Investigating Global Imbalances: Empirical evidence from a GVAR approach

Economic Modelling 64 (2017) 201–210

Contents lists available at ScienceDirect

Economic Modelling journal homepage: www.elsevier.com/locate/econmod

Investigating Global Imbalances: Empirical evidence from a GVAR approach

MARK

Timo Bettendorf1 Deutsche Bundesbank, Germany

A R T I C L E I N F O

A BS T RAC T

JEL classification: F10 F32 F41

This paper investigates the development of external imbalances from an international perspective by estimating a Global VAR model. Specifically, we simulate the effects of shocks to relevant macroeconomic variables in the United States and the oil price on international trade balances and quantify the relative importance of these shocks using variance decompositions. Overall, we find evidence for the joint dynamics of our variables as drivers of the imbalances and provide further evidence for different hypotheses of Global Imbalances. The results suggest that shocks to US’ real GDP, real stock prices and the oil price are of particular importance for international trade balances and may thus be interpreted as support for parts of the global saving glut hypothesis as well as the international wealth channel.

Keywords: Global Imbalances Global var International trade Open economy macroeconomics

1. Introduction In the early 1990s, external balances of several major economies began to widen considerably. The increased divergence of current account balances is known as Global Imbalances (GIs). The objective of this research is to investigate drivers of trade balances from an international perspective by estimating a GVAR model, which allows us to measure the effects of country-specific shocks on international trade balances. Since the trade balance is a major component of the current account balance, our research is related closely to the phenomenon of GIs. In this context, we relate the country-specific shocks to theories of GIs, which enables us to shed light on the relevance of those theories. Specifically, we employ a data driven approach and estimate a GVAR model for the period 1981Q1–2011Q2. We provide generalized impulse response functions (GIRFs) of international trade balances following shocks to variables which have been named as potential determinants of GIs by different theories, such as US’ real equity prices, real GDP and the oil price. Generalized forecast error variance decompositions (GFEVDs) for the trade balances of countries subject to large surpluses or deficits such as the United States, Germany, China, Japan, Spain and the United Kingdom suggest that those variables are of particular importance for international trade balances. These empirical findings thus enable us to draw conclusions about the

relative importance of those shocks for the corresponding trade balances and the relevance of certain theories on GIs. The importance of the oil price and equity prices may be interpreted as empirical support for parts of the global saving glut hypothesis and the international wealth channel. Both hypotheses will be discussed in the following section. Only a few authors have considered the multinational perspective of this phenomenon so far. We follow the ideas of Bussière et al. (2012), but contribute in several ways to existing literature. Firstly, we introduce short-term and long-term real interest rates as well as real equity prices as possible drivers of trade balances. The new variables provide us with a deeper insight into the importance of asset prices in the context of GIs. As argued earlier, asset prices enable us to shed light on the importance of an international wealth channel. Secondly, we provide a disaggregated analysis on how oil price shocks affect international trade balances. Thirdly, the results from the GVAR model allow us to draw conclusions about the country-specific relative importance of trade balance determinants. Since we are particularly interested in external balances, we deviate from the standard GVAR literature (see Dees et al. (2007)) and rely on variables in real terms, which allows us to abstract from the common monetary policy of the EMU members and to model all EMU members separately. This is of particular importance, because we observe a strong divergence of current account balances within the EMU.

E-mail address: [email protected]. This paper respresents the authors' personal opinions and does not necessarily reflect the views of the Deutsche Bundesbank or its staff. The author is grateful to Ron Smith, HansMartin Krolzig, Miguel León-Ledesma, Michael Binder, Reinhold Heinlein, Andrew Mountford, Keisuke Otsu and audiences at the 3rd Conference on Recent Developments in Macroeconomics (ZEW, Mannheim), Money, Macro and Finance Conference 2012 (Trinity College, Dublin) as well as seminar participants at Kent and Goethe University Frankfurt for valuable comments and suggestions. 1

http://dx.doi.org/10.1016/j.econmod.2017.03.033 Received 4 November 2016; Received in revised form 19 March 2017; Accepted 30 March 2017 0264-9993/ © 2017 Elsevier B.V. All rights reserved.

Economic Modelling 64 (2017) 201–210

T. Bettendorf

He argues that Asian savings flooded international capital markets and led to a decrease in world interest rates. These low interest rates then stimulated US consumption and discouraged households savings. Consequently, the US current account balance deteriorated, while Asian current account balances improved. Moreover, high oil prices increased revenues in oil-exporting countries so that these countries became lenders on international financial markets. Hence, their current account positions improved remarkably as well. The IMF (2005) points out that global saving rates have declined, rather than increased. Furthermore, investment rates of East Asian countries which were affected by the 1997 Asian Crisis dropped sharply to levels far below the saving rates. The theory of a global saving glut hypothesis could thus also be interpreted as an investment drought hypothesis. Nevertheless, one has to consider country-specific saving rates, because the Chinese rate, for instance, increased strongly during the 2000s (see León-Ledesma and Mihailov (forthcoming)). But also institutions contributed to high savings. The wish for precautionary foreign reserves following the 1997 Asian currency crisis induced central banks to buy foreign reserves. Others tried to prevent exchange rate appreciations (see Aizenman and Marion (2003)). Since central banks tend to hold their reserves in government bonds, these actions also led to a decline in US long-term interest rates. Taylor (2008) sees the current account imbalances from a different perspective. He argues that persistently low interest rates are the result of a loose monetary policy rather than a global saving glut. During the 1990s, the Fed kept the federal funds rate for a at levels far below what would have been suggested by the Taylor-Rule. According to Bems et al. (1995), the low interest rates led to a decline in savings and thus, a current account deficit. These long-run developments have in common that the readjustment of the imbalances is a smooth process. This is not the case for the theories, which we group as short-run developments. The implication of these theories is that readjustment may be sudden and severe. According to Laibson and Mollerstrom (2010), asset price bubbles may induce (international) wealth effects, which increase consumption and translate thus into trade deficits, assuming the consumption basket of households includes foreign goods. According to their theory, the appearance of an asset bubble affects household savings negatively and consumption positively. This model is in line with the findings of Case et al. (2005), who conclude that an increase in stock and housing prices leads to an increase in private consumption. A collapse of the bubble, would thus lead to a sudden readjustment of the current account. We do not discuss further theories such as the twin deficit hypothesis, because we are not able to analyze them within our model.

