The impact of global liquidity on the Eurozone

(1)

UNIVERSITY OF AMSTERDAM

Master Thesis

MSc. Economics, Specialization International Economics and Globalization

Student Name: David Mischke Student Number: 10604464

Supervisor: Kostas Mavromatis Date: 29.08.2014

Abstract

Global liquidity is the aggregated effect of monetary policy around the world. In this context I analyse the transmission of monetary policy shocks of major adavanced economies and its effects on the Eurozone, using a structural vector autoregression model. The novelty of this research is the

incorporation of market liquidity into the estimation. Impulse response functions show that innovations in global liquidity lead to a money inflow into the Eurozone economy and cause a temporary increase in GDP and inflationary pressures in the future. Market liquidity appears to be

stimulated by increased monetary liquidity.

(2)

1

1 Introduction: Monetary Policy and Global Liquidity

Economists have studied the effect of international monetary policy shocks for a long time. How are monetary policy shocks transmitted internationally? What are the effects on domestic output of a foreign monetary expansion? These and other questions were investigated by previous researchers. In the early 1990’s researchers reconciled the predictions of theoretic models with econometric evidence (Christiano L. , 1994). Since then, a rich body of literature evolved from single country studies, examining the domestic effects of a monetary policy shock, to comprehensive cross country studies.

The aim of this thesis is to investigate the phenomenon of international monetary policy

transmissions in context of an aggregated, global level, which has been termed “global liquidity” (forthcoming). The prime focus is the concept of global liquidity and its effect on the Eurozone. Set against this background, this thesis seeks to integrate the well-established monetary global liquidity analysis with the widely acknowledged but empirically barely implemented aspect of market liquidity into a single estimation. To that end a structural vector autoregression model (SVAR) with two global liquidity variables and five macroeconomic time series identifies monetary liquidity shocks and their effects on the economy of the Eurozone. The estimation shows that global liquidity shocks are transmitted by pushing money into the Eurozone, increasing real GDP while also leading to inflationary pressures in the future.

This thesis is organized as follows: Firstly the concept of global liquidity and international policy transmission will be introduced. Secondly an overview of relevant literature on the topic is presented. Next, the empirical section presents data, econometric methods and estimation results. Lastly, the thesis concludes with a discussion of the findings.

1.1 A recent phenomenon: Global liquidity

Modern economies do not exist in isolation. For their monetary policy the global entirety of world monetary policies and private capital flows matters. Figure 1 gauges the development of the monetary policy stance for selected major economies. It is noticeable that recently the money supply increased considerably over a short time horizon. The way in which these developments aggregate and influence economies around the world has been the subject of a large variety of studies. Therefore, the term global liquidity is used to refer to the combined

(4)

3 effect of monetary policy stances of major economies (Domanski, Fender, & McGuire, 2011). It tries to capture the intricate channel through which the joint monetary policies of central banks around the world interact on a global level. Therefore global liquidity is believed to be an important driver of cross border capital flows and crucial for global financial stability.

Moreover it has very important implications for policymakers around the world. With increasing financial integration, global liquidity conditions

increasingly determine the effective provision of funds in national financial markets. It also plays a crucial role during the buildup of financial bubbles (BIS, 2012). A real-world example thereof provides the policy responses to the 2008 subprime mortgage crisis. In the aftermath of the crisis, monetary authorities tremendously expanded the money supply and lowered policy rates. Hence global liquidity is recently also termed “global excess liquidity”, which reflects the view of the existence of ample cross border liquidity (Domanski, Fender, & McGuire, 2011).

1.2 International transmission mechanisms of monetary policy

In theory there is quite some ambiguity about the transmission mechanisms through which one country’s monetary policy is emitted internationally. Given the multiplicity of possible mechanisms I will focus on those which are extensively discussed in the literature and those that can be empirically

tested for. I will distinguish between financial Figure 1: Nominal money supply of selected economies 0 500 1000 1500 2000 2500 3000 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 US money supply M1 (billions of $)

0 100000 200000 300000 400000 500000 600000 700000 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 Japanese money supply M1 (billions of ¥)

0 1000 2000 3000 4000 5000 6000 1979 1981 1983 1985 1986 1988 1990 1992 1993 1995 1997 1999 2000 2002 2004 2006 2007 2009 2011 2013 Eurozone money supply M1 (billions of €)

(5)

4 and real (trade related) transmission channels.

Pertaining financial transmission channels, it is thought that in the long run the nominal money supply does not affect real variables such as output, while in the short run monetary policy does have an effect on aggregate output. Monetary variables can only have an influence on real variables in the presence of price rigidity. Given the assumption of price rigidity, changes in the money supply alter the relative yields of imperfectly substitutable monetary assets, mostly “money” and “bonds”. Thus, an increase in domestic money supply decreases domestic interest rates and hence increases the ease of financing and boosting aggregate demand (Borio & Zhu, 2012). This negative effect of a monetary expansion on interest rates is termed liquidity effect (Romer, 2012).

The effect of foreign variables on domestic variables is less clear cut. If the foreign country is large and financial markets are integrated, its policy may depreciate world real interest rates and thus stimulate the domestic economy (Kim (2001)).

A related idea in terms of financial transmission mechanisms was postulated by Baks & Kramer (1999): They distinguish a “push” and a “pull” channel. The push channel postulates that excessive foreign money growth gives rise to money outflow of that country with the pursuit of higher yields abroad. Hence money is “pushed” into the domestic economy from abroad. This should in theory coincide with an increase in domestic bond/asset prices (i.e. a decrease in interest rates). As a consequence one should observe a negative correlation between foreign money growth and domestic interest rates. In contrast, the pull channel has an opposite effect. If excessive money growth fuels foreign asset prices, investors might view this as real asset price inflation, leading to higher yields. As a corollary, foreign capital should flow out of the domestic economy with interest rate increases in both countries. The extent to which financial

transmission channels exist depends crucially on the international integration of capital markets. Hence an empirical investigation is more adequate concerning countries with high capital

market integration and no investment and trade restrictions.

