Bank of Belgium Working Paper No. 119.
Balduzzi P, Elton EJ, Green TC. 2001. Economic news and bond prices: evidence from the U.S. Treasury
market. Journal of Financial and Quantitative Analysis 36: 523-543.
DOI: 10.1002/jae
Barsky RB. 1989. Why don't the prices of stocks and bonds move together? American Economic Re
79: 1132-1145.
Becker KG, Finnerty JE, Friedman J. 1995. Economic news and equity market linkages between the
and U.K. Journal of Banking and Finance 19: 1191-1210.
Bekaert G, Harvey CR. 2004. A chronology of important financial, economic and political events in emerg
markets. Duke University.
Bekaert G, Hodrick RJ, Zhang X. 2009. International stock return comovements. Journal of Finance
2591-2626.
Bernanke BS. 2007. Globalization and monetary policy. Speech at the Fourth Economic Summit, Stanford
Institute for Economic Policy Research: Stanford, CA, 2 March 2007.
Bernanke BS, Kuttner KN. 2005. What explains the stock market's reaction to federal reserve policy. Journal
of Finance 60: 1221-1257.
Beyer A, Juselius K. 2008. Does it matter how to measure aggregates? The case of money trans
mission mechanisms in the euro area. In 9th Workshop of the Euro Area Business Cycle Network.
http://www.eabcn.org/workshops/cambridge_2008/documents/Juselius.pdf [21 February 2010].
Bollerslev T, Cai J, Song FM. 2000. Intraday periodicity, long memory volatility, and macroeconomic
announcement effects in the U.S. Treasury bond market. Journal of Empirical Finance 7: 37-55.
Boyd JH, Jagannathan R, Hu J. 2005. The stock market's reaction to unemployment news: why bad news is
usually good for stocks. Journal of Finance 60(2): 649-672.
Briiggemann R, Liitkepohl H, Marcellino M. 2008. Forecasting euro area variables with German Pre-EMU
data. Journal of Forecasting 27: 465-481.
Campbell JY, Ammer J. 1993. What moves the stock and bond markets? A variance decomposition for
long-term asset returns. Journal of Finance 48: 3-37.
Cass D, Pavlova A. 2004. On trees and logs. Journal of Economic Theory 116: 41-83.
Connolly R, Wang AF. 2003. International equity market comovements: economic fundamentals or conta
gion? Pacific-Basin Finance Journal 11: 23-43.
David A, Veronesi P. 2008. Inflation and earnings uncertainty and volatility forecasts: a structural form
approach. Manuscript, Graduate School of Business, University of Chicago, http://papers.ssrn.com/sol3/
papers.cfm?abstractJd=507122 [21 February 2010].
Diebold F, Yilmaz K. 2009. Measuring financial asset return and volatility spillovers, with application to
global equity markets. Economic Journal 119: 158-171.
Dungey M, Martin VL. 2007. Unravelling financial market linkages during crises. Journal of Applied
Econometrics 22(1): 89-119.
Ehrmann M, Fratzscher M. 2004. Taking stock: monetary policy transmission to equity markets. Journal of
Money, Credit and Banking 36(4): 719-737.
Ehrmann M, Fratzscher M. 2005. Equal size, equal role? Interest rate interdependence between the euro area
and the United States. Economic Journal 115: 930-950.
Ehrmann M, Fratzscher M, Rigobon R. 2005. Stocks, bonds, money markets and exchange rates: measuring
international financial transmission. NBER Working Paper 11166.
Engle RF, Ito T, Lin WL. 1990. Meteor-showers or heat waves: heteroskedastic intradaily volatility in the
foreign exchange market. Econometrica 55: 391-407.
Faust J, Rogers J, Wang S, Wright J. 2005. The high frequency response of exchange rates and interest rates
to macroeconomic announcements. Journal of Monetary Economics 54(4): 1051-1068.
Flannery MJ, Protopapadakis A. 2002. Macroeconomic factors do influence aggregate stock returns. Review
of Financial Studies 15: 751-782.
Fleming MJ, Remolona EM. 1997. What moves the bond market. Economic Policy Review 3: (December):
31-50.
Fleming MJ, Remolona EM. 1999. Price formation and liquidity in the U.S. Treasury market: the response
to public information. Journal of Finance 54: 1901-1915.
Forbes KJ, Chinn M. 2004. A decomposition of global linkages in financial markets over time. Review of
Economics and Statistics 86: 705-722.
Forbes KJ, Rigobon R. 2002. No contagion, only interdependence: measuring stock market co-movements.
Journal of Finance 57(5): 2223-2261.
Goldberg L, Leonard D. 2003. What moves sovereign bond markets? The effects of economic news on U.S.
and German yields. Current Issues in Economics and Finance 9: 1-7.
DOI: 10.1002/jae
Griffin JM, Karolyi GA. 1998. Another look at the role of the industrial structure of markets for international
diversification strategies. Journal of Financial Economics 50: 351-373.
Hamao Y, Masulis RW, Ng V. 1990. Correlations in price changes and volatility across international stock
markets. Review of Financial Studies 3: 281-307.
Hartmann P, Straetmans S, de Vries C. 2004. Asset market linkages in crisis periods. Review of Economics
and Statistics 86: 313-326.
Heston SL, Rouwenhorst KG. 1994. Does industrial structure explain the benefits of international diversifi
cation? Journal of Finance 36: 3-27.
King M, Sentana E, Wadhwani S. 1994. Volatility and links between national stock markets. Econometrica
62: 901-934.