Aggregating the EMU members would overshadow these effects. The model set-up in real terms relates this research close to the class of international real business cycle (IRBC) models such as Mendoza (1991) or Baxter (1995), while the standard GVAR literature is closely related to the New-Keynesian paradigm. This paper is organized as follows. Section (2) provides a literature review covering the theories of GIs and related empirical studies. In Section (3), we discuss the data and methodology. We perform an analysis of our data and present the results in Section (4). We present our conclusions in Section (5). 2. Literature review GIs may be defined as “external positions of systemically important economies that reflect distortions or entail risks for the global economy” (Bracke and Bussière, 2010, p.12). They attracted early attention from policy makers and researchers because capital started to flow from emerging economies into developed economies, which is against the textbook wisdom. According to the standard neoclassical growth model, one would expect that economies with higher marginal products (i.e. higher returns on investment) attract more foreign capital. This is not always the case. León-Ledesma and Mihailov (forthcoming) argue that although risk-adjusted rates in developed economies may be higher due to lower probabilities of default and higher legal investment protections compared to emerging markets, countries with lower investment-output ratios tend to attract more capital. This refers to the so-called ”Allocation Puzzle” discussed by Gourinchas and Jeanne (2013). Another reason why GIs attracted attention is the high persistence of the imbalances. This pattern worried policy makers in particular, because long and persistent current account deficits entail a high risk for the economy. To discuss the different theories of GIs, we follow León-Ledesma and Mihailov (forthcoming), and classify them into long-run and shortrun developments. Both views refer to economic distortions, but theories which are related to the long-run developments tend to explain the high persistence of current account imbalances, whereas theories referring to short-run developments are more concerned about the cyclical component of GIs. However, there are also theories stating that the imbalances may be explained by an intertemporal model, which does not refer to any distortions. Engel and Rogers (2006) for example find that using survey data of future GDP growth, the intertemporal model is able to explain the motion of the US current account balance very well. But Hoffmann et al. (2011) also show that using growth expectations based on observed changes in productivity, the intertemporal model can explain the US current account remarkably well. Several other theories refer to the long-run developments, one such being the Bretton Woods II hypothesis. Dooley et al. (2003, 2004, 2009) argue that the currency pegs of Asian economies to the US dollar kept the relative prices of Asian goods at a low level. This increased price competitiveness of Asian economies translated into current account surpluses. Moreover, the degree of financial market integration has implications for household savings. If financial markets are less developed, there might be a lack in the supply of safe assets, which leads to capital flows from economies with less developed financial markets to economies with highly developed markets. Caballero et al. (2008) demonstrate in a theoretical model that an insufficient supply in safe assets causes such a capital flow. Similarly, Mendoza and Quadrini (2009) argue that insufficient social insurance may increase precautionary savings and thus a high demand for safe assets to insure against income shocks. This, however, requires safe assets, which are domestically not available. Both authors deal with the same question, but provide different answers. Whereas Caballero et al. (2008) explain the capital flows with an imbalance in the supply of safe assets, Mendoza and Quadrini (2009) argue that there is an imbalance in the demand for safe assets. These explanations refer directly to the Allocation Puzzle mentioned above. Bernanke (2005) supplements these theories.

2.1. Empirical studies The empirical literature on GIs focuses mainly on specific theories rather than the phenomenon as a whole. There is a large strand of literature, which tests the intertemporal model of the current account. Lee and Chinn (2006) for example estimate bivariate SVAR models with long-run restrictions a la Blanchard and Quah (1989) using real exchange rate and current account data from the G7 countries. They find that temporary (i.e. monetary) shocks explain more of the current account variation than permanent (i.e. productivity) shocks and lead to an improvement of the current account. Karadimitropoulou and LeónLedesma (2009) provide a more comprehensive set-up by distinguishing between temporary and permanent shocks to domestic net output, preference shocks and external supply shocks. With the exception of France, the authors find empirical evidence for the intertemporal model in the G6 countries (G7 minus the US). Moreover, they conclude that preference shocks and external supply shocks explain most of the current account fluctuations in their system. Bracke and Fidora (2008) focus on several theories which we refer to as long-run developments by applying a SVAR model with sign restrictions on the impulse response functions to identify different shocks. The authors account 202

Economic Modelling 64 (2017) 201–210

T. Bettendorf

but differs in an important way. We discuss the effects of shocks on trade balances and the inclusion of real short-term and long-term rates as well as real equity prices which allows us to be more specific about the effects of shocks to asset prices. Our scope is thus a different one.

for monetary, preference and investment shocks which they relate to monetary policy, saving glut and investment drought hypotheses. The authors find that monetary shocks seem to be more important than preference or investment shocks. The advantage of this approach is that shocks are structural, meaning that they can directly be related to a specific theory. Moreover, the authors may draw conclusions about the relative importance of different theories. However, the disadvantage of their approach is that they only consider the United States and a huge aggregate representing the Emerging Markets. This does not allow them to draw any conclusions about the importance of specific countries in the Emerging Markets. Southeast Asia is for example an extremely heterogeneous region, meaning that one would expect different countries in the region to respond differently to specific shocks. Moreover, Bracke and Fidora (2008) do not account for other advanced economies like the UK or European countries. The IMF (2006), in contrast, postulated a GVAR model to analyze the role of oil prices in the context of GIs. The model consists of the United States, China and three large aggregates, namely other advanced countries, other developing countries and oil exporters. The IMF finds that oil price shocks have a positive effect on the external balances of oil exporters and a negative effect on the US trade balance. However, the aggregation of advanced, developing and oil exporting countries does most likely overshadow important aspects of GIs, because aggregates do not account for country-specific differences, so that the heterogeneity across the countries in these aggregates is not accounted for. A disaggregated perception of the world may be more helpful in this context as we are interested in identifying countries where an oil price shock has particularly strong effects on the current account (or trade balance). Studies that test the short-run developments focus mainly on the effects of asset price bubbles. Fratzscher (2010) for example estimate a structural Bayesian VAR model with sign restrictions to identify shocks to real equity, housing and the real effective exchange rate. More precisely, they estimate the model using the differences of US and G6 (G7 minus the US) variables. Their results indicate that shocks to real equity explain more of the trade balance movements than shocks to housing prices, while shocks to the REER are the least important among these three. This study, however, is also silent about the international effects of the US shocks. Another related study by Holinski and Vermeulen (2012) apply the GVAR model to find evidence for international wealth effects, which overcomes this aggregation problem. Their hypothesis is that an increase of real equity prices affects domestic consumption positively, which again leads to a deterioration of the trade balance. They find evidence for their hypothesis in the cases of the United States, the United Kingdom and France. The authors show that the real equity prices are at least as important as the REER for explaining trade balance movements. No evidence was found in the cases of Germany and Japan. However, they only look at responses in countries where the shocks originate. Hence, they are silent about the international transmission of the shocks. Moreover, they employ GIRFs, which makes it impossible to give the shocks a structural interpretation, meaning that they cannot be directly compared with those by Fratzscher (2010). A much more comprehensive study of global trade flows by Bussière et al. (2012) uses the GVAR framework as well. The authors model the real exports and real imports jointly, in combination with the real output and the REER. Their dynamic analysis shows that a shock to US’ real output primarily affects the exports of neighboring countries (Canada and Mexico) as well as the European and Asian economies. Moreover, the authors show that an appreciation of the US’ REER would increase the exports of Japan and several European countries. The impact of a shock to German GDP has positive effects mainly on exports of European countries, but also on those of the US. The authors’ findings concerning a shock to Chinese imports on GDP and exports from other countries are only significant in the case of a few Asian countries. Our analysis is closely related to Bussière et al. (2012),