Real transmission channels can be traditionally thought of in a Mundell-Fleming-Dornbusch modeling setup. According to this line of thought, an expansionary monetary policy has two opposing effects on the trade balance and output of the foreign country. A foreign monetary

(6)

5 expansion may constitute a beggar-thy-neighbor policy, depreciating the foreign exchange rate and thereby improving the trade balance through cheaper exports, leading to higher income (expenditure switching effect) for the foreign country. In a two country world this would imply an appreciation of the domestic exchange rate and a possible reduction of income through more expensive imports. Due to the higher foreign income, this effect may be offset by a subsequent increase in import demand (income absorption effect). Hence the effect on the domestic trade balance may initially be detrimental but revert itself eventually (Kim, 2001). Another real economy transmission-channel might stem from global inflation dynamics. If an increase in money supply abroad raises foreign expected inflation, domestic import prices might rise (assuming a constant exchange rate) and domestic firms experience an increase in marginal costs due to intra industry trade in intermediate products. This “cost-push” mechanism might fuel domestic inflation and have adverse effects on the output of the domestic country (Rüffer & Stracca, 2006). Moreover the increasing inflation can influence domestic interest rates, as interest rates and (expected) inflation are cointegrated. The extent of such a foreign-driven inflation is also moderated by exchange rate pass through. If foreign exporters pass unfavorable movements of the exchange rate onto domestic consumers, then the cost push channel is likely to be more visible.

The empirical estimation of the effect of foreign monetary policy onto the Eurozone and the identification of transmission channels are the main goals of the empirical section. Thus I propose the following research questions:

What is the effect of a global liquidity shock on Euro area output, inflation, money supply, interest rates and the exchange rate?

Which transmission channels of monetary policy shocks can be identified?

2 Literature Review

2.1 Official and market liquidity

Early research on global liquidity has been done by Baks & Kramer (1999). Already in this pioneering paper the authors make an important distinction between official or monetary

(7)

6

liquidity and private or market liquidity. Liquidity is defined as the (perception of) “ease of

financing”. Likewise, the liquidity of an asset is the degree to which it can quickly be sold in the market.

Official liquidity is created by monetary authorities, denoted in the respective domestic currency. It refers to funding that is available to settle claims through official channels (BIS, 2012), that is, the central bank creating money. The central bank might however be constraint in its operation through the domestic policy framework such as the exchange rate regime. On the international level, (cross border) official liquidity is created through central bank swap lines or IMF facilities (SDR’s). Official liquidity can hence be seen as exogenous, since theoretically it does not depend on the funding conditions in a particular market (Domanski, Fender, & McGuire, 2011).

By contrast market liquidity is created by private institutions, most notably banks. The most important determinant for liquidity is the willingness of banks to provide funding to other financial institutions1 and the real sector. For example banks provide overnight loans to each other on the interbank market or loans for trade credit to the real sector. The ability to lend of private institutions is nonetheless constrained by the legal framework they are operating in such as reserve requirements. Therefore, market liquidity is endogenous, since it is dependent on macroeconomic and financial market conditions. These conditions under which financial intermediaries can finance their operations depend on systemic risk perceptions and counterparty risks. Since major financial intermediaries operate on an international level, private liquidity is a major determinant of international spillovers. In addition private liquidity exhibits a strong procyclicality, since the (perception of) systemic risk determines risk taking behavior, which gives rise to large credit booms and busts (BIS, 2012). In recent years financial innovation, most notably the process of securitization, is believed to have led to a strong increase in (perceived) market liquidity (Domanski, Fender, & McGuire, 2011).

Albeit central banks have the mandate to act independently and to ensure financial stability by interfering in private markets, the actual relationship between monetary and market liquidity might however be more complicated. A contrasting view is that official liquidity is not an enabler of private liquidity but a substitute to market liquidity and hence not exogenously determined. As the recent financial crisis has shown, official liquidity is needed when market

(8)

7 liquidity dries up. Hence during economic hardship, central banks have an important role in providing official liquidity in supporting market liquidity (Baks & Kramer, 1999).

The preceding section already foreshadowed that global liquidity has multiple facets. In fact, several authors emphasize that multiple indicators are potentially useful to assess global liquidity (Eickmeier, Gambacorta, & Hofmann, 2013). Often empirical studies have employed credit aggregates or broad money measures as quantity based measures of official global liquidity. They may be seen as stock variables, indicating outcomes of financing conditions over time. On the other hand price based variables were used to capture current financing conditions. Measures of official liquidity may include major policy rates such as the federal funds rate. Cross border credit volumes of banks and non-banks can be deployed as indicators of market liquidity. A larger amount of cross border credit should be associated with greater liquidity.

There is a widespread notion in the literature that global liquidity is in fact “excessive”. This concept refers to the idea that there exists more liquidity than what would be the equilibrium value at the given interest rate, thus creating a “monetary overhang”. Put differently, the

money supply is larger than what people are willing to hold: Too much money is chasing too few goods. Research points out that such dynamics have caused massive inflation in Russia after the breakdown of the Soviet Union (Kim B.-Y. , 1999). As a consequence the recent massive increase in money supply after the financial crisis of 2008 fueled the fear that excessive liquidity might cause strong inflation in the future.

2.2 Empirical Research

Ample research has been done on monetary policy in open economies. The aforementioned influential study by Baks & Kramer (1999) shows that excess money growth in G7 countries is consistent with higher G7 real stock returns and lower real interest rates (in the other country). In general they obtain stronger results when narrow definitions of money are used as opposed to broad definitions.

Kim (2001) is a prime contributor to open economy structural vector autoregression modeling studies. He investigates the effect of U.S. monetary policy shocks on interest rates, trade balance and output in major industrialized countries. It seems that trade linkages play a minor

(9)

8 role in the transmission channels. A monetary expansion in the U.S. leads to an increase in economic activity in other major advanced countries.

The recent strand of literature revolves around estimating the effects of global liquidity on an array of macroeconomic variables in a variety of settings: Building upon Kim’s (2001) work, Sousa & Zaghini (2008) study the effects of global liquidity shocks on the euro area. They utilize a VAR model with a measure of global liquidity (see Christiano et al. 1998), which is constructed as the sum of non-Eurozone G7 countries’ monetary aggregates. Hence they only consider monetary liquidity but not market liquidity. Analysis via impulse response functions shows that a shock to global liquidity causes a significant temporary increase of Eurozone GDP and the real effective exchange rate, and a permanent increase of EU money, and the consumer price index. Furthermore global liquidity is found to be the prime source of output variability.

A similar study by the ECB discusses several transmission mechanisms of global liquidity on inflationary pressures (Rüffer & Stracca, 2006). A useful extension to standard transmission channels is the consideration of dynamic effects among central banks. If a central bank wants to neutralize foreign beggar-thy-neighbor policies, one should observe a positive correlation among domestic and foreign money aggregates. The authors conduct a VAR analysis similar to Sousa and Zaghini (2008), in which they analyze the effect of a global money shock on output, price level and the exchange rate in the US, Japan and the Euro area. Interestingly, Japanese and European variables react in line with previous results while US variables appear to be less

reactive to global effects.