Lin W, Engle RF, Ito T. 1994. Do bulls and bears move across borders? International transmission of stoc
returns and volatility. Review of Financial Studies 7: 507-538.
Pavlova A, Rigobon R. 2007. Asset prices and exchange rates. Review of Financial Studies 20(4
1139-1180.
Rigobon R. 2003. Identification through heteroskedasticity. Review of Economics and Statistics 85: 777-792.
Rigobon R, Sack B. 2003a. Measuring the reaction of monetary policy to the stock market. Quarterly Journal
of Economics 118: 639-669.
Rigobon R, Sack B. 2003b. Spillovers across U.S. financial markets. NBER Working Paper No. 9640,
Cambridge, MA.
Rigobon R, Sack B. 2004. The impact of monetary policy on asset prices. Journal of Monetary Economics
51: 1553-1575.
Sentana E, Fiorentini G. 2001. Identification, estimation and testing of conditionally heteroskedastic factor
models. Journal of Econometrics 102: 143-164.
Shiller RJ, Beltratti AE. 1992. Stock prices and bond yields. Journal of Monetary Economics 30: 25-46.
Wright P. 1928. The Tariff on Animal and Vegetable Oils. Macmillan: New York.
Zapatero F. 1995. Equilibrium asset prices and exchange rates. Journal of Economic Dynamics and Control
19: 787-811.
DOI: 10.1002/jae
TECHNICAL APPENDIX: IDENTIFICATION THROUGH HETEROSKEDASTICITY (IH)
The theoretical derivation of the methodology is provided in Rigobon (2003). This technical
appendix provides an exposition of the methodology as applied to the international transmission
of financial market shocks. Assume that there is a set of N endogenous variables denoted by X,
and assume that the 'true' model (structural form) is denoted as
A X c -j-
In general it is assumed that the diagonal of the matrix A is equal to 1 (the standard normalization
assumption).16 The constant terms are denoted by the vector c. The residuals e are known as the
structural shocks and in most macroeconomic and development applications they are assumed to
be uncorrelated. In other words:
a,S, 1 0 0
0
7l 2
var (e) = =
0 0 o:
s,N
The reduced form of the model is given by
X=Axc+A~xs
with the reduced form residuals defined as
n) = A~le
Although the structural form cannot be estimated, the reduced form can. In this particular case,
we can estimate the reduced-form constant terms and the covariance matrix of the reduced form
residuals. In fact, these are the only statistics that can be obtained from the data. The question
of identification has to do with whether the structural coefficients can be recovered from the
reduced-form estimates. In general, the answer is no.
Note that in a system with N endogenous variables the matrix A has N(N 1) terms, and the
covariance matrix of the structural form residuals has N termswhich accounts for a total of
N2 unknowns. On the other hand, the covariance matrix of the reduced-form residuals provides
N(N + l)/2 equations. This leaves N(N l)/2 unidentified parameters, the number of reduced
form coefficientsor moments that can be estimated from the reduced formis smaller than the
number of unknownsor structural parameters.
The idea of identification through heteroskedasticity is to increase the number of available
moments or equations such that the problem of identification can be solved. Let us assume,
therefore, that the sample can be split in two (s 1 and s2) in a manner that satisfies the following
properties:
AXsl = c + esl
16 We simplify the problem at hand for expositional purposes, but it should be clear that the inclusion of lags does not
affect the arguments made here.
DOI: 10.1002/jae
AXs2 = C + Ss2
var (V1) = E*1
var (V2) = s2
This structure adds the assumption that the data are heteroskedastic. All other aspects of the
structure remain identical. In particular, the model assumes that the constant c and the set of
coefficients on the endogenous matrix A are the same in the cross-section. The only difference is
in the variance of the residuals of the structural equations across the two subsamples. In fact, this
is the typical assumption we would make if we were estimating an ARCH model.
This simple assumption implies that we can estimate a reduced-form covariance matrix for each
subsample, providing N(N + l)/2 additional equations, but only N additional unknowns. Therefore,
the existence of two regimes will already be sufficient for exact identification of the model: the
second regime adds N(N + l)/2 N N(N l)/2 free parameters, equal to the unidentified
N(N l)/2 parameters.
It is important to highlight another identifying assumption in this procedurethe fact that the
structural shocks are uncorrected. If the shocks were correlated and there is no restriction on
the variation of such covariance, then every heteroskedastic regime adds as many equations as
unknowns, and the problem of identification could not be solved. However, note that there is a
crucial difference between parameter stability and the covariance assumption. Parameter stability
can be tested if there are enough regimes. The reason is that it can be estimated assuming that
parameters are stable and check the overidentifying restrictions. The zero correlation assumption,
on the other hand, cannot be tested. This is the identifying assumption, and therefore maintained.
DOI: 10.1002/jae

