International spillovers of U.S quantitative easing on emerging market economies

INTERNATIONAL SPILLOVERS OF U.S

QUANTITATIVE EASING ON EMERGING

MARKET ECONOMIES

Utku Dogus Kiran

11374810

Master Thesis

International Economics and Globalization

2017

Date: 10.07.2017

Supervisors

dhr. dr. K.B.T. (Boe) Thio

Abstract

This paper conducts an empirical research about possible spillover effects of Federal Reserve’s quantitative easing measures on main macroeconomic variables of

emerging market economies. A PVAR model is built on 15 emerging market economies to investigate cross-border spillovers of liquidity and real asset purchase operations of the Federal Reserve. Empirical findings indicate that there exist long and short-run relationships between Fed interventions and macroeconomic variables of EMEs. It is found that liquidity operations lead to a decline in industrial

production and inflation. Also, they cause real exchange rate appreciations and increase government bond yields in emerging markets. By contrast, spillovers arisen due to the real asset purchases of Fed are found to be an increasing factor for

industrial production in EMEs. Besides, treasury purchases of Fed lowered EM bond yields and led to real exchange rate appreciations in EMEs. Regarding inflation, effects of treasury purchases are found to be ambiguous while mortgage-backed security purchases contributed to an increase in consumer prices. These findings allow concluding that during the first phase of the financial crisis, liquidation of the U.S economy led to capital outflows from emerging markets through several international transmission channels. Starting from 2010 these outflows turned into capital inflows because of the relative increase in EM asset yields.

Keywords: Monetary policy, quantitative easing, spillovers, emerging markets, capital flows, panel data, vector auto-regression.

Statement of Originality

This document is written by Utku Dogus Kiran, who declares to take full responsibility for the contents of this document.

I claim that the text and the work presented in this paper is original and that no sources other than those mentioned in the text and its references used in creating it.

The Faculty of Economics and Business of the University of Amsterdam is responsible solely for the supervision of completion of the work, not for the contents.

Table of Contents

I- INTRODUCTION……….………...……....6

II- INTERNATIONAL TRANSMISSION CHANNELS OF QE AND RELATED LITERATURE….………...…...…………..9

III- EMPIRICALINVESTIGATION OF US QE ON EMERGING MARKETS’ MACROECONOMIC VARIABLES……...……….…..14

IV- CONCLUSIONS.………….………..………...………....25

V - APPENDIX ..……….……….………..………. 28

Abbreviations

ADF: Augmented Dickey-Fuller

BIS: Bank for International Settlements

CADF: Cross-Sectionally Augmented Dickey–Fuller CPI: Consumer Price Index

DHg: Durbin-Hausman Group DHp: Durbin-Hausman Panel EM: Emerging Market

EMEs: Emerging Market Economies Fed: Federal Reserve

FEVD: Forecast Variance Error Decompositions FOMC: Federal Open Market Committee

GMM: Generalized Method of Moments GSE: Government-Sponsored Enterprise IP: Industrial Production

IR: Interest Rate

IRF: Impulse Response Function(s) LFI: Lending to Financial Institutions LKCM: Liquidity to Key Credit Markets LSAP: Large-Scale Asset Purchase LSDV: Least Squares Dummy Variable LTTP: Long Term Treasury Purchases MBS: Mortgage Backed Security

MBSP: Mortgage-Backed Security Purchases MEP: Maturity Extension Programme

OMO: Open Market Operations RER: Real Exchange Rate VAR: Vector Autoregression

PVAR: Panel Vector Autoregression FAVAR: Panel Vector Autoregression QE: Quantitative Easing

SVAR: Structural Vector Autoregression UMP: Unconventional Monetary Policy US: United States

I- Introduction

Central Banks use monetary policy tools to balance business and financial cycles to create a sustainable economic environment. Controlling short-term interest rates was broadly used to regulate economic activity before the 2008 crisis. However, weak economic growth and deflation risks in the major economies led central banks to decrease their policy rates gradually. As a result, they reached the zero lower bound (ZLB), and low-interest rates persisted for years. Conventional counter-cyclical monetary policy tools as interest rate adjustments may not lead to desired outcomes when central banks reach the ZLB. That is because the further decrease in policy rates to boost economic activity either is not possible arithmetically or might create a default risk for financial and banking sectors. Therefore, central banks have begun to apply unconventional monetary policies (UMP) such as quantitative easing,

qualitative easing, and forward guidance. Federal Reserve is one of the most active and effective central banks in terms of implementing several UMP measures in different time periods in the recent past.

Before the global financial crisis, the Fed, as the other major central banks did, gradually lowered its policy rate and it reached the zero lower bound, which made conventional monetary policy tools ineffective. After Bear Stearns’ insolvency and the collapse of Lehman Brothers, the U.S. economy was critically threatened by a financial crisis. Banks were hesitant to lend each other because they did not know the details of other banks’ balance sheets. Trade volumes dropped rapidly, and volatility jumped. During 2008, as a lender of last resort, many of the financial institutions and credit markets were supported by the Fed to stimulate economic activity and to solve liquidity problems. The balance sheet of Federal Reserve is expanded by some $1.8 trillion during these operations. These interventions were known as liquidity

operations. Besides, starting from March 2009, real asset purchase programs enlarged Federal Reserve’s balance sheet even further, and these actions are known as Large Scale Asset Purchase (LSAP) programs. ‘Open Market Operations’ (OMO) of the Federal Reserve were conducted by The Federal Open Market Committee (FOMC) which was assigned to pursue LSAP operations. Overall, these liquidity and

real asset purchase operations were different kinds of unconventional monetary policy tools of the Fed, which were identified as credit easing or quantitative easing (QE) policies (Bernanke, 2009).