Another perspective, which has been taken by empirical researchers, is the effect of global liquidity on developing or emerging countries. In this context, studies found, that excess liquidity is in fact a driver of commodity and/or asset price inflation in these countries and

therefore potentially harmful to the economic development of certain countries (see e.g. (Brana, Djigbenou, & Prat, 2012), (Mackowiak, 2005)).

Recently Eickmeier et al. (2013) took a new approach towards measuring global liquidity. The authors emphasize that no single variable can capture global liquidity. Therefore they propose three factors, which constitute global liquidity: global monetary policy, capturing official liquidity, global credit supply and global credit demand, which measure market liquidity. They emphasize that ultimately market liquidity matters for economic activity and hence should be

(10)

9 included in a study of global liquidity. Therefore, explicitly incorporating market liquidity in an open economy framework should improve the quality of such an analysis.

To summarize, global liquidity is an important determinant of macroeconomic variables in open economies: Several studies confirm that domestic inflation and output react significantly to global liquidity shocks. Furthermore, one should differentiate between official and market liquidity when thinking about global liquidity. Empirically, SVAR models have been the econometric tool of choice when analyzing open economy monetary policy.

Since previous authors have acknowledged the importance of market liquidity but not explicitly incorporated such a variable in the empirical estimation. This research is a first attempt to do so. Such an extension of the existing literature should make the investigation more accurate and realistic. Due to data availability and the sake of comparability with previous studies, I will focus on the effect of global liquidity on the Euro area. Moreover the Euro area is an economy which is suitable for an empirical investigation as it is characterized by free movement of capital and labor within itself and other countries.

3 Empirics

3.1 Measurement of global liquidity and data

In order to investigate the effect of global liquidity a measure thereof needs to be constructed. While in theory and practice a variety of indicators may be employed I will, based on the preceding discussion of potential measures, follow most researchers by considering a form of official liquidity deriving from the quantity theory of money. The final measurement of global liquidity will be most suitable for further empirical investigation due to its ease of computation and widespread use in other empirical studies. The quantity theory of money is summarized by: , with M being the total money supply, V being the velocity of money, P the price level and Y being real output (Brana, Djigbenou, & Prat, 2012). By taking the logarithm and differentiating over time the equation can be rewritten as:

̇ ̇ ̇ ̇

(11)

10 growth of money and real output in period t respectively. One simplifying assumption needs to be made before the final measurement for global liquidity is derived: The velocity of money is constant over time. Given the practical difficulties to adequately measure the velocity of money this assumption is convenient and widely applied in previous research. The final measure of global liquidity will therefore be:

̇ ̇ ̇

Therefore I define official (excess) global liquidity in terms of a surplus of money growth over output growth. Since the growth of GDP is subtracted from the growth of money, it is implicitly assumed that only that part of liquidity has international effect, which is not already “used” for GDP growth. For example, if ̇ equals zero for a given period, GDP and the money stock grew at the same rate whilst a positive value indicates that money growth exceeded GDP growth in that period. Most central banks do not try to achieve zero inflation but conduct inflation targeting. In case of the ECB the proclaimed target is to keep inflation close to 2% annually. Hence if

̇ exceeds the target, the growth of monetary liquidity might be judged excessive given a certain value of real GDP growth and velocity of money.

Note that this measure is based on growth differentials and therefore omits potential information incorporated in stock variables. However the intuitive appeal and ease of

computation have made equation a widespread measure of global liquidity. Global liquidity will be computed according to (2) by only using GDP and money supply data of non-Eurozone G5 countries (USA, UK, Canada, Japan), because the aim of this analysis is to investigate how

foreign liquidity affects the Eurozone. Hence the measure of global liquidity only covers that

part of global liquidity which is not created by the Eurozone.

Alternatively the money equation can be rewritten to . Thus, by assuming a constant velocity of money, one should expect a fairly stable ratio of money to GDP if the objective of price stability is maintained. Set against this background, earlier research employed the ratio of “world” (i.e. non-Euro zone G7 countries) monetary base to “world” nominal GDP as measure of global excess liquidity (Brana, Djigbenou, & Prat, 2012). Albeit this measure may carry additional information, by being comprised of stock variables it runs into the practical difficulty of

(12)

11 liquidity is used.

Figure 2 displays ̇ over time. The global liquidity measure averages 0.853% over the whole time horizon with a standard deviation of 0.906%. It is noticeable that during the majority of the time there is an excess money growth over real GDP growth. The recent financial crisis in 2008 seemed to have coincided with a large increase in excess liquidity.

In general the international conversion of economic variables into the same currency by market exchange rates is impeded by their volatility. In order to construct meaningful measures of global money growth, monetary aggregates of the remaining non-Eurozone G7 countries (U.S.A., Japan, United Kingdom and Canada) are converted into Euros with relative PPP exchange rates. Hereto the nominal exchange rate of the first quarter of 1999 (the introduction of the Euro) of the Euro against the several currencies is used as basis. The relative change of the consumer price index is then used to construct the PPP exchange rate based on the period by period changes2.

As a measure of market liquidity the growth rate of international bank claims to non-bank institutions will be used, as it is proposed by the Bank of International Settlements3. This credit aggregate is particularly useful since it captures the international dimension of integrated financial markets. Moreover, credit is a good indicator of market liquidity since it stands at the end of the financial intermediation chain (Domanski, Fender, & McGuire, 2011). Since an increase in official global liquidity should in principle result in easier funding conditions in credit markets, a positive effect of the official global liquidity index and international bank claims is expected. Figure 3 shows the growth rate of these bank claims over time. As aforementioned, a problem with quantity based measures of market liquidity is that they do not indicate market conditions as such but outcomes over time. Therefore this measure of market liquidity might be

Figure 2: official global liquidity, GL = excess money growth over real GDP growth in %

-2 0 2 4 6 GL 1980q1 1985q1 1990q1 1995q1 2000q1 2005q1 2010q1 2015q1 time

(13)

12 somewhat noisy, since the growth in

credit at a given point in time is likely to be (partially) determined by funding conditions in previous and expectations of future periods and not just in the same period. Another shortcoming of this measure is that it only captures cross border positions of banking claims but not domestic credit. This relates to the idea that the measure of market

liquidity should be global (i.e. not Eurozone-based). Generally there is a certain bias to opt for domestic sources of funding (home bias). Thus this measure represents only a fraction of the total credit market in the world. However, the credit markets of the G5 countries are highly integrated and I expect the variation in the variable to be sufficiently descriptive to delineate the behavior of market liquidity.