3. The GVAR model The GVAR modeling approach has been proposed by Pesaran et al. (2004). It allows exploring international linkages of variables by linking country-specific VARX*(pi,qi) models of i = 1, 2, …, N countries with each other, where the X* denotes a vector of foreign variables which enter the country-specific VAR models. These models account for pi lags of the domestic and qi lags of the foreign variables. Hence, it is possible to explore international linkages between different variables and to trace shocks through a worldwide system of single country models. Pesaran and Smith (2006) show that the VARX*(pi,qi) models can be derived as solutions of dynamic stochastic general equilibrium (DSGE) models. Moreover, they demonstrate in this context that shortrun and long-run restrictions can be imposed. Dees et al. (2007) build a GVAR on these findings and impose restrictions on the long-run relationships of several variables by identifying the cointegrating vectors of the country-specific vector error-correction models (VECM). An approximation of the GVAR model to a common factor model has been derived by Dees et al. (2007). Chudik and Pesaran (2013) discuss the case in which one of the units is dominant. The estimation of our GVAR model follows a two-step approach. While the dominant unit and country-specific VARX*(pi,qi) models are estimated separately in the first step, these country-specific VARX* models are then simultaneously linked with each other using trade weights in the second step. The GVAR is then augmented with the dominant unit where commodity prices are modeled. After an explanation of the variables, this methodology will be explained, as it is applied in the GVAR Toolbox 2.0 (Smith and Galesi, 2014) which we use for the estimation. 3.1. Data and variables Overall, we model 33 countries (see Table 1) to analyze the effects of certain shocks on external balances. We use data from Smith and Galesi (2014) and extend the dataset with trade data from the IMF Directions of Trade statistics. The dataset ranges from 1981Q1 to 2011Q2. The sample begins with the availability of Chinese trade data and it ends with the unwinding of GIs. Lane and Milesi-Ferretti (2012), for example, state that the determinants of the emergence and the unwinding of GIs were not the same. Considering a larger sample would thus bias the results of this research. The variables for the GVAR model are closely related to those of international real business cycle (IRBC) models such as Mendoza (1991). Using real variables enables us to model all EMU countries separately by abstracting from the common monetary policy (see Bussière et al. (2012)). We include the real GDP (yit), real short-term Table 1 Countries in the GVAR. Countries ARGENTINA AUSTRALIA AUSTRIA BELGIUM BRAZIL CANADA CHINA CHILE FINLAND FRANCE GERMANY

203

INDIA INDONESIA ITALY JAPAN KOREA MALAYSIA MEXICO NETHERLANDS NORWAY NEW ZEALAND PERU

PHILIPPINES SOUTH AFRICA SAUDI ARABIA SINGAPORE SPAIN SWEDEN SWITZERLAND THAILAND TURKEY UNITED KINGDOM USA

Economic Modelling 64 (2017) 201–210

T. Bettendorf

interest rates (rsrit), real long-term interest rates (rlrit), real equity prices (reqit), real exchange rates (epit) and trade balances (tbit). Additionally, the foreign real GDP (yit*), foreign real short-term interest rates (rsrit*), foreign real long-term interest rates (rlrit*) and foreign real equity prices (reqit*) are included as exogenous variables into the system. These are defined as N

yit* =

∑ ωijyjt ,

N

rsrit* =

j =1

∑ ωijrsrjt , j =1

N

reqit* =

∑ ωijreqjt ,

Table 2 Model Specification. Variables

xit

j =1

Others

xit*

xit

xit*

Real GDP

yit

yit*

yit

yit*

∑ ωijrlrjt

Real Short-Term Rate

rsrit

–

rsrit

j =1

Real Long-Term Interest Rate

rlrit

–

rlrit

rsrit* rlrit*

Real Equity Prices

reqit

–

reqit

reqit*

RER

–

epit*

epit

–

Trade Balance

tbit

–

tbit

–

N

rlrit* =

US

where ωij denotes the trade weight of country i with j. Note here that ωii = 0 . Our trade weights are fixed and represent the average total trade between country i and j over the sample period (1981–2011). As explained earlier, the core of the GVAR model is a trade weight matrix which has a dimension of 33×33.2 Not surprisingly, the United States, followed by Germany, China, Japan and the United Kingdom are the key economies for our system. The matrix exhibits that the linkages between these economies are particularly strong, as the trade shares are relatively high. Moreover, the table unveils strong border effects. China's column displays relatively high trade shares with other countries of the Asia-Pacific region. Consequently, the main dynamics in our model are expected to occur between these economies. Since nominal GDP, which would be the natural variable to normalize the trade balance, is not available for all countries, we model the trade balance as exports over imports. A dominant unit (see Smith and Galesi (2014)) contains oil (oilit), material (matit) and metal (metit) prices. Variable transformations are given by3:

endogenous variables are captured by the ki × ki coefficient matrix Φ. η stands for the IID error term with mean zero. Note that feedback effects from other variables are not taken into account. 3.2.2. Single country models We rely on the AIC to determine the lag lengths pi and qi for our VARX*(pi,qi) models (see Table 5). To present the GVAR we use the VARX*(1,1) model representation, which can be written as

xit = ai 0 + ai1t + Φi1xit −1 + Λi 0 xit* + Λi1xit*−1 + Ψi 0wit + Ψi1wit −1 + uit ,