The United States’ QE refers to enlarging the asset side of the Fed’s balance sheet by liquidity operations and purchasing long-term financial assets from the financial sector. Interventions aimed to decrease the yields of US financial assets to support and improve credit conditions in the economy. It is also widely used after the global financial crisis by major central banks. One of the aims of QE policy is to boost economic activity and to regulate the inflation by injecting liquidity into the markets (Dudley, 2010). Regarding QE operations in the US, the primary goal of the first program (QE1) was decreasing borrowing costs for mortgage credits to activate housing markets, as well as to improve financial conditions in the economy. QE1 includes the measures; (i) Liquidity operations to support financial institutions and credit markets and (ii) LSAP of Mortgage-backed securities (MBS), government-sponsored enterprises (GSE) debt, agency debt, and treasury securities. It continued until March 2010 and enlarged the assets side of Fed’s balance sheet by $1.4 trillion. Even though these remarkable interventions took place, the U.S economy was not stable enough, and Mr. Bernanke expressed his opinions about the risk of deflation, which was a signal for further measures to spur economic growth. In 2010, the Fed applied its second unconventional intervention (QE2), from November 2010 to June 2011. QE2 included monthly purchases of US treasury assets up to $600 billion. Furthermore, $667 billions of government debt securities were issued under the name of Maturity Extension Program (MEP). QE2 was aimed to decrease long-term

interest rates on the purchased assets while increasing the inflation rate. Despite monthly injections of dollars, labor markets were not active enough, which led to the implementation of the QE3 program in September 2012. So, Fed purchased $160 billion worth of MBS.

Most of the literature focuses on the domestic effects of these interventions. However, because of the QE operations, the global liquidity of the U.S dollar increased. This led to the presence of higher yields on EM assets due to the

depreciation of the dollar, so that capital flows to developing nations were

significantly increased (Bhattarai et al., 2015). Free movement of capital may lead to an efficient provision of resources which in turn prompts productivity and economic development. However, the surge of capital inflows also leads to appreciation pressures on EM currencies, which may create financial imbalances. That might cause substantial challenges for policymakers in the EMEs. Cross-border impacts of QE interventions on EMEs not only occur because of the capital flows but also through various transmission channels, which will be explained in detail in the following section.

In general, QE operations have substantial effects on EMEs. It increases output, lowers unemployment and interest rates, also leads to decrease in the risk premium on financial assets. For a better understanding of international transmission

mechanism of QE on EMEs, understanding the dynamics of spillover effects are crucial. Therefore, this paper investigates the possible dynamics of such spillover effects.

Moreover, most of the literature has focused on the outcomes of these policies on the major economies, but studies on the emerging market economies have been less abundant. This is another incentive to pursue this research on EMEs. Overall, the primary concern of this study is to investigate cross-border impacts of both real financial asset purchases and liquidity operations of Fed on emerging market economies’ macroeconomic variables and in this way to contribute to the literature. Therefore, this research conducts an empirical study, through a panel vector

autoregression analysis (PVAR), on EMEs to analyze international spillover effects of the Federal Reserve's quantitative easing operations.

The model used in this paper was employed by Turkay (2016) to show short-run relationships between the government bond and MBS purchases of Fed on the one hand, and macroeconomic variables in the EMEs on the other hand. In this research, Mr. Turkay’s model will be extended by considering each subheading of the different currency injection tools of Fed as a different QE variable. Next, they will be tested separately with the EM macroeconomic variables to discover the various spillover

effects of separate interventions. Specifically, the four different currency injection tools to be considered are distinguished as, on the one hand, LSAP operations which refer to long-term treasury purchases (LTTP) and purchases of mortgage backed securities (MBS), and, on the other hand, Liquidity operations which refer to lending to financial institutions (LFI) and liquidity to key credit markets (LKCM). Through bilateral testing the effects of these policy tools on EM macroeconomic variables, the aim is to get an overall picture of cross-border effects of Fed’s QE operations. To my knowledge, testing spillover effects due specifically to the liquidity operations of the Fed with the PVAR approach is new to the literature.

The remainder of the paper is organized as follows. The next section contains the literature review that summarizes associated information from highlighted articles in the literature. Besides, it provides a theoretical background about international transmission mechanisms of QE. Section 3 starts with the Methodology and Data for the empirical research on EMEs. Precisely, I choose 15 EMEs as in Uribe and Schmitt-Grohe (2017). Next, it applies a panel autoregressive (PVAR) model with a monthly time interval from 2008:1 to 2015:12. Empirical results and interpretations are in the form of impulse response functions (IRF) and forecast error variance decomposition (FEVD) analyses of EM macroeconomic variables. Key findings and directions for future work will be presented in the final section.