The other variables in the estimation are real GDP (GDP), the consumer price index (CPI), money stock M2 (M), the short term interest rate (int) and the real effective exchange rate (XR) of the Eurozone. The choice of these variables was largely dictated by data availability and previous research, as these variables are well established in the empirical modelling of monetary policy shocks (see for example Smets & Peersman (2001)). The quarterly data ranges from Q3 1980 until Q3 2013. GDP, money stock and CPI is seasonally adjusted. More detailed data sources and definitions see the table in the appendix.

3.2 Model

The empirical estimation is based on a structural vector auto regression method. This modelling approach has become a widely used methodology in empirical macroeconomics. The technique is built upon multiple equations, whose time path is affected by current and past realizations of other variables in the system (Enders, 2003). Unfortunately, simple VAR models suffer from the drawback that a structural shock cannot be distinguished from a composite shock if variables

-1 0 0 10 20 30 ML 1980q1 1985q1 1990q1 1995q1 2000q1 2005q1 2010q1 2015q1 time

Figure 3 Global Market liquidity, measured as growth in international bank claims in %

(14)

13 share too many contemporaneous relationships. For example in a two equation VAR in which both variables affect each other in the same period, a shock to one variable creates a feedback onto itself through the second variable, making it impossible to distinguish a structural (i.e. a “pure” shock to only that variable) shock of that variables from the composite shock, comprised of the initial true shock and the contemporaneous feedback effect. Therefore shocks to the system do not carry a meaningful economic interpretation if structural shocks cannot be identified. More formally said, innovations have no structural interpretation due to the

underlying correlation of residuals across estimation equations (Krolzig, 2003). For forecasting purposes this does not pose a problem however it is inappropriate for economic analysis, because it is impossible to distinguish true structural shocks and forecast errors from one another.

One way to identify pure structural shocks in the system is by using a recursive structure of the contemporaneous relationships of the variables. This method is known as Choleski

decomposition. By imposing a lower triangular contemporaneous relationship on the system of equations, exact identification can be achieved (see below). By doing so, one makes essentially strong assumptions about the underlying structural errors. Consider again the two-equation VAR but with only one simultaneous relationship. A shock to the first variable is interpreted as a pure structural shock while a shock to the second variable is a composite shock of the

contemporaneous feedback effect from the first variable and its own structural shock (Enders, 2003). A major caveat of the recursive structure is that it often does not fit economic theory. Since the system is recursive, the contemporaneous relationship of e.g. money and interest rates cannot be dynamic (i.e. both variables affect each other in the same period).

One way to deal with this issue of “incredible identifying restrictions” is proposed by Sims (1980). The restrictions on the simultaneous relationships can be generalized towards an economic structure. With this identification scheme each equation in the model is based on economic theory or previously derived stylized facts about variable relationships. Consequently, the assumption that a shock to one variable does not affect another variable within the same period must be justified by economic theory. This is the distinctive advantage the structural model has over the recursive structure, since only then the system of equation models the structural (i.e. economic) interactions of the variables. For example a recursive scheme forbids a

(15)

14 contemporaneous effect of the interest rate after an exchange rate shock (assuming the

interest rate is ordered before the exchange rate), which is contradictory to the uncovered interest parity condition (Kim & Roubini, 2000). Moreover structural schemes allow for a better modeling of the monetary authority’s preferences, since the money supply can be modelled as reacting to variables before and after its own ordering.

Formally a SVAR model can be written in the following way:

∑

In an n-equation system, A is the matrix of contemporaneous relations, being the vector of endogenous variables and the vector of structural innovations. are

assumed to be pure white noise disturbances ( ) and serially uncorrelated. B0 is the vector of constants and B1 the matrix of coefficients of the lagged terms. Lastly Dt are vectors of deterministic terms such as a linear trend and dummy variables. The SVAR methodology allows to impose restrictions on the contemporaneous relationship of the variables i.e. matrix A such that it maps the structural relationships among the variables. Thus an entry ank in A shows the contemporaneous effect of a shock to variable n on variable k. By pre-multiplying with A-1 we obtain:

∑

With , , and . This is the system in reduced form

with the composite error term et. The variance/covariance matrix of the error terms, Σe, is diagonal and variances are assumed to be constant ( ). This matrix has n(n+1)/2 distinct entries, while the A has n2 different entries. Since this poses an identification problem. In order to identify the n2 unknown elements of A from the entries of Σe it is required to impose restrictions on the A matrix to achieve exact identification. In Choleski decomposition the system is always exactly identified, while in a structural model the system might be overidentified with more restrictions than required. When

(16)

15 enough restrictions on A have been placed, the reduced form can be estimated and structural innovations can be recovered from the composite error terms.

The estimation is based on the following identification scheme:

[ ] [ ]

This identification scheme largely follows previous research with the additional consideration of market liquidity. The first two equations represent the global liquidity dynamics captured by official and market liquidity. The first variable is the measure of official global liquidity (GL). As argued before, official liquidity is assumed to be the most exogenous variable, since it is set by central banks. It shares no contemporaneous effect with any other variable. Since GL is global liquidity created by the remaining G5 countries, I assume that non-ECB central banks react only with a delay to Euro variables.

ML is global market liquidity as defined in the data section. It is affected by official liquidity and Euro area money supply within the same period. This assumption captures the notion that internationally operating banks can be affected by either official global liquidity outside the Euro area and by money supply increases by the ECB.

The next two equations represent the goods market. GDP is the Euro area real gross domestic product4 and CPI is the Euro area consumer price index. GDP is assumed to react to global market liquidity contemporaneously but not to official liquidity. This assumption is based on the fact that ultimately credit market conditions matter for economic growth but not official money supply. Since official money supply increases first need to be converted into loanable funds, global liquidity cannot affect GDP immediately. Moreover I assume that can only GDP react to financial variables other than credit (since credit is the immediate source of funding) with a delay, because upon unexpected changes to financial variables firms do not adjust output and prices immediately due to menu costs. For the same reason, the price level is assumed to react

(17)

16 only sluggishly to other variables, except for the domestic money supply. Increases in domestic money supply can raise inflation expectations and thereby bring price level increases forward in time (Romer, 2012).

The variables M for money supply and int for the short term interest rate represent the money market. Note that M represents the money demand equation, depending on current

realizations of global liquidity, the short term interest rate, income and the price level. The short term interest rate is assumed to specify the central bank’s reaction function. The central bank can observe official global liquidity and domestic financial variables within the same period but reliable information on output, prices and market liquidity are not available within the same period (Sousa & Zaghini, 2008). With this specification I assume that the ECB can readily observe the level of global liquidity but information on market liquidity is subject to informational delays, since this kind of data comes from private institutions around the world and needs time to be processed by the ECB.