(2)

where ai0 denotes the coefficients of constants and ai1 the coefficients of time trends.5 Φil and Λil are coefficient matrices of ki × ki dimension for the vectors of domestic and foreign variables. The error-term uit is a ki × 1 vector and assumed to be IID and have a zero mean with a covariance matrix Σii. The variables of the dominant unit enter the single country models exactly as foreign variables do (with the same lag order). Their effects on the endogenous variables is captured by the ki × ki coefficient matrices Ψ. We now define zit as a (ki + ki*) × 1 vector of the domestic and foreign variables as

yit = ln(RGDPit ), rsrit = 0.25*ln(1 + Ritshort /100) − (ln(CPIit ) − ln(CPIit −1)), rlrit = 0.25*ln(1 + Ritlong /100) − (ln(CPIit ) − ln(CPIit −1)), reqit = ln(EQit ) − ln(CPIit ), epit = ln(Eit ) − ln(Pit ),

⎛ xi ⎞ zi = ⎜ *⎟ . ⎝ xi ⎠

tbit = ln(Exportsit ) − ln(Importsit ), oilit = ln(OILt ), matit = ln(MATt ), metit = ln(METt )

Substituting it into (2) yields

where Rt represents the annualized nominal interest rate. Table 2 explains the general specification of the vectors of our model. Foreign trade balances are not included as exogenous variables into the model, as the computation of trade-weighted aggregates of our foreign trade balances would not reflect the foreign trade balances, but a mismatch of those. Trade balances are thus not linked to the weight matrix. Since data is not available for all countries, the model specification of the estimated model deviates slightly from Table 2. Mainly, the longterm rates and equity prices are missing in some cases. Also, Saudi Arabia does not provide a nominal short-term interest rate for our sample. Table 4 displays the exact specification for every single country.

Ai 0 zit = ai 0 + ai1t + Ai1zit −1 + Ψi 0wit + Ψi1wit −1 + uit ,

(3)

where

Ai 0 = (Iki, − Λi 0 ),

Ai1 = (Φi1, Λi1).

For the estimation of the dominant unit and the single-country models, we rely on a VECM* representation. But as a VECM can be mapped back into a VAR representation, we continue with the representation given by Eq. (2). Table 5 displays the number of cointegration relationships for each model. 3.3. Step two: Solving the GVAR In order to solve the GVAR, we now define the vector zit in terms of the global vector xt = (x 0t ′, x1t ′, …, xNt ′) as

3.2. Step one: Estimating the dominant unit and single-country models of the GVAR

zit = Wx i t, 3.2.1. Dominant unit The dominant unit of the GVAR has a VAR(1) specification4:

wt = μ0 + μ1t + Φ1wt −1 + ηt

where Wi denotes a matrix of identity matrices, zeros and trade weights with a dimension of (ki + ki*) × k . Given the global vector xt, Wi yields exactly the same vector zit, which we defined earlier. The vector xt contains the domestic variables of all countries. Hence, we get the

(1)

wt is a ki × 1 vector, containing the endogenous commodity prices (oil, metal and raw materials). μ0 denotes the coefficients of constants and μ1 the coefficients of time trends, t. The lagged effects of the

5 Although economic theory suggests that rsr is stationary, we include a linear trend in all models. Our sample starts in 1981; from then on, real interest rates follow a downward trend. Without explicitly incorporating a trend, the trend would be captured in the VECM coefficients and would produce misleading impulse responses. Estimates without a trend demonstrate exactly this behavior, meaning that the trend is found to be atheoretic, but empirically necessary.

2

The trade weight matrix is provided in an online Appendix. The real exchange rate is defined as in Dees et al. (2007). 4 The lag order of the estimated model is based on the AIC. 3

204

Economic Modelling 64 (2017) 201–210

T. Bettendorf

CAN

CHN

FRA

1

0.5

0.6

0.5

0

0.4

0

-0.5

0.2

5

10

15

20

-1

5

DEU

0.5

10

15

20

-0.5

0.2

2

0

1

20

5

KOR

10

15

20

5

NOR

10

15

20

15

20

15

20

2 1

-0.5

0

20

SAR

0

0.5

15

0

-0.4 15

10

JAP

-0.2 10

5

ITA

0

5

0

0 -1

-0.5 5

10

15

20

-1 5

ESP

1

10

15

20

UK 1

-0.2 0

0

-0.4

-1

-0.6 -1

-0.8 10

15

20

10

USA

0

5

5

5

10

15

20

-2

5

10

Fig. 1. Effects of a 1SD shock to US real GDP on international trade balances (in percent). Note: The time (in quarters) following the shock is plotted on the x-axes and the responses of the trade balances (in percent) are plotted on the y-axes. The solid lines of the impulse response functions correspond to the medians and the dashed lines to the 90% confidence intervals (1000 draws).

H0yt = h 0 + h1t + H1yt −1 + ζt ,

expression

Ai 0 Wx i t = ai 0 + ai1t + Ai1 Wx i t −1 + Ψi 0wit + Ψi1wit −1 + uit .

(4)

where

When stacking all the country models, we obtain the equation

G0xt = b0 + b1t + G1xt −1 + Ψ0wt + Ψw 1 t −1 + ct ,

⎡G0 ⎡b ⎤ −Ψ0 ⎤ ⎡b ⎤ ⎥ , h 0 = ⎢ 0 ⎥ , h1 = ⎢ 1 ⎥ , H0 = ⎢ 0 ⎣ μ1⎦ ⎣ μ0 ⎦ ⎣ m w × k Im w ⎦

(5)

where

⎛ a 00 ⎞ ⎜a ⎟ b0 = ⎜ 10 ⎟ , ⎜ ⋮ ⎟ ⎝ aN 0 ⎠

⎛ a 01⎞ ⎜a ⎟ b1 = ⎜ 11 ⎟ , ⎜ ⋮ ⎟ ⎝ aN1⎠

⎛ u 0t ⎞ ⎜u ⎟ ct = ⎜ 1t ⎟ ⎜⋮⎟ ⎝ uNt ⎠

⎡G1 −Ψ1⎤ ⎡u ⎤ ⎥ , ζt = ⎢ ηt ⎥ , H1 = ⎢ ⎣ t⎦ ⎣ 0m w × k Φ1 ⎦ or

and

⎛ A00 W0 ⎞ ⎟ ⎜ A W G0 = ⎜ 10 1 ⎟ , ⎜ ⋮ ⎟ ⎟ ⎜ ⎝ AN 0 WN ⎠

(6)

yt = c0 + c1t + C1yt −1 + H0−1ζt ,

⎛ A01W0 ⎞ ⎜ ⎟ A W G1 = ⎜ 11 1 ⎟ . ⎜ ⋮ ⎟ ⎜ ⎟ ⎝ AN1WN ⎠

with

c0 = H0−1h 0 , c1 = H0−1, C1 = H0−1H1.