II- International Transmission Channels of QE and Related Literature

The literature defines four major transmission channels related with international externalities of QE. First, there is the portfolio rebalancing channel. It is the most frequently cited transmission channel in the literature that takes its roots from theories of prominent monetary economists as James Tobin (1961, 1963, and 1969) and Milton Friedman (1978). These rely on the idea that financial assets in the market should be considered as substitutes for each other. Hence, changes in the

yields and supply of these financial assets alter investors’ preferences accordingly. For instance, LSAP of the Federal Reserve decreases the amount of US Treasuries and MBS as well as their yields in the market. Bernanke (2012) states that investors have different risk aversions and preferences of maturities of the assets. As yields on the assets are reduced due to the LSAP interventions, investors rebalance their portfolios to the financial assets with higher yields. Besides, abundant liquidity in the market encourages market attendants to invest in riskier assets (Borio and Zhu, 2008). In other words, because of the lack of substitution for US financial assets whenever their yields decrease, their substitutes including emerging markets assets attract investors’ attention so that it leads to an increase in demand for emerging market assets. To illustrate, Gagnon et al. (2011) and Hamilton et al. (2012), show empirically that LSAP immediately lowers the yields on financial assets, which causes the rebalancing of portfolios towards their substitutes. Portfolio rebalancing causes a reduction in risk premiums and an increase in prices, so it leads to the presence of lower yields on emerging market economies’ financial assets (Robert Lavigne et al., 2014). Consequently, real asset purchase operations lead to portfolio rebalancing through more risky and high return assets, including EM financial assets. The second transmission mechanism of the QE is the trade channel. According to the Mundell-Fleming model (Mundell,1963), as a conventional policy measure,

monetary easing increases the output level in the domestic country. Likewise, among UMP tools, QE operations have similar effects as the monetary easing on the

countries’ economic activity. Indeed, increased spending and easier trade credit conditions during QE operations, boost demand for imported goods and services from the emerging market economies. This would contribute to EM countries’ output. Meinusch and Tillmann (2016) and Kapetanios et al. (2012) tested the effects of QE on the emerging market countries’ economic activity and found that the output level in these countries increased due to the QE policies. A similar result is presented by Dahlhaus (2014), who concluded that production growth in the country raises the demand for imported goods and services, which contributes to the economy of

foreign exporters. However, the effectiveness of the trade channel varies according to the import elasticity in the U.S. regarding commodities imported from emerging

markets (Chen et al., 2012). Besides, some authors such as Peersman (2011), and Ugai (2007) argued that such additional output increases due to the QE interventions are limited and temporary. Thus, the spillover effects of the trade channel are

anticipated to be ambiguous and the least active channel among the transmission channels.

The third is the exchange rate channel, also called the expenditure switching effect. The Mundell-Fleming model states that expansionary monetary policy leads to currency depreciations. Correspondingly, during QE implementations by the US, the country may experience an exchange rate depreciation, Glick and Leduc (2013), Joyce et al. (2011) and Rosa (2012) empirically showed these currency depreciations. As a result, the terms of trade worsen for the U.S (goods and services become less expensive for foreigners), and net exports increase. This situation created the beggar-thy-neighbor effect1 which had a negative impact on output of foreign countries through the expenditure switching effect (Dahlhaus, 2014). The influence of the exchange rate channel is stronger for the major trade partners of the U.S. Moreover, since the dollar is the primary reserve currency in world markets, depreciation of the US-dollar is much more influential than that of other currencies also engaging in QE. Thus, QE in the US might have an adverse effect on emerging market economies (Sun, 2015). To illustrate, Canada is a major trade partner of the U.S. so that QE implemented by Federal Reserve would have a substantial impact on the Canadian economy through the exchange rate channel. Furthermore, since the U.S. has a significant share of the world economy and a reserve currency, depreciation in the U.S dollar affects emerging markets’ exchange rates, interest rates, and size and volatility of capital flows (Fic, 2013). It is important to notice that the trade channel on the one hand, and the exchange rate channel on the other hand, may have contrary effects on the output of foreign countries. There is no consensus in the literature regarding the dominance of either of these two effects. If the exchange rate effect

1 _{The beggar-thy-neighbor strategy aims to increase demand for domestic country’s exports while decreasing}

local reliance on imports. This situation often occurs by a devaluation of the currency to gain a trade advantage on export goods and services.

dominates, production declines in the foreign country, whereas foreign output increases when the effect of the trade channel dominates (Kawai, 2015).

Another transmission channel is the signaling channel (expectation effect). This refers to the forward guidance of the central banks, which gives signals to the markets about the current situation of the economy and future implications of the central banks’ policies. A commitment to QE may cause expectations about policy rates of the central bank to remain close to zero for a considerable time. One needs to notice that this perception seems accurate since, for instance, a subsequent increase in interest rates would lead to significant losses on Fed’s purchased assets. Therefore, large and increasing interest rate differentials relative to the EMEs may be

anticipated by investors. These expectations increase the demand for emerging market economies’ financial assets. When the Fed expands its balance sheet with the liquidity and asset purchase operations, it boosts carry trades2 and capital flows into the EMEs. Because of investors’ higher risk-adjusted return towards emerging economies’ assets, inflation and assets prices increase. Chen et al. (2012), conclude that spillovers of QE would lead to the presence of healthy macroeconomic

fundamentals and solid economic growth in the emerging markets.