The last equation (XR) represents the foreign exchange market with XR being the real effective exchange rate of the euro against 12 trading partners. It is the most reactive of all variables, because the exchange rate is typically very erratic and responds to all other economic variables. Two versions of the model are estimated: Model 1, covering data from second quarter of 1980 until the second quarter 2008 (n=110). Model 2 spans the whole estimation period until 2013 quarter two (n=133). The former represents a model based on pre-crisis data while the latter incorporates the time period of the subprime mortgage crisis and the European sovereign debt crisis. A linear time trend and in the second model a dummy variable for the 2008 financial crisis and the European sovereign debt crisis are added5. Based on eliminating autocorrelation in the residuals, a lag length of 2 is chosen in model 1 and a lag length of 6 in model 2. All variables except for official and market liquidity are expressed in levels and converted into their natural logarithm. This is motivated by more stable behaviour of the impulse response functions. Furthermore Sims (1980) argues that differencing variables leads to great loss of information and long run relationships are harder to identify. The model is estimated by maximum likelihood with the JMulTi6 software.

(18)

17

3.3 Estimation results

After the estimation impulse response functions are computed from the results. These describe the path of a given variable over time after it has been hit by an exogenous shock to another variable in the system7 (Enders, 2003). The size of the global liquidity (GL) shock is one standard deviation of 0.82% in model 1 and 0.90% in model 2. The time horizon is chosen to be 16

quarters.

Concerning model 1 (Figure 4), real GDP exhibits a hump shaped trajectory, peaking at eight periods and returning to normal levels in the longer run. After a slightly negative reaction, the consumer price index increases after 2 years and appears to be rising in the long run after a global liquidity shock. Both Euro area money supply and interest rate show a significant

persistent positive reaction. The exchange rate appreciates significantly upon impact and seems to remain on an appreciated level also in the long run. Market liquidity exhibits a positive response after two quarters and returns to zero in the long run.

Regarding model 2 (Figure 5), it is noticeable that GDP has a very similar path as in the first model. In contrast to model 1 however, the drop in the price level is much more pronounced in model 2 and no significant positive reaction can be observed over the four year horizon. The most noticeable difference to model 1 occurs in the money market. Contrasting to model 1 the Euro area money supply appears to be insensitive to global liquidity shocks. No significant reaction over 4 years can be observed. The short term interest rate behaves quite differently when incorporating post-crisis data. It initially exhibits a previously not observed significant negative response until this trend reverses itself and eventually leads to a positive reaction after two years. The appreciation of the exchange rate seems to be only short lived. It appreciates upon impact but returns to zero already after 3 quarters. Market liquidity exhibits the same hump-shaped response as in model 1.

Lastly in both models the effect of the shock on official global liquidity itself decays fairly quickly, becoming insignificant after 4 quarters.

(19)

18 The LM test for autocorrelation of the residuals is conducted to check the validity of the two proposed models. This test checks the aforementioned assumption that residuals of the models are not autocorrelated. The test rejects serial correlation in both models at the 95% level, suggesting no bias in the residuals. For details see appendix.

-0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 2 4 6 8 10 12 14 16

GL

-0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0 2 4 6 8 10 12 14 16

ML

-1.00E-03 -5.00E-04 0.00E+00 5.00E-04 1.00E-03 1.50E-03 2.00E-03 2.50E-03 3.00E-03 3.50E-03 4.00E-03 4.50E-03 0 2 4 6 8 10 12 14 16

GDP

-0.002 -0.0015 -0.001 -0.0005 0 0.0005 0.001 0.0015 0.002 0.0025 0.003 0 2 4 6 8 10 12 14 16

CPI

(20)

19 -0.002 0 0.002 0.004 0.006 0.008 0 2 4 6 8 10 12 14 16

M

-0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1 0 2 4 6 8 10 12 14 16

int

0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0 2 4 6 8 10 12 14 16

XR

-0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 2 4 6 8 10 12 14 16

GL

-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 0 2 4 6 8 10 12 14 16

ML

Figure 4 Impulse Response Functions Model 1

(21)

20 -0.006 -0.004 -0.002 0 0.002 0.004 0.006 0.008 0 2 4 6 8 10 12 14 16

M

-0.003 -0.002 -0.001 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0 2 4 6 8 10 12 14 16

GDP

-0.005 -0.004 -0.003 -0.002 -0.001 0 0.001 0.002 0.003 0 2 4 6 8 10 12 14 16

CPI

-0.15 -0.1 -0.05 0 0.05 0.1 0.15 0 2 4 6 8 10 12 14 16

int

-0.02 -0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02 0.025 0 2 4 6 8 10 12 14 16

XR

(22)

21

3.4 Estimation Interpretation

The models show that the Eurozone is indeed affected by global liquidity. However due to the two different estimation periods, the results of the two models differ quite a bit. While model 1 is largely in line with results of earlier research, model 2 delineates a different picture.

3.4.1 Model 1:

The model predicts significant rise in real GDP after a global liquidity shock. Hence I find that the assumption of nominal rigidity holds, making GDP reactive to the global liquidity variable. The growth in money (M) and the rising price level (CPI) reflect the inflow of money into the Eurozone associated with the global liquidity growth. Due to nominal rigidity, the price level is not immediately rising as money flows into the economy. These findings are strong evidence in favour of the “push” channel, whereby foreign money is pushed into the Eurozone economy. The rise (i.e. appreciation) of the exchange rate further supports this notion of money flowing into the economy. Despite the appreciation of the exchange rate, the price level is rising. This might be attributed to the cost-push channel through which inflation is “imported” via price level increases in foreign countries. If global excess liquidity raises inflation in non-Eurozone countries, it is transmitted via intra industry trade and hence also the price level in the Eurozone is affected. As expected, the increase in market liquidity indicates that an official liquidity shock raises international bank lending and stimulates economic activity. Somewhat contradictory to the push-channel is the increase in the interest rate, since money inflow should decrease the domestic interest rate. However the rise in the interest rate can be interpreted as the central bank’s reaction to increase the interest rate in response the increase in the price level in order to counteract unwanted inflation (Sousa & Zaghini, 2008).

3.4.2 Model 2:

The inclusion of post-crisis data yields some surprising results. Even with the inclusion of dummy variables for the subprime mortgage crisis and the European sovereign debt crisis the mechanics of global liquidity are quite different from the pre-crisis period.

There is evidence in favour of a much weaker push-channel or money inflow into the Eurozone, since domestic money responds only insignificantly to a global liquidity shock. The behaviour of

(23)

22 the exchange rate supports this interpretation as it only shows a slight appreciation in the first quarter with a quick reversion to non-significance. One possible explanation is that the only short lived appreciation is the immediate reaction of the exchange market to restore

equilibrium. However, the lack of the persistent appreciation if the Euro indicates that there is less demand for Euros in the foreign exchange market. Hence less money is flowing into the Eurozone.