In order to obtain a VAR representation of the GVAR, we define yt = (x′t , w′t )′ as (k + m w ) × 1 vector and write the GVAR as

Eq. (7) represents our final GVAR model. 205

(7)

Economic Modelling 64 (2017) 201–210

T. Bettendorf

CAN

CHN

0

FRA 0.8 0.6 0.4 0.2 0 -0.2

1 0

-1

-1 -2

5

10

15

20

5

DEU

15

20

5

-0.5 -1

10

15

20

-0.5

20

15

20

15

20

0 5

10

15

20

5

NOR

10

SAR

1

6 4

0

0

15

2

KOR 1

20

4

0

-1.5

15

JAP

0.5

5

10

ITA

0

-1

10

2 0

-1 5

10

15

20

5

ESP

10

15

20

5

UK

0

USA 1

0

-1 -2 -3 5

10

15

20

10

0

-0.5

-1

-1

-2 5

10

15

20

5

10

Fig. 2. Effects of a 1SD shock to the US real equity prices on international trade balances (in percent). Note: The time (in quarters) following the shock is plotted on the x-axes and the responses of the trade balances (in percent) are plotted on the y-axes. The solid lines of the impulse response functions correspond to the medians and the dashed lines to the 90% confidence intervals (1000 draws).

3.4. Generalized Impulse responses and generalized forecast error variance decompositions

equation in country i. Hence, no direct policy implications result from these responses. Nevertheless, they are still informative as they show the most likely response following a shock. In order to analyze the proportion of forecast error variance of the trade balance that is explained by shocks to variables in our system, we perform a generalized forecast error variance decomposition (GFEVD). The philosophy of the GFEVD is related to the GIRFs. Shocks are not orthogonalized, meaning that the output is invariant to the ordering of the equations. Since we are allowing for correlations between different shocks by using a non-diagonal covariance matrix, the GFEVDs do not sum to 1. Hence, they cannot be interpreted as the relative contribution of a shock to the forecast error variance. We refer to Dees et al. (2007) for a detailed explanation of the GFEVD.

A common approach for the estimation of impulse responses in structural VAR models is to follow Sims (1980) and to impose k (k − 1)/2 restrictions in the form of a Cholesky decomposition on the variance covariance matrix (see for example Sims (1986)). Others impose long-run restrictions (see for example Blanchard and Quah, 1989), or directly restrict the impulse responses (see for example Uhlig, 2005). These restrictions give the shocks an economic interpretation and thus have implications for policy makers. We employ generalized impulse response functions (GIRFs), in order to investigate the effects of shocks. These GIRFs were introduced for nonlinear models by Koop et al. (1996) and for linear multivariate models by Pesaran and Shin (1998). The advantage of this method is that the impulse responses are invariant to the ordering of the equations. However, shocks are not orthogonalized, so that they cannot be interpreted in the same way as for example structural shocks. The GIRFs show us what is most likely to happen after a shock to the lth

4. Data analysis 4.1. Impulse response analysis In this section, we perform a dynamic analysis of our model. Given 206

Economic Modelling 64 (2017) 201–210

T. Bettendorf

CAN

1.5

CHN 1

1 0 -0.5

-0.5

0

0.5

5

10

15

DEU

-0.5 -1 -1.5 10

-1

-2

-1.5 5

15

20

-1

-4

-1.5

-6 5

10

15

20

15

20

ESP 1

10

15

20

15

20

15

20

4 2 0 -2 5

10

15

20

5

UK

2

2

20

6

-0.5 10

15

SAR

0

5

5

NOR

0.5

-2

10

JAP

0

1

-1

5

-2

1.5

0

15

-0.5

20

KOR

10

ITA

0

0

5

-1 20

FRA

0

10

USA 2 1

1

0

0

-1

-1 5

10

15

20

0

5

10

15

20

5

10

Fig. 3. Effects of a shock to the oil price on international trade balances (in percent). Note: The time (in quarters) following the shock is plotted on the x-axes and the responses of the trade balances (in percent) are plotted on the y-axes. The solid lines of the impulse response functions correspond to the medians and the dashed lines to the 90% confidence intervals (1000 draws).

illustrates that the shock to the US real GDP leads to a deterioration of the US trade balance. However, the shock does not affect the trade balance contemporaneously. We observe a significant temporary decrease of approximately 0.7%. Corresponding significant increases of foreign trade balances can be found in France, Germany, Japan and Spain. In France, the trade balance increases by about 0.3%. The German trade balance responds in a significantly positive manner only during the short run (+0.3%). A similar pattern can be observed in Japan, where the trade balance increases significantly over the first two quarters. We observe a contemporaneous effect of 0.4%. International Real Business Cycle (IRBC) models, which allow not only for intertemporal utility maximization of households, but also for investment decisions of firms, can explain this pattern. Baxter (1995), for example, provides a two-country model of international trade with capital accumulation and international investment flows which is able to explain the negative correlation between output and the trade balance. She shows that following a permanent productivity shock in a SOE with fixed labor input and incomplete markets, the trade balance deteriorates. Additionally, she provides stylized facts of business cycles

the large size of the model, we focus on a smaller set of countries and shocks which we find to be important in the context of GIs. We focus on the United States, the country with the highest current account deficit in absolute terms and its’ main trading partner, Canada. Moreover, we consider the two Eurozone core countries Germany and France as well as the two largest periphery countries, namely Italy and Spain. China, Japan and Korea account for the importance of Asia in the context of GIs. Norway and Saudi Arabia are oil exporters. The United Kingdom is an important trading partner for most of the countries. Because of high trade shares, one would expect strong spillovers. We estimate GIRFs of shocks to the US’ real GDP, US’ real equity prices an the oil price. As discussed in Section (2), a shock to real equity prices relates to the hypothesis of international wealth effects, while the oil price shock may be attributed to the global saving glut hypothesis. The shock to real GDP relates to literature on IRBC models in which GDP is negatively correlated with the trade balance.