International spillover effects of QE transmitted through afore-mentioned channels, and overall effects can be divided into two main subheadings. First, we consider International Financial Spillover Effects. In this domain, Panel Regressions and Event Study Approaches are broadly used. The latter approach was applied by Fratzcher et al. (2013) on 42 EMEs and 21 developed economies to examine effects of QE1 and QE2. The authors found that QE policies, especially starting from the QE2, boosted equity markets globally. Moreover, the authors conclude that QE policies led to significant portfolio rebalancing. In QE1 where liquidity operations took place, the authors showed portfolio outflows from emerging market economies, while opposite effects were observed for QE2. Similarly, Chen et al. (2012)

empirically showed a significant appreciation in the exchange rates, and a decrease in

2

Carry trade refers to a trading strategy to benefit from profit arises due to arbitrage opportunities by borrowing a currency that has a low-interest rate and investing borrowed money in a currency that has a high-interest rate.

government bond yields, in an event study conducted on EMEs in Latin America and Asia. Similarly, Georgiadis and Grab (2016) and Neely (2015), through an event study approach on main emerging markets, supported the previous authors’ conclusions: a global increase in equity prices and a decrease in the corporate and government bond yields. Such global alterations of dynamics in global financial markets, attract investors’ attention towards financial assets of the EMEs. Using a panel study approach, Moore et al. (2013), Chen et al. (2014), and Bowman et al. (2015b) conducted empirical studies on global financial effects of QE and

highlighted parallel conclusions.Ahmed and Zlade (2014), Lim and Mohapatra (2016) state that capital flows into EMEs increased during the period of Fed’s QE2. Kiendrebeogo (2016) found similar results regarding capital flows and concludes possible tapering (exit) from QE would probably have opposite effects on capital flows to the EMEs. Furthermore, Meinusch and Tillman’s research in 2016, showed explicit effects on government interest rates, exchange rates and equities where a Qual VAR model is built upon emerging markets’ financial indicators to observe the impact of quantitative easing.

Second, we consider the question whether International Macroeconomic Spillover Effects of QE in the US are substantial for macroeconomic variables of emerging markets. Many researchers apply Vector Autoregression (VAR) Models to answer these types of questions. Among those, Gambacorta et al. (2014) built a SVAR model for selected developed economies and presented that QE increased output and prices. Barosso et al. (2015) used SVAR analysis to show that QE contributed to the growth and led to credit booms in Brazil. Similarly, Carrera et.al. (2015) found, in a study regarding Latin American economies, that significant increases in capital flows led to currency appreciations and credit growth. Such effects lead us to think about the sensitivity of the EMEs for monetary policy shocks. Chen et al. (2016) underline that effects in emerging markets are relatively larger than those in developed

countries. Hence, macroeconomic spillover effects of QE vary across countries and time. Output and inflation effects presented with a FAVAR model for Canada, applied by Dahlhaus et al. (2014), were significant and positive. QE spillovers vary across countries regarding channels and characteristics of the countries. Chen et. al

(2014) conducted both a panel regression and an event study on 21 EMEs and found that countries with better fundamentals are affected relatively less by QE policies. Thus, in general, QE has clear effects on EMEs. These effects show differences regarding the characteristics of interventions and they mainly occur due to the fluctuation in capital flows. In the case of capital inflows caused by QE

interventions, typically output increases, while unemployment and the interest rate on government bonds decrease in EMs. Furthermore, it causes a substantial reduction in the risk premium on financial assets. As a result, confidence increases which leads to improved financial conditions in EMEs. In this context, this paper tries to

contribute to the existing literature by conducting a PVAR model on selected EMEs’ macroeconomic variables to observe international spillovers of different QE

interventions of Fed.

III- Empirical Investigation of US QE on Emerging Markets’ Macroeconomic Variables

3.1 PVAR Model

The aim of PVAR model is to show the dynamic short-run relationships between Federal Reserve’s QE interventions on industrial production, consumer price index, government bond interest rate, and real exchange rate in selected emerging market economies. Except for government bond yields, all data is in logarithmic form and seasonally adjusted. Government bond yields are used in levels. The optimal lag order is chosen to be 1 according to the Akaike, Bayesian and Hannah-Quinn information criteria. As a result, a PVAR (1) equation can be presented as:

Where i= 1,2,…,M (M=15 countries), t=1,2,…,T (T=96) and

b

t is the lag operator.

Y

i,t is a five variables vector that includes QE, IP, CPI, IR, and RER. The variable QE denotes Fed’s data for quantitative easing interventions, so it refers to a different data set in each PVAR test. Once a particular dataset is chosen as QE variable, it will be the same for all countries throughout the test. Specifically, these different QE datasets are ‘lending to financial institutions (LFI)’, ‘liquidity to key credit markets (LKCM)’, ‘long-term treasury purchases (LTTP)’, and ‘mortgage-backed security purchases (MBSP)’ of Federal Reserve. Other variables in the vector belong to the selected emerging market economies where IP represents industrial production, CPI is consumer price index, IR is 10-year government bond yield, and RER is the real effective exchange. For rest of variables, a represents intercept of

Y

i,t, Di is dummy variable for ‘ith’ country,

e

i,t accounts for the error term. Including a dummy variable as well as an intercept for each country would lead to perfect multi-collinearity so one of the dummies is dropped from the equation. In other words, since we have a constant, 𝑀 − 1 dummies are included in the equation. The advantage of including dummies for all countries is that individual dummies demonstrate the specific effects to the countries, so it allows controlling unobserved heterogeneity across nations.

PVAR model represented above will be tested by using a least square dummy variable estimator (LSDV) of Cagala and Glogowsky (2014). The way in which LSDV estimator operates is described below:

“LSDV estimator is a Stata module which allows multivariate panel regression of each dependent variable on lags of itself and lags of all the other dependent variables. This module also produces forecast error variance decomposition and impulse

response function graphs by using Monte Carlo simulation.” (Cagala and Glogowsky, 2014).