The interest rate in model 2 behaves in stark contrast to model 1. While formerly the increase in interest rates can be attributed to the raising price level, the negative response of the interest rate can now be attributed to the falling price level. In that respect it confirms the earlier interpretation that the interest rate behaviour is driven by inflation dynamics. This leads to the conjecture that the causal chain runs from GL → global inflation → Eurozone inflation → Eurozone interest rate. To further investigate this proposition a variable for global inflation might be incorporated into the SVAR model (see also Kim & Roubini (2000)).

Market liquidity exhibits a similar trajectory in both models. In so far this is good news for policy makers, since it seems that also in the post crisis period official and market liquidity share a positive relationship in the sense that the central bank can stimulate credit growth and thus possibly economic activity.

3.5 Forecast Error Variance Decomposition

In multivariate time series analysis, forecast error variance decomposition (FEVD) can be used to determine how much information each variable contributes to other variables in the regression (Enders, 2003). It shows the contribution of each forecast error variance, which can be

attributed to exogenous shocks to another variable. In the context of the analysis at hand, one can compute the contribution a shock of e.g. GL has on the variability of the interest rate in the Eurozone over time. Table 2 shows the FEVD for model 1 and 2. The time intervals are chosen to be years, averaged from quarters.

(24)

23

variation in GL GL ML GDP CPI M int XR

1 yr 0.919 0.024 0.000 0.000 0.036 0.014 0.006

2 yrs 0.807 0.072 0.002 0.001 0.053 0.049 0.016

3 yrs 0.780 0.074 0.004 0.001 0.072 0.052 0.017

4 yrs 0.743 0.075 0.004 0.001 0.104 0.056 0.017

variation in ML GL ML GDP CPI M int XR

1 yr 0.025 0.910 0.000 0.001 0.020 0.025 0.019

2 yrs 0.067 0.777 0.002 0.005 0.040 0.057 0.053

3 yrs 0.081 0.646 0.015 0.011 0.072 0.059 0.115

4 yrs 0.076 0.571 0.037 0.012 0.091 0.087 0.125

variation in GDP GL ML GDP CPI M int XR

1 yr 0.005 0.019 0.957 0.001 0.004 0.010 0.002

2 yrs 0.053 0.088 0.734 0.001 0.004 0.117 0.003

3 yrs 0.122 0.148 0.556 0.001 0.004 0.162 0.007

4 yrs 0.158 0.180 0.492 0.002 0.008 0.151 0.009

variation in CPI GL ML GDP CPI M int XR

1 yr 0.016 0.004 0.001 0.927 0.016 0.022 0.014

2 yrs 0.051 0.004 0.019 0.808 0.078 0.016 0.023

3 yrs 0.044 0.012 0.055 0.597 0.192 0.066 0.034

4 yrs 0.036 0.055 0.071 0.402 0.270 0.141 0.025

variation in M GL ML GDP CPI M int XR

1 yr 0.034 0.098 0.006 0.014 0.215 0.355 0.278

2 yrs 0.102 0.185 0.018 0.003 0.311 0.192 0.189

3 yrs 0.118 0.235 0.023 0.003 0.367 0.117 0.137

4 yrs 0.131 0.256 0.031 0.006 0.360 0.086 0.131

variation in int GL ML GDP CPI M int XR

1 yr 0.007 0.015 0.050 0.009 0.601 0.176 0.142

2 yrs 0.018 0.074 0.107 0.008 0.556 0.148 0.091

3 yrs 0.083 0.156 0.079 0.010 0.389 0.208 0.075

4 yrs 0.144 0.224 0.053 0.010 0.311 0.177 0.081

variation in XR GL ML GDP CPI M int XR

1 yr 0.031 0.028 0.006 0.000 0.001 0.527 0.407

2 yrs 0.045 0.051 0.014 0.001 0.002 0.382 0.505

3 yrs 0.065 0.081 0.033 0.004 0.022 0.293 0.501

4 yrs 0.079 0.109 0.053 0.007 0.066 0.234 0.452

variation in GL GL ML GDP CPI M int XR

1 yr 0.912 0.012 0.021 0.027 0.011 0.013 0.004

2 yrs 0.747 0.026 0.065 0.050 0.062 0.033 0.017

3 yrs 0.649 0.049 0.066 0.059 0.091 0.057 0.029

4 yrs 0.567 0.119 0.087 0.057 0.084 0.057 0.030

variation in ML GL ML GDP CPI M int XR

1 yr 0.022 0.893 0.037 0.003 0.018 0.017 0.010

2 yrs 0.059 0.667 0.068 0.013 0.116 0.052 0.026

3 yrs 0.190 0.486 0.092 0.029 0.079 0.032 0.094

4 yrs 0.194 0.427 0.139 0.032 0.067 0.028 0.113

(25)