4.1.1. A positive shock to United States’ real GDP Following an 1SD shock, the US’ real GDP increases by 0.4%. Fig. 1 207

Economic Modelling 64 (2017) 201–210

T. Bettendorf

Table 3 Generalized forecast error variance decomposition. CHINA

CHN CHN DU CHN CHN USA DU USA DU USA JAPAN

GERMANY

tb ep mat rir y eq oil rir met rlr

JAP rir JAP rlr DU oil JAP tb DU mat CHN y USA eq DU met USA y SAR y UNITED KINGDOM UK DU UK USA SAR UK UK ITA UK DU

tb oil rir y y y mat rir rlr met

Year 0

Year 1

Year 2

Year 5

Year 10

0.72 0.00 0.04 0.01 0.01 0.03 0.03 0.00 0.00 0.01

0.45 0.13 0.10 0.05 0.03 0.02 0.02 0.02 0.01 0.01

0.33 0.21 0.11 0.08 0.04 0.02 0.02 0.01 0.02 0.01

0.25 0.26 0.10 0.09 0.05 0.02 0.02 0.02 0.03 0.01

0.22 0.28 0.09 0.09 0.05 0.02 0.03 0.02 0.03 0.01

Year 0 0.00 0.00 0.09 0.52 0.01 0.07 0.00 0.01 0.02 0.01

Year 1 0.25 0.13 0.12 0.06 0.05 0.01 0.01 0.01 0.01 0.01

Year 2 0.25 0.13 0.12 0.03 0.06 0.01 0.02 0.02 0.01 0.01

Year 5 0.22 0.11 0.11 0.02 0.08 0.01 0.02 0.04 0.01 0.01

Year 10 0.21 0.11 0.11 0.02 0.09 0.01 0.02 0.05 0.00 0.01

Year 0 0.79 0.09 0.00 0.02 0.01 0.00 0.01 0.01 0.00 0.00

Year 1 0.57 0.16 0.09 0.02 0.01 0.01 0.01 0.01 0.01 0.01

Year 2 0.52 0.17 0.09 0.02 0.01 0.01 0.02 0.01 0.01 0.01

Year 5 0.44 0.17 0.09 0.02 0.01 0.01 0.03 0.01 0.01 0.02

Year 10 0.41 0.16 0.09 0.01 0.01 0.01 0.03 0.01 0.01 0.03

Year 0

Year 1

Year 2

Year 5

Year 10

DEU DEU DU DEU DEU DU DU DEU CHN USA SPAIN

tb rlr oil rir y met mat ep y y

0.72 0.00 0.01 0.00 0.00 0.02 0.01 0.01 0.02 0.02

0.27 0.11 0.10 0.09 0.07 0.05 0.01 0.01 0.01 0.01

0.20 0.12 0.09 0.09 0.09 0.08 0.01 0.01 0.01 0.01

0.15 0.12 0.06 0.07 0.11 0.08 0.02 0.01 0.00 0.00

0.13 0.12 0.05 0.06 0.11 0.06 0.03 0.01 0.00 0.00

ESP ESP ESP DU USA ESP DEU ESP ESP DU USA

tb y rlr met eq rir y eq ep oil

Year 0 0.81 0.01 0.02 0.01 0.01 0.01 0.00 0.01 0.00 0.01

Year 1 0.16 0.15 0.13 0.10 0.04 0.02 0.02 0.02 0.02 0.02

Year 2 0.08 0.16 0.11 0.11 0.06 0.03 0.01 0.02 0.01 0.02

Year 5 0.03 0.16 0.07 0.06 0.10 0.03 0.01 0.02 0.01 0.02

Year 10 0.01 0.14 0.06 0.04 0.10 0.03 0.01 0.02 0.01 0.02

USA USA DU USA USA DU DU USA IND CAN

tb rir met y rlr mat oil eq tb y

Year 0 0.88 0.00 0.02 0.00 0.01 0.01 0.05 0.01 0.01 0.00

Year 1 0.40 0.12 0.06 0.05 0.05 0.04 0.02 0.02 0.01 0.01

Year 2 0.21 0.19 0.13 0.03 0.06 0.07 0.02 0.02 0.00 0.00

Year 5 0.10 0.21 0.18 0.01 0.08 0.08 0.01 0.02 0.00 0.00

Year 10 0.08 0.21 0.20 0.01 0.08 0.07 0.01 0.02 0.00 0.00

Note: The table shows the shares of generalized forecast error variance (median based on a bootstrap procedure with 1000 draws) for 8 different country-specific trade balances explained by the 10 most important shocks over time. Shocks are ordered by their contribution to the forecast variance 1 year after the shock. DU stands for the dominant unit model.

statistically significant positive response of approximately 2.8%, trade balances of Canada (-0.8%), China (-0.9%) and Spain (-1.9%) deteriorate.

in several countries. She demonstrates that savings and investments are supposed to be procyclical, but if the decrease in investment dominates the decrease in savings, the trade balance becomes countercyclical and this is commonly the case. Other authors like Mendoza (1991) or Correiaa et al. (1995) present similar findings. Bussière et al. (2012) provide empirical evidence for a significant increase in US imports by approximately 2% after one year, but not for the effect on exports. Hence, we cannot draw any conclusions about the trade balance.

4.2.1. A positive oil price shock The oil price increases by 9.7% due to a positive 1SD shock. As argued earlier, the effect of a positive shock to the oil price on the trade balance is expected to depend upon the country-specific economic structure. The trade balance of oil-exporting countries are supposed to increase with a price rise, while those of oil importing countries deteriorate. Oil-neutral countries, e.g. economies which do not depend on foreign oil resources, should not be affected by a shock. Our results (see Fig. 3) show that a positive oil price shock causes statistically significant improvements in the trade balances of Canada (+0.5%), Saudi Arabia (+5.9%) and the UK (1.1%). These trade balances react contemporaneously to the shock. On the other hand, we find significantly negative responses in France (-0.9%), Germany (-1.0%), Italy (-0.9%), Japan (-4%), South Korea (-1.5%) and the United States (-0.4%). The GIRFs show that the effect of an oil price shock on the trade balance is extremely heterogeneous and that aggregating specific groups of countries would cause misleading results. Even among oilexporting countries, the effects are heterogeneous. The IMF (2006) finds similar results for the current account of the United States, China and an aggregate of oil-exporting countries. For the aggregates with other advanced and developing countries, the current account deteriorates.

4.2. A positive shock to the US’ real equity prices A positive shock to the US real equity prices increases real equity prices by 4.5%. Fig. 2 shows that following a positive equity price shock in the United States, the domestic trade balance deteriorates, but not in a statistically insignificant way. Nevertheless, foreign trade balances show significant responses. The so called international wealth channel (see for example Fratzscher and Straub, 2010 and Holinski and Vermeulen, 2012), may explain the trade balance's response. The equity shock increases household wealth, which translates into an increase in demand. The higher demand causes an increase in imports, which explains the trade deficit. Foreign trade balances may improve due to the increase in US imports. If, however, the spillover effect on foreign stock markets is strong or if foreigners own US equity, the same effect might occur abroad. We find no significant response for the domestic trade balance, but several significant responses abroad. While Saudi Arabia shows a 208

Economic Modelling 64 (2017) 201–210

T. Bettendorf

Table 4 Country-specific model specification.