Article 1

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Article 1

Hochgeladen von

Copyright:

Verfügbare Formate

STOCKS, BONDS, MONEY MARKETS AND EXCHANGE RATES: MEASURING

INTERNATIONAL FINANCIAL TRANSMISSION

STOCKS, BONDS, MONEY MARKETS AND EXCHANGE

MICHAEL EHRMANN,3* MARCEL FRATZSCHERa AND ROBERTO RIGOBONb

Received 11 April 2008; Revised 15 November 2009

Copyright 2010 John Wiley & Sons, Ltd.

3. ESTIMATION METHODOLOGY AND DATA

3.1. Identification through Heteroskedasticity

A y, = & + n (L) yt-\ + V (L) zt + Mr (1)

y, =A-li} + A~ln (L) yr_! +A~lV (L) z,+A~^t

it should be obvious that heteroskedasticity can be combined with other assumptions

3.2. Overidentifying Restrictions

Identification through heteroskedasticity is a powerful method to identify systems of equ

/ 1 0l\2 ai3 /$i4 ^15 yn\

A = @41 Pa2 Pa3 1 <*45 <*46 747

\ e ! \ Yi\ yi2 Yi3 Yia Yi5 Yi6 1 >

between markets, whereas the parameters of the reduced-form model capture

/ 1 a\2 < 0 <*13 1 < pu < 0 0 0 K17\

3.3. Estimation Procedure

The estimation procedure is as follows:

min g'g with g = A'Y,jA Q,

s.t. E,- is diagonal, A restrictions

var(yf) = [/ -A-In (L)]var(/z,) {[/ - A"'n (L)]r}-1and

4.1. International transmission

4.1.1 Direct Effects

r,us = 0.0008 r,EA - 0.0038 e, + ... (3)

b,vs = 0.4402 b,EA + 0.0678 e, + ... (4)

s,vs = 0.0594 srEA - 0.1324 <?, + ... (5)

r,EA = 0.1704 r,us + 0.0043 e, + ... (6)

b,EA = 0.4892 btvs - 0.0596 e, + ... (7)

s,EA = 0.6269 s,vs + 0.7071 e, +... (8)

Table I. Parameter estimates and bootstrap results of structural form model

Point estimate Bootstrap

a 12 -0.2850*** -0.2988 0.0474 0.0000

<*45 -0.0761*** -0.0802 0.0379 0.0040

USA to euro area

P52 -0.4892*** -0.4668 0.0429 0.0000

063 -0.6269*** -0.6376 0.0716 0.0000

Euro area to USA

/14 -0.0008 -0.0078 0.0157 0.1756

063 -0.0594 -0.0397 0.0860 0.2515

Exchange rate effects

Y31 0.1324 0.1303 0.1053 0.1018

K57 0.0596*** 0.0596 0.0047 0.0000

Y61 -0.7071*** -0.6758 0.0634 0.0000

Yll 3.4602*** 3.4396 0.1046 0.0000

YlA 0.3800*** 0.3840 0.0952 0.0100

Yt 5 -4.2461*** -4.1854 0.1435 0.0000

Table II. Parameter estimates and bootstrap results of reduced-form model

Point estimate Bootstrap

044 1.0074*** 1.0095 0.0081 0.0000

[ASA (o euro area

&41 0.2099*** 0.1654 0.0572 0.0000

Euro area to USA

b\4 0.0084*** 0.0193 0.0229 0.0040

Exchange rate effects

Table II. (Continued)

Point estimate Bootstrap

Ctf 0.0356*** 0.0337 0.0057 0.0000

4.1.2 Overall Effects

r,us = 0.0084 /j** + 0.1859 ^ + 0.0000 ^ + 0.0071 /jeJ + ... (9)

r,EA = 0.2099 + 0.1176 + 0.0014 + 0.0010 ne,t + ... (12)

Figure1.Distrbuonfstruc aloeficntsofhebncmark odelin50botsrapelictons

A = @41 Pa2 Pa3 1 <45 <46 747

l^e,t 0.14% 2.54% 1.72% 0.01% 2.47%* 7.43%*** 80.25%'

^c,t 1.70%* 1.01%* 0.25% 0.04% 1.76%* 3.00%*** 0.61%'

<45 -0.0761 -0.2656* -0.1654* -0.0579 -0.0675* 0.0000 -0.0190* 0.0000

$14 -0.0008 -0.0216 -0.0128 -0.0455 -0.0444* -0.1433 -0.0941*** -0.0046*