24

variation in GDP GL ML GDP CPI M int XR

1 yr 0.019 0.148 0.822 0.004 0.004 0.002 0.002

2 yrs 0.020 0.177 0.700 0.053 0.025 0.014 0.010

3 yrs 0.114 0.144 0.537 0.091 0.041 0.031 0.042

4 yrs 0.229 0.151 0.409 0.076 0.034 0.026 0.077

variation in CPI GL ML GDP CPI M int XR

1 yr 0.023 0.045 0.076 0.025 0.737 0.025 0.068

2 yrs 0.159 0.172 0.169 0.017 0.388 0.011 0.085

3 yrs 0.196 0.208 0.258 0.046 0.217 0.007 0.069

4 yrs 0.158 0.179 0.307 0.123 0.166 0.009 0.058

variation in M GL ML GDP CPI M int XR

1 yr 0.003 0.036 0.002 0.951 0.004 0.003 0.001

2 yrs 0.009 0.142 0.005 0.819 0.011 0.011 0.003

3 yrs 0.011 0.171 0.003 0.761 0.037 0.011 0.005

4 yrs 0.023 0.113 0.022 0.765 0.052 0.007 0.017

variation in int GL ML GDP CPI M int XR

1 yr 0.218 0.094 0.064 0.012 0.029 0.114 0.469

2 yrs 0.182 0.271 0.192 0.066 0.024 0.036 0.230

3 yrs 0.160 0.226 0.159 0.212 0.042 0.042 0.158

4 yrs 0.225 0.214 0.123 0.224 0.033 0.038 0.143

variation in XR GL ML GDP CPI M int XR

1 yr 0.057 0.005 0.007 0.006 0.007 0.725 0.193

2 yrs 0.041 0.083 0.019 0.044 0.008 0.634 0.171

3 yrs 0.048 0.136 0.022 0.187 0.017 0.462 0.127

4 yrs 0.042 0.111 0.040 0.332 0.029 0.347 0.100

The tables confirm that global liquidity is a large determinant of the variation in GDP, money and the interest rate. In fact official and market liquidity are the two strongest sources of variation in GDP except the variable itself. A comparison between the two models yields some interesting observations: In model 2 the contribution of the interest rate to GDP variability seems to be greatly increased. Possibly the inclusion of unrecorded low nominal interest rates and the uncertainty about the future trajectory of the economy in the aftermath of the crises are responsible for GDP variation. Official liquidity seems to be a greater cause of variation in interest rates in model 2. One possible explanation is that the large increases in money supply around the world and the associated surge in global liquidity as a countermeasure to the financial crises have contributed to interest rate variability. In sum the comparison of the two models suggests that in recent years global liquidity has become a stronger source of variability of macroeconomic variables.

(26)

25

3.6 Discussion

In this section the results are discussed and how they fit in existing literature and economic theory. Thereby I will mainly refer to model 1 (pre-crisis), as the results are more comparable with previous studies and not marred by the crises.

The analysis strongly points in the direction of the Eurozone experiencing an inflow of money following a global liquidity shock. The results are very similar to what previous researchers have found (Sousa & Zaghini (2008), Rüffer & Stracca (2006)).

The empirical findings suggest that multiple, partly opposite effects are at work. The rise in price level and money supply clearly show the inflow of money. Thus I confirm the existence of the push-channel as a form of international monetary policy transmission channel. While in a standard textbook, the upshot of an increase in money supply is a decrease in the interest rate, the estimation predicts the opposite. This is a potential threat to the economy as credit

becomes more expensive and economic activity is likely to slow down. Furthermore the appreciation of the exchange rate makes European goods more expensive and is thought to further depress the economy. Surprisingly, real GDP is not decreasing after a global liquidity shock. These seemingly contradictory findings can be reconciled by considering more advanced, micro-founded economic models, which allow for intertemporal expenditure switching and feature more consumption driven exchange rate dynamics (Obstfeld & Rogoff, 1995). Forward looking agents will anticipate a future price level increase and substitute consumption forward in time, as later on the rising price level lowers their real income. Therefore real GDP rises as a consequence of this consumption driven stimulus. Moreover the increase in consumption and the exchange rate might thus share a dynamic relationship as an appreciation caused by the money inflow triggers the intertemporal substitution, which in turn raises consumption and GDP, which increase import demand, appreciating the exchange rate further. Future research could investigate this conjecture by incorporating a variable for consumption into the estimation and see whether the mechanism proclaimed here can be found.

The effects of such intertemporal substitutions are profoundly linked to the cost push

transmission channel. Since this channel postulates the transmission of (expected) inflation to the Eurozone and consumption is substituted forward in time, the cost push channel can be seen as the reason for the increase in real GDP.

(27)

26 Empirical studies about global liquidity and monetary policy shocks are certainly no novelty. The contribution to the literature of this thesis is twofold:

Firstly, while the dichotomy of official and market liquidity is widely recognized there was no empirical study explicitly incorporating market liquidity to be found. The findings suggest that official liquidity indeed stimulates market liquidity since a shock the former induces a significant positive reaction to the latter. Surprisingly the effect is much stronger in the post-crisis model, possibly due to “hunt for yield” dynamics by which multinational banks engage in increased cross border lending because of very low nominal interest rates in advanced countries. Set against this background the notion of official liquidity as enabler of market liquidity is confirmed. Therefore, the provision of official liquidity could potentially counteract the adverse effects of a drought in market liquidity (see also (Eickmeier, Gambacorta, & Hofmann, 2013)).

Secondly, the results from model 2 incorporate data after the great crisis of 2008. While the general dynamics, which drive the system, seem to be similar there are some important differences. Most notably, the initial drop of the price level is much more pronounced. It is certainly debatable whether one should combine pre- and post-crisis data. In the post-crisis era we observed unseen amounts of global liquidity coupled with an array of non-standard

monetary policy instruments (such as the purchase of troubled assets by central banks). Hence there might be fundamentally different dynamics at work in a world where these policies are in place and the nominal interest is close to hitting the lower bound. However this period is also part of history and it is certainly interesting to see how empirical methods such as the SVAR model perform during such unusual times.

As much as any empirical study in economics, also this thesis is naturally subject to a number of limitations. Firstly, the measurement of global liquidity, although a widely utilized one due to its intuitive appeal and ease of computation, stems from the quantity theory of money, which is a fairly simplistic way of describing money, as it is not micro-founded. Moreover the notion of “excess” liquidity is hard to construct from that theory as it does not incorporate any notion of equilibrium level of money supply, thus making it hard to term a certain level of global liquidity as “excessive” (i.e. more than some equilibrium value). Next, the measure of global liquidity is not truly global, because it only incorporates money supply of G5 countries. Nowadays

(28)

27 emerging economies such as the BRICS8 countries play an important role in the global economy. Not incorporating these countries was mainly determined by the limited amount of data for these countries. Hence future studies should definitely try to incorporate these countries in the study of global liquidity as much as data availability allows.

Secondly the measure of market liquidity is somewhat crude, as it only captures cross border credit to nonbank institutions. There are however many more sources of funding for firms other than international credit (domestic credit, bonds, stocks…). Therefore this study is only a first attempt to look at the interaction of official and market liquidity. Future studies can explore the dynamics of official global liquidity and other markets and possibly utilize a variety of different market liquidity measures (for a summary see BIS, 2012).

Lastly the estimation method appeared to be somewhat sensitive to specification of

contemporary restrictions. Therefore I largely followed the identification method of previous SVAR studies, which delivered reasonable results.

4. Conclusion

In recent years major economies have expanded their money supply tremendously over a short period of time. Economists have termed the phenomenon of aggregated money supplies on a global scale global liquidity. Already early on, it was acknowledged that there exists an

important dichotomy of distinguishing official global liquidity, which refers to the creation of money by the central bank, and market liquidity, which is the degree to which sources of

financing can be quickly bought and sold in a financial market. This distinction is important since market liquidity ultimately matters for economic activity and hence the central bank can only indirectly influence the economy by tighten or loosen monetary policy and thereby affect market liquidity. Since global liquidity is mostly created outside of any single country, it is particularly interesting to see how shocks of global liquidity are transmitted internationally. A vast body of empirical literature evolved around the issue of global liquidity, originating from the study of such international transmissions of monetary policy shocks.