The LSDV estimator is selected, rather than a Generalized Method of Moments (GMM), because the properties of GMM estimators are appropriate when M is large, and they cause biased and inconsistent outcomes when the panel has a small M

(Bruno, 2005). With large macro panels, results of the LSDV estimator are consistent and unbiased. Besides, the least square dummy variable method is appropriate to use for a relatively small number of observations (M). It prevents from having too many dummy variables. Since my panel has relatively small M and large T (T>M), it is appropriate to use LSDV estimator instead of GMM.

It is important to elucidate the issue of stationarity of variables in the VAR system. According to Sims (1980) and Sims et al. (1990), the distribution of test statistics is not affected by non-stationarity. In other words, there is no need to transform the system into a stationary form. The authors claim that the aim of the VAR systems is to investigate the relationship between variables, not to estimate parameters. Sims et. al. (1990) argue that, if variables in the system are cointegrated, the VAR model may be tested in levels. Following this, the PVAR model is estimated in levels.

Another important issue is the Cholesky ordering in the PVAR system. The model assumes a lower triangular Cholesky form which means that variables which appear early in the system are comparatively more exogenous than the variables which appear later. The former variables affect the following ones both contemporaneously and with a lag. Instead, variables that appear later in the system have an influence on the previous variables through a lag only. The first variable in the system is QE which is the most exogenous variable for EMEs. National dynamics are thought to lag behind the global dynamics in spillover analysis. Industrial production and

inflation come earlier than government bond yield and real exchange rate because the former variables influence the latter ones simultaneously while the latter variables have an impact on the former ones through its lag only. The most endogenous variables in the system are government bond yields and real exchange rates. The Cholesky ordering of the PVAR system follows those applied in the literature {Sousa and Zaghini (2008), Belke et al. (2010) and Brana et al. (2012)} so that output and prices come before the financial indicators.

After the coefficients of the model are estimated, the forecast error variance

show the short-run impact of the Fed’s QE on macroeconomic variables of the emerging market economies. To clarify, the FEVD represents the information amount of an individual variable contained as a contribution to the other variables in the regression. That is to say, how much of the forecast error variance of a single variable can be explained by an exogenous QE shock. IRF graphs denote response of an endogenous macroeconomic variable to a one-standard-deviation positive shock from an exogenous QE variable.

3.2- Data

The paper uses monthly data from 2008:1 to 2015:12 for 15 EMEs. Countries included in this study have been chosen according to the book “Open Economy Macroeconomics” written by Uribe and Schmitt-Grohe (2017). Selected emerging market economies are China, Chile, Colombia, Czech Republic, Brazil, India, Indonesia, Malaysia, Mexico, Philippines, Poland, South Africa, South Korea, Taiwan, and Thailand. Data for the US quantitative easing and EM government bond yields are taken from DataStream; consumer price index and industrial production data are obtained from the World Bank Global Economic Monitor database. Data for real exchange rate are taken from Bank for International Settlements’ database. To observe the dynamic short-run effects of different QE interventions on EMEs, outcomes of the PVAR model should be unbiased, consistent and economically meaningful so there is a necessity to check two statistical requirements, which are stationarity and cointegration.

In the time series, regression applies to the variables against their own lags. When dependent variables are plotted, spikes (ups and downs) appear in the data. This movement of the data represents volatility (or in this case: an effect of shocks). The idea of checking for stationarity is to determine whether the effects of these shocks are permanent or temporary. If the effect of a shock is temporary, the value of the dependent variable will return to its long-run equilibrium level, but not so if it is permanent. The former variables are called stationary or I(1) while the latter ones are

non-stationary or I(2). Due to the non-stationary variables shocks are absorbed into the system and lead to deviations from long-run equilibrium of dependent variables that is one rationale for checking stationarity. Another rationale is unit root testing for determination of trends that might result in possible bias in the auto-regression analysis. If a variable follows a non-stationary process, it has a unit root and a stochastic trend3. Therefore, it might cause spurious regression4 results in the analysis. To avoid this risk, it is essential to check for cointegration to eliminate common stochastic trends in the system.

Hence, before interpreting PVAR outcomes, panel unit root tests and a panel cointegration test are implemented to figure out whether analyzed variables are stationary, also to investigate long run cointegration relationships among variables.

3.3 Unit Root and Cointegration Tests

3.3.A - Unit Root Tests

Among panel data variables, Quantitative Easing is a cross-sectional time-invariant variable and the same for all countries, so a Phillips-Perron (1988) unit root test is implemented to investigate whether the QE is stationary.

Additionally, the model uses Maddala and Wu (1999) and Pesaran (2007) unit root tests because of their wide range of application in the literature (Turkay 2016). The Maddala and Wu test5 is a Fisher-type test which pools individual p-values of

3

A stochastic trend is a persistent but random long term movement of a variable over time. A priori not clear how to model trends so instead of modeling, possible bias – might arise due to the existence of stochastic trends- is preferred to be eliminated by testing for cointegration.

4

Non-stationary variables might contain common stochastic trends. In the system, these variables can be observed as correlated with the other variables even if they are not related at all. This situation called as the existence of spurious regression in empirical analysis. Biased and inconsistent outcomes can be prevented by checking for cointegration. According to results of cointegration test, if variables are cointegrated, there will be no more common trends and spurious regression risk.