Table 5 Lag orders and cointegration relationships.

Models

tb

y

req

ep

rsr

ARGENTINA AUSTRALIA AUSTRIA BELGIUM BRAZIL CANADA CHINA CHILE FINLAND FRANCE GERMANY INDIA INDONESIA ITALY JAPAN KOREA MALAYSIA MEXICO NETHERLANDS NORWAY NEW ZEALAND PERU PHILIPPINES SOUTH AFRICA SAUDI ARABIA SINGAPORE SPAIN SWEDEN SWITZERLAND THAILAND TURKEY UNITED KINGDOM USA

x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x

x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x

x x x x

x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x

x x x x x x x x x x x x x x x x x x x x x x x x

x x x x x x x x x x x x x x x x x x x x x x

x x x x x x x x

rlr

x x x x

x x

x x x

x x x

x

x x x

x x

Note: An x denotes that the variable is entering the corresponding model. Empty spaces imply that data for the country-specific variable was not available.

Country

pi

qi

# of CR

ARGENTINA AUSTRALIA AUSTRIA BELGIUM BRAZIL CANADA CHINA CHILE FINLAND FRANCE GERMANY INDIA INDONESIA ITALY JAPAN KOREA MALAYSIA MEXICO NETHERLANDS NORWAY NEW ZEALAND PERU PHILIPPINES SOUTH AFRICA SAUDI ARABIA SINGAPORE SPAIN SWEDEN SWITZERLAND THAILAND TURKEY UNITED KINGDOM USA

2 1 1 2 2 2 2 2 2 2 1 2 1 2 1 2 1 2 1 2 2 2 2 2 2 1 2 1 1 2 1 1 2

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

2 6 (5) 4 1 1 3 1 2 2 2 2 2 2 1 2 3 2 2 3 4 2 2 0 1 2 (1) 2 2 2 2 1 1 2 2

Note: Lag orders are estimated using the AIC. # of cointegration relationships are obtained by the Johansen trace test. Numbers in brackets denote the # of cointegration relationships used in the model in order to ensure model stability.

4.3. Generalized forecast error variance decomposition foreign variables. The Spanish trade balance is also mainly affected by domestic variables. Only US real equity prices and German real GDP are important foreign variables. The trade balance of the United Kingdom is mainly driven by the oil price, domestic real short-term rates, US real GDP. Shares for other variables are relatively low. The statistics show for the United States that shocks to domestic variables and commodity prices are the important drivers. Shares of foreign variables are approximately zero. This result indicates that foreign variables may be of less importance for the US trade deficit and that the deficit is mainly driven by domestic factors. Overall, we observe that besides domestic factors, foreign variables such as US real equity prices, US real GDP and the oil price are important determinants of international trade balances.

In this section, we discuss the GFEVDs of different trade balances. We shall focus on China, Germany, Japan, Spain, the United Kingdom and the United States, because these countries are the world's largest economies and of major importance in the context of external imbalances. Our model consists of 177 equations of endogenous variables, meaning that we obtain the same number of possible determinants for each trade balance. In order to minimize the output, without losing important information, we report the 10 most important determinants, ordered by their contribution to the forecast variance 1 year after the shock. Table 3 displays these 10 variables in the corresponding countries. We find in all cases relatively high percentages for the domestic trade balances on impact, which implies that the trade balance itself explains a lot of its own forecast error variance. But we also observe that the importance of own shocks decays quickly. Holinski and Vermeulen (2012) find a low persistence trade balance shocks, as well. We focus in the following analysis on mainly on foreign variables, as we are particularly interested in spillovers and international effects. We find for China that the US real equity prices and the oil price play a role in the short-run, while the RER and RIR are relatively important in the long-run. In Germany, we see that the trade balance is mainly driven by domestic variables and commodity prices. Foreign variables play a minor role. But Chinese and US real GDP are the most important ones. In Japan, we observe high shares of explained forecast error variance for the real short-term and long-term rates, as well as the oil price. This finding may indicate that the persistently low interest rates in Japan are non-neutral for the high external surplus, as they might have caused capital outflows. Besides US real equity prices, we find that Chinese, US and Saudi Arabian real GDP are important

5. Conclusion This paper has sought to identify the international drivers of trade balances by estimating a Global VAR model for 33 countries using a relatively large set of variables. Our results reveal the joint dynamics of the variables involved, which can illuminate the relevance of certain channels in the emergence of GIs. We assess the importance of specific variables that can be related to theories of GIs for international trade balances using GIRFs and GFEVDs. Our results indicate that a positive shock to the US’ real equity prices has a negative but not statistically significant effect on the domestic trade balance. Nevertheless, several foreign trade balances show statistically significant responses. The sign can be positive or negative, depending on how the shock is transmitted into the foreign country. The GFEVDs underline these findings, implying that US’ equity prices are an important driver for international trade balances. 209

Economic Modelling 64 (2017) 201–210

T. Bettendorf

US’ real GDP is an important determinant itself. This finding is consistent with the literature on IRBC models in which GDP is negatively correlated with the trade balance. This study shows that GIs cannot be explained by one lone theory. This has important implications for policy makers. Policy makers should continuously monitor a broad set of variables and react to undesired developments in such a way that disequilibria do not occur. The study at hand highlights the importance of asset prices in the context of GIs. Policy rates far below adequate levels, for instance, could cause excessive asset price inflation and economic misalignments (see Taylor (2008) and Laibson and Mollerstrom (2010)). If such market developments could be prevented by policy makers, there would be a lower risk regarding the emergence of potentially harmful external imbalances.