In this paper a measure of global liquidity is used, which defines global liquidity as the difference of money growth and real GDP growth of G7 countries. While this measure is subject to

(29)

28 substantial simplifications, its intuitive appeal and ease of computing have made it a widely used measure in empirical studies. It appears that during most periods since 1980 there was a positive difference of money growth over GDP growth. The financial crisis of 2008 appears to have caused a massive increase in global liquidity.

In order to empirically assess the effect of global liquidity on the Eurozone, a SVAR model with official global- and market liquidity and other macroeconomic variables is estimated. A positive shock to global liquidity appears to push money into the Eurozone economy, visible by a rise in money and the price level and an appreciation of the Euro. The interest rate increases as a result of its cointegrated relationship with expected inflation. Furthermore global liquidity raises output by triggering an intertemporal substitution effect caused by the raising price level. When including post-crisis data into the sample it appears that the push channel is weakened while a positive reaction of real GDP to the shock is maintained. Concerning the interaction of official and market liquidity, it appears that a shock to official liquidity increases cross border lending. Set against this background, the analysis shows, despite an appreciation of the exchange rate, that foreign money supply increases do not constitute a beggar-thy-neighbour but rather a

prosper-thy-neighbour policy by increasing lending and GDP. Nevertheless the ECB should be

aware that foreign money supply increases may lead to unwanted future inflationary pressures in the Eurozone.

Acknowledgements

I want to thank A. Zaghini for the provision of a dataset, which was of great help for constructing my own dataset for the econometric analysis.

(30)

29

Appendix

The table provides an overview of the raw data, which has been used in the empirical section. All data has been gathered in April 2014.

Variable Definition Source

GDP, GDP deflator Nominal GDP (current prices) GDP deflator base year = 2010

OECD Main Economic Indicators

Money aggregates Euro Zone: M2 USA: M2 Japan: M2 UK: M4 Canada: M2 ECB Federal Reserve Bank of Japan Bank of England Federal Reserve

GDP deflator, CPI CPI (HICP) base year = 2010 OECD Main Economic Indicators, ECB

Interest rate Before 1999: 3 month interbank rate, thereafter three month EURIBOR

ECB

Exchange rate Real effective exchange rate of the Euro zone against 12 trading partners

ECB

Global liquidity Market liquidity: international (non)bank lending, all BIS reporting banks: cross border credit and loans in foreign currency

BIS – Global Liquidity indicators

Model Checking

LM test for autocorrelation of residuals H0 = no autocorrelation at lag j (Johansen, 1995) Model 1 Χ² p-value Lag 1 63.82 0.07 Lag 2 56.31 0.22 Lag 3 36.28 0.91 Model 2 Lag 1 59.33 0.15 Lag 2 55.66 0.24 Lag 3 55.00 0.26

(31)

30

Bibliography

Baks, K., & Kramer, C. (1999). Global liquidity and asset prices: Measurement, Implications, and Spillovers.

IMF Working Paper.

Belke, A., Bordon, I., & Volz, U. (2012). Effects of Global Liquidity on Commodity and Food Prices. World

Development.

Berger, H., & Harjes, T. (2009). Does Global Liquidity Matter for Monetary Policy in the Euro Area?

International Finance 12:1, 33-55.

BIS. (2012). Global Liquidity - Concept, measurement, policy implications.

Borio, C., & Zhu, H. (2012). Capital regulation, risk-taking and monetary policy: A missing link in the transmission channel? Journal of Financial Stability.

Brana, S., Djigbenou, M., & Prat, S. (2012). Global excess liquidity and asset prices in emerging countries: A PVAR approach. Emerging Markets Review.

Christiano, L. (1994). Modeling the liquidity effect of a monetary shock.

Christiano, L., Eichenbaum, M., & Evans, C. (1998). Monetary Policy Shocks: What have we learned and to what end?

Domanski, D., Fender, I., & McGuire, P. (2011). Assesing Global Liquidity. BIS Quaterly Review.

Eickmeier, S., Gambacorta, L., & Hofmann, B. (2013). Understaning Global Liquidity. BIS working paper. Enders, W. (2003). Applied Econometric Time Series.

Galati, G., & Moessner, R. (2012). Macroprudential Policy - A Litearture Review. Journal of Economic

Surveys.

Johansen, S. (1995). Likelihood based inference in conintegrated vector autoregression models. Kim, B.-Y. (1999). The Income, Savings and Monetary Overhang in Soviet Households. Journal of

Comparative Economics.

Kim, S. (2001). International transmissions of U.S. monetary policy shocks: evidence from VAR's. Journal

of monetary economics.

Kim, S., & Roubini, N. (2000). Exchange rate anomalities in industrial countries: A solution with a structural VAR approach. Journal of Monetary Economics.

Krolzig, H. (2003). General to Specific Model Melection Procedures from Structural Vector Auto Regressions. Oxford Bulletin of Economics and Statistics.

(32)

31 Mackowiak, B. (2005). What does the Bank of Japan do to East Asia? Journal of International Economics. Mishkin, F. (2001). Monetary Policy Strategy: Lessons from the Crisis. NBER Working Papers.

Obstfeld, M., & Rogoff, K. (1995). Exchange rate dynamics redux. Romer, D. (2012). Advanced Macroeconomics. McGraw Hill.

Rüffer, R., & Stracca, L. (2006). What is Global Liquidity and does it matter? ECB Working Paper. Sims, C. (1980). Macroeconomics and Reality. Econometrica.

Smets, G., & Peersman, F. (2001). The monetary transmission mechanism in th Euro area: More evidence

from VAR analysis.

Sousa, J., & Zaghini, A. (2008). Monetary policy shocks in the euro area and global liquidity spillovers.

International Journal of Finance and Economics.

1

Financial intermediaries are banks but also shadow banks such as money market mutual funds, investment banks, hedge funds etc.

2

The PPP exchange rate at time t is formally constructed as ( ) (

) with being

the price level of the euro at time t and being the price level of country i at time t. This does not guarantee that absolute PPP holds, however for the issue at hand not the level of the exchange rate is relevant but only its changes over time.

3

The BIS publishes global liquidity statistics on its website: http://www.bis.org/statistics/gli.htm (accessed August 2014)

4

Real GDP in 2010-prices, deflated by GDP deflator 5

The mortgage crisis dummy has a value of 1 from 2008 quarter 2 until 2012 quarter 3. The European sovereign debt crisis dummy has a value of 1 from fourth quarter 2009 until the end of the estimation period.

6

Applied Time Series Econometrics, Lütkepohl, H. and Krätzig, M. Cambridge University Press, 2004 7

For detailed explanation see Enders (2003)

8