5

every cross-section ‘i’ for determination of unit roots in the panel data. The rationale of using a Maddala and Wu unit root test is that it allows for serial correlation among error terms and heteroscedasticity. Also, from the Monte Carlo simulation, authors found that Fisher-type test gives consistent and unbiased results with small sample sizes and large observation periods (Maddala and Wu, 1999).

Another unit root test implied is the Pesaran (2007) 6 unit root test. The main reason for application of Pesaran (2007) panel unit root test is that spatial dependency7 or spillover effects might, for instance, create cross-sectional dependencies across countries. In other words, there is a possibility of the presence of unobserved common factors among countries. The Pesaran (2007) test takes this issue into account. This unit root test is an extended version of ADF regression with the lagged sectional mean and first-differences of individual series. This is called a cross-sectionally augmented Dickey-Fuller (CADF) test.

3.3.B - Cointegration Test

To investigate long run cointegration relationship among variables Durbin-Hausman co-integration test is implemented8. It is first presented by Westerlund in 2008 and the test can be applied when variables are integrated of a different order. Moreover, unlike first generation cointegration tests, the Durbin-Hausman test allows for cross-sectional dependence, heteroskedastic and serially correlated error terms.

6

Detailed calculation to obtain test statistics of Pesaran (2007) unit root test can be found in Appendix 1B 7

Spatial dependency is the joint variability (co-integration) of two random variables within a topographical area: features of specific localities appear to be either positively or negatively interconnected. Therefore, spatial dependency causes the spatial autocorrelation that is diametrically opposite to fundamental statistical assumption of conditional mean zero. This assumption requires random samples of a larger set need to be independent (not correlated) from each other to get unbiased outcomes from a linear regression.

8

Detailed calculation to obtain test statistics of Westerlund (2008) Durbin-Hausman cointegration testcan be found in Appendix 1C

The classical Durbin-Hausman test has two scopes: (i) Panel scope (DHpanel)

(ii) Group scope (DHgroup)

The assumption for DHpanel test is that all cross-section units have the same autoregressive parameter. Because of this assumption, if the null hypothesis is rejected, cointegration is present for all cross-sections. The Durbin-Hausman group test assumes that the autoregressive parameter can change within cross-sections under the alternative hypothesis, so that rejection of the null hypothesis tells us there is a cointegration relation for some individuals.

3.3.C Unit Root and Cointegration Test Results

Application of the Phillips-Perron unit root test shows QE data is I(1), it is non-stationary. Additionally, according to Maddala and Wu (1999) and Pesaran (2007) panel unit root test results, some variables are I(0) while the others are I(1). In other words, there is mixed evidence regarding stationarity of the variables. Thus, it is crucial to check for cointegration test results. This is the best possible way to ascribe an economic meaning to PVAR outcomes. Durbin-Hausmann cointegration test results reveal that there exists a long-run cointegration relationship among Federal Reserve’s interventions and macroeconomic variables of the emerging market economies. Test results are presented in tables 1 - 4 below.

Table 1. Phillips-Perron () Unit Root Test Results for QE Dataset

*, **, *** show significance level of at 10%, 5% and 1%, respectively

Table 2. Maddala-Wu (1999) Unit Root Test Results

*, **, *** show significance level of at 10%, 5% and 1%, respectively.

Table3. Pesaran (2007) Unit Root (CIPS) Test Results

Table 4. Westerlund Durbin-Hausman Cointegration Test Results

*, **, *** show significance level of at 10%, 5% and 1%, respectively

3.4 PVAR Test Results

Since long-run cointegration relation among variables is revealed, PVAR outcomes can be interpreted as short-run dynamics of the system. In the appendix section, the first four figures are associated with the LSAP operations, while the second four relate to the liquidity operations of Fed. Therefore, figures related to ‘lending to financial institutions’ (LFI) and ‘liquidity to key credit markets’ (LKCM) provide a basis for interpreting the spillovers of Fed’s liquidity operations. Besides, figures related to ‘long-term treasury purchases’ LTTP and ‘mortgage-backed security’ (MBS) purchases denote the spillovers of LSAP operations of Fed in the short-run. For each kind of QE - operations, figures are presented both for the impulse response functions (IRF) and for the forecast error variance decomposition (FEVD). These have different graphical representations as well as interpretations. Concerning the IRF graphs, impulses and responded variables are mentioned on the top of the chart and each column represents impulse responses over 20 months. Functions in the graph show impulse responses of specific macroeconomic variables to a one-standard-deviation positive QE shock. Specifically, among the plotted lines in each graph, the upper (green) and lower (red) lines are one-standard-error bands, while the blue line in the middle is the mean response of the variable. Regarding representation and interpretation of FEVD graphs, the particular EM macroeconomic variable of interest is labeled on the top of the chart and red columns display forecast error

variances of labeled variables (each column represents nearly 5 months). Blue columns, which are located inside the red ones symbolize the amount of information, due to the exogenous QE shock, that a specific macroeconomic variable contains. Because most of the studies in the literature are focused on real asset purchase operations of Fed, it is appropriate to start with the interpretation of LSAP operations of the Federal Reserve.