The results are in line with those of Fratzscher (2010) and support the hypothesis of international wealth effects (see Laibson and Mollerstrom (2010)). Our results also demonstrate that the oil price was an important driver of trade balance surpluses and deficits for many countries. While trade balances of oil importing countries tend to deteriorate following an oil price shock, trade balances of oil exporting countries tend to improve. These statistics are highly significant in most cases. The results are thus keeping with those of the IMF (2006) and support parts of the saving glut hypothesis postulated by Bernanke (2005). Additionally to the findings of the IMF (2006), we have demonstrated that the responses are not only heterogeneous in sign across countries, but also heterogeneous in size. Apart from conventional theories on GIs, this study implies that the

Appendix: Tables See Tables 4 and 5. Appendix A. Supplementary data Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.econmod.2017.03.033. CERP Discussion Paper, (DP8080). Gourinchas, P.-O., Jeanne, O., 2013. Capital flows to developing countries: the allocation puzzle. Rev. Econ. Stud. 80 (4), 1484–1515. Hoffmann, M., M. Krause, and T. Laubach. 2011. Long-run growth expectations and global imbalances, Deutsche Bundesbank Discussion Paper Series 1: Economic Studies, (01/2011). Holinski, N., Vermeulen, R., 2012. The international wealth effect: a global errorcorrecting analysis. Empir. Econ. 43 (3), 985–1010. IMF, 2005. Global Imbalances: A Saving And Investment Perspective, World Economic Outlook, pp. 91–124. IMF, 2006. Oil Prices and Global Imbalances, World Economic Outlook, pp. 71–96. Karadimitropoulou, A., and León-Ledesma, M.A., 2009. Sources of Current Account Fluctuations in Industrialized Countries, University of Kent, Department of Economics Discussion Papers, (KDPE 0910). Koop, G., Pesaran, M.H., Potter, S., 1996. Impulse response analysis in nonlinear multivariate models. J. Econ. 74 (1), 119–147. Laibson, D., Mollerstrom, J., 2010. Capital flows, consumption booms and asset bubbles: a behavioural alternative to the Savings Glut Hypothesis. Econ. J. 120 (544), 354–374. Lane, P.R., Milesi-Ferretti, G.M., 2012. External adjustment and the global crisis. J. Int. Econ. 88, 252–265. Lee, J., Chinn, M.D., 2006. Current account dynamics and real exchange rate dynamics in G7 countries. J. Int. Money Financ. 25 (2), 257–274. León-Ledesma M.A., Mihailov A. (forthcoming): Advanced International Macroeconomics and Finance Oxford University Press (forthcoming). Mendoza, E.G., 1991. Real business cycles in a small open economy. Am. Econ. Rev. 81 (797-818), 1089–1113. Mendoza, E.G., Quadrini, V., Ríos-Rull, J.-V., 2009. Financial Integration, Financial Development and Global Imbalances. J. Political Econ. 117 (3), 371–416. Pesaran, H.M., Shin, Y., 1998. Generalized Impulse Response Analysis in Linear Multivariate Models. Econ. Lett. 58 (1), 17–29. Pesaran, M.H., Schuermann, T., Weiner, S.M., 2004. Modelling Regional Interdependencies Using a Global Error-Correcting Macroeconometric Model. J. Bus. Econ. Stat. 22 (2), 129–162. Pesaran, M.H., Smith, R., 2006. Macroeconometric Modelling with a Global Perspective 74. Manchester School, University of Manchester, 24–49. Sims, C.A., 1980. Macroeconomics and reality. Econometrica 48 (1), 1–48. Sims, C.A., 1986. Are forecasting models usable for policy analysis? Fed. Reserve Bank Minneap. Q. Rev. 10 (1), 2–16. Smith, V. L., and A. Galesi, 2014. GVAR Toolbox 2.0, 〈https://sites.google.com/site/ gvarmodelling/gvar-toolbox〉. Taylor, J.B., 2008. The Financial Crisis and the Policy Responses: An Empirical Analysis of What Went Wrong, Keynote Address at Bank of Canada. Uhlig, H., 2005. What are the effects of Monetary Policy on Output? Results from an Agnostic Identification Procedure. J. Monet. Econ. 52 (2), 381–419.

References Aizenman, J., Marion, N.P., 2003. The High Demand for International Reserves in the Far East: what's Going On? J. Jpn. Int. Econ. 17 (3), 370–400. Baxter, M., 1995. International Trade and Business Cycles. NBER Work. Pap. Ser. (5025). Bems, R., Dedola, L., Smets, F., 1995. US Imbalances - the role of technology and policy. J. Int. Money Financ. 26 (4), 523–545. Bernanke, B, 2005. The Global Saving Glut and the U.S. Current Account. Remarks at the Sandridge Lecture, in Virginia Association of Economics, Richmond, VA, March 10. Blanchard, O.J., Quah, D., 1989. Are forecasting models usable for policy analysis? Am. Econ. Rev. 79 (4), 655–673. Bracke, T., Bussière, M., Fidora, M., Straub, R., 2010. A Framework for assessing Global Imbalances. World Econ. 33 (9), 1140–1174. Bracke, T., Fidora, M., 2008. Global Liquidity Glut or Global Savings Glut? A Structural VAR Approach. Working Paper Series 911, European Central Bank. Bussière, M., Chudik, A., Sesteri, G., 2012. Modelling Global Trade Flows: results from a GVAR Model. Fed. Reserve Bank Dallas Work. Pap. (119). Caballero, R., Farhi, E., Gourinchas, P.-O., 2008. An equilibrium model of ’Global Imbalances' and low interest rates. Am. Econ. Rev. 98 (1), 258–393. Case, K.E., Quigley, J.M., Shiller, R.J., 2005. Comparing Wealth Effects: the Stock Market versus the Housing Market, Berkeley Program on Housing and Urban Policy. Work. Pap. Ser. (45). Chudik, A., Pesaran, H., 2013. Econometric analysis of high dimensional VARs featuring a dominant unit. Econ. Rev., 32. Correiaa, I., Neves, J.C., Rebelo, S., 1995. Business cycles in a small open economy. Eur. Econ. Rev. 39 (6), 1089–1113. Dees, S., di Mauro, F., Pesaran, M.H., Smith, L.V., 2007. Exploring the International Linkages of the Euro Area: a Global VAR Analysis. J. Appl. Econ. 22 (1), 1–38. Dees, S., Holly, S., Pesaran, M.H., Smith, L.V., 2007. Long run macroeconomic relations in the global economy. Econ. - Open-Access, Open-Assess. E-J. 1, 3. Dooley, M.P., Folkerts-Landau, D., Garber, P., 2003. An essay on the revived Bretton Woods System. NBER Work. Pap. Ser. (9971). Dooley, M.P., Folkerts-Landau, D., Garber, P., 2004. Interest rates, exchange rates, and international adjustment: bw II dynamics. Int. Econ. Econ. Policy 3 (3-4), 293–302. Dooley, M.P., Folkerts-Landau, D., Garber, P., 2009. Bretton Woods II Still Defines The International Monetary System. NBER Work. Pap. Ser. (14731). Engel, C., Rogers, J.H., 2006. The U.S. current account deficit and the expected share of world output. J. Monet. Econ. 53 (5), 1063–1093. Fratzscher, M., Juvenal, L., Sarno, L., 2010. Asset prices, exchange rates and the current account. Eur. Econ. Rev. 54 (5), 643–658. Fratzscher, M., and R. Straub. 2010. Asset Prices, News Shocks and the Current Account,

210