Figure 1 gives the impulse response graphs of respective EM macroeconomic variables as derived from an exogenous and positive long-term treasury purchase shock of Fed. The first column of ‘logip’ graph (from 0-5) covers more than 20 months’ response of the industrial production. As it shows, a positive LTTP shock is followed by an increase in industrial production. The outcome is permanent because the one-standard-error band does not include the zero line and impulse responses stays on the positive side through the time interval. It can be seen in the CPI graph that although there is a signal of a negative impact on inflation in the mean response, the one-standard-error band includes the zero line, which makes net effect

ambiguous. Figure 2 supports this conclusion because there are no blue columns found inside the red columns of CPI graph. In other words, there is not enough information found that explains inflation in emerging markets decreased by

externalities of long-term treasury purchases in the short-run. The highest effect of LTTP on industrial production occurs in about 1 year. In addition, impulse response results of IR indicate that a positive shock to long-term treasury purchase lowers 10-year government bond yields in emerging market economies and leads to exchange rate appreciation in the short-run. The main trigger is the surge in capital flows into EMEs through transmission channels such as portfolio rebalancing and signaling. Mortgage-backed securities were another type of financial assets, which have been purchased during LSAP operations. Similar results are obtained in terms of industrial production (figure 3). One can see from the graph that MBSP operations increased economic activity in emerging markets both positively and permanently. MBSP by the Fed led to an increase in consumer prices, and also it pointed to exchange rate depreciations. Thus, it hypothetically means that upward pressure on inflation due to

the increase in output was larger than the downward pressure on emerging markets currencies due to exchange rate depreciations. It is important to mention that the first two columns of the CPI graph in figure 3 demonstrate an exchange rate depreciation in emerging markets that corresponds to a 3 years time period in which liquidity operations also took place. Starting from 2010, there were just LSAP operations by the Fed, and a reverse effect on inflation is observed. Therefore, it is hard to make an assessment about overall spillovers of MBSP on inflation. The mean response of Figure 3 shows MBSP increases sovereign bond yields in EMEs. However, the one-standard-error band includes the zero line and no significant information is found in figure 4 to explain that this increase is because of an exogenous MBSP shock. Therefore, spillover effects of MBSP on government bond yields are found to be ambiguous.

In general, findings of the PVAR model in terms of real asset purchases of Fed are similar with the previous studies in the literature. For instance, Bowman et al. (2015), Chen et al. (2012), Dahlhaus et al. (2014) Georgiadis and Grab (2016) Moore et al. (2013) found that LSAP interventions lowered government bond yields and led to exchange rate appreciations in emerging market economies. Moreover, previous papers of Ahmet and Zalde (2014), Carrera et al. (2015), Gambacorta et al. (2014), Meinusch and Tillmann (2016) have shown a surge in capital flows and output in EMEs during LSAP operations. These findings are also parallel with the PVAR outcomes.

On the other hand, results related to the liquidity operations provide a different perspective on spillovers of QE operations. It is essential to mention that liquidity operations of the Fed occurred in the early phase of the crisis and lasted for one year (2008-2009). During these operations, the asset side of the Federal Reserve’s balance sheet were dramatically enlarged ($1.8 trillion) and, starting from 2009, provided liquidity gradually decreased. Figure 5 shows that the first liquidity operation, lending to financial institutions (LFI), leads to a decline in the emerging markets’ economic activity and decreases inflation. Effects reach a maximum in about three years. The spillover effects of LFI on government bond yields are unclear. Although

in figure 6 there are blue columns inside the first two red columns which cover about 1 year of time, which effect appears to be positive from the mean response in the ‘ir’ graph in figure 5, the one-standard-error band includes the zero line which makes the spillover effects unclear. There exist exchange rate appreciations in the short-run. Without capital inflows, exchange rate appreciations in EMEs is a sign for

depreciation of the U.S dollar. Likewise, the second liquidity operation - liquidity provided to key credit markets (LKCM) - increased the Fed’s balance sheet by $400 billion at the beginning of the financial crisis. Findings regarding the spillover effects of LKCM are parallel to that of LFI. We can see from figure 7 that industrial

production decreased and inflation temporarily decreased in EMEs. The first column of figure 7 illustrates liquidation of the key credit markets in the U.S. leads to an increase in government bond yields which are a sign for capital outflows from EMEs. Furthermore, these operations lead to provisional exchange rate appreciations (figure 7). Evidently, liquidation of financial institutions and key credit markets by some $1.8 trillion led to depreciation of the U.S dollar. As a result, terms of trade deteriorated in the U.S which in turn boosted net exports. It is crucial to remark that liquidity operations create a beggar-thy-neighbor effect through the exchange rate channel, and has a negative impact on foreign countries’ output due to the

expenditure switching effect. One can detect these currency depreciations, for instance, by looking at the historical data between the years 2008-2009.

These findings correspond with the results provided by Dahlhaus (2014), Glick and Leduc (2013), Joyce et al. (2011), Neely (2015), Rosa (2012). Authors conclude that QE policies lead to a depreciation of the home currency. Besides, it is necessary to conclude liquidation of the domestic economy in the United States led to the

depreciation of the dollar and during the first phase of crisis it led to capital outflows from EMES. However, starting with the QE2, abundant liquidity in markets caused investors to capitalize EM assets through portfolio rebalancing channel. These results are similar, for instance, with the findings of Borio and Zhu (2008), and Fratzscher (2013).

IV- Conclusions

Major central banks and their unconventional monetary policy implementations raised attention in academics regarding possible effects of these policies. Federal Reserve was the leading central bank in terms of the application of several UMP measures during and after the global financial crisis. So far, studies predominantly focused on domestic consequences of QE. Since U.S has a reserve currency and a major share in world trade, these measures have some crucial spillover effects on foreign economies. Among these economies, studies on emerging markets are scarce so that my paper contributes to the literature by trying a spillover analysis of Fed’s QE on the emerging market economies’ macroeconomic variables. In addition, to the best of my knowledge, it is the first PVAR analysis on spillover effects of Fed’s liquidity operations.

Empirical results in this research demonstrate that there are, indeed, long and short-run relationships among quantitative easing operations and macroeconomic variables in emerging market economies. In the short-run, it is found that liquidity operations of Fed between the years 2008-2009 led industrial production in EME to decline, and they temporarily decreased inflation. Moreover, during these operations government bond yields increased and exchange rates appreciated in EME. These findings allow me to conclude in the first phase of global financial crisis U.S dollar depreciated and there were capital outflows from emerging market economies. Several transmission channels, particularly portfolio rebalancing and exchange rate channels, were main drivers of these spillover effects.

On the other hand, reverse effects were observed in relation to the large-scale asset purchase operations of the Federal Reserve. According to the PVAR outcomes, long-term treasury purchases led to the presence of lower government bond yields in emerging market economies. The rationale was a surge in capital flows to the EMEs. Economic activity in EMEs is permanently increased due to the LSAP interventions

of Fed. That is observed from impulse response functions as an increase in industrial production. Increase in EM countries’ industrial production reached a maximum in about 12 months. Analysis of spillover effects on LSAP operations leads us to think that effects are transmitted through various transmission channels such as portfolio rebalancing, trade, and signaling. Although spillover effects of long-term treasury purchases on inflation are found to be ambiguous, an increase of inflation due to mortgage-backed security purchases is observed. Additionally, treasury purchases of Federal Reserve affected the emerging market economies adversely and led to the appreciation of exchange rates in the short-run.

Overall, results obtained from the PVAR model are similar with the existing literature, and they provide some noticeable conclusions for authorities.

Policymakers in the advanced countries should not be naïve by setting policies unilaterally. They should increase their awareness of possible cross-border consequences of their monetary policies. Since emerging market economies represent a large share of the world economy, these spillover effects can have significant feedback effects on advanced economies. That should be taken into consideration for sustainability of the policies as well as manageability of the crises. Furthermore, policymakers in the EM countries should track the monetary policies of the advanced countries for possible spillovers which would help to alter their future targets and expectations accordingly.

V- APPENDIX

Figure 1. Federal Reserve’s QE – Long Term Treasury Purchases IRF of EM Macroeconomic Variables

IRF of EM Macroeconomic Variables

IRF of EM Macroeconomic Variables

Figure 6. Federal Reserve’s QE – Lending to Financial Institutions FEVD of EM Macroeconomic Variables

IRF of EM Macroeconomic Variables

Appendix 1A

According to Turkay, 2016:

Fisher-type test of Maddala and Wu (1999):

G = -2

"

𝑙𝑛𝑔

#$- i

(1)

Where -2lngihas a chi-squared distribution with 2 degrees of freedom. In other words, G is distributed as X2 with 2M degrees of freedom as a total number of observations goes to infinity for finite M.

The null hypothesis is:

H

0

= G

i

= 1, i=1,2,…,M

(2)

Against its alternatives:

H

A

= G

i

< 1, i=1,2,…3,M

1

; G

i

=1, i=M

1

+1, M

1

+2,…, M

(3)

Augmented Dickey-Fuller (ADF) test is applied for each cross-section units separately.

ADF regression:

Y

i,t

= a

i

+ G

i

Y

i,t-1

+

_0$-/#

𝑄

i,j

DY

i,t-j

+ e

i,t

; t=1,2,…,T

(4)

t-statistics are calculated for each cross-sectional unit and P-value is used to reckon test statistics and it is compared with the critical value (Baltagi, 2013).

Appendix 1.B

According to Turkay, 2016:

A CADF regression can be characterized as AR(p) equation which is augmented with current and lagged values of

y

t as below:

Y

i,t

= a

i

+ g

i

𝑦

i,t-1

+…+ d

i0

𝑦

t

+ d

i1

𝑦

t-1

+ … + d

ip

𝑦

t-p

+

e

i,t

(1)

To obtain CIPS statistic, the equation above is transferred into first difference equation and individual ADF statistics (CADFi) is measured for different cross- section units. Average of CADFi statistics with the equation below gives CIPS statistics.

CIPS =

2345#

6 789

Appendix 1.C

According to Turkay, 2016:

Calculation of Durbin-Hausman test is presented below: Assume following panel data model;

Y

i,t

= a

i

+ b

i

x

i,t

+ h

i,t

(1)

X

i,t

= dx

i,t-1

+ w

i,t

(2)

In addition, assume that

h

i,t follow below set of equations

h

i,t

= 𝜆

i

N

t

+ e

i,t

(3)

N

j,t

= 𝜌

j

N

j,t-1

+ u

j,t

(4)

e

i,t

=

µi

e

i,t-1

+ 𝜈

i,t

(5)

In this system,

N

t is a k-sized vector of common factors

N

j,t

𝜆

i is the conformable vector of factor loadings. To attain Durbin-Hausman test it is needed use first difference of equation (3) and it becomes:

∆h

i,t

= 𝜆

i

∆N

t

+∆e

i,t

(6)

To estimate

𝜆

iand

∆N

t,

∆h

i,t should be known. Since it is unknown, it is necessary to get OLS approximations and implement primary components which can be written as:

Δ

h

i,t

= Δy

i,t

- 𝛽

i

Δx

i,t

(7)

times the eigenvector from the highest eigenvalues of the 𝑇 − 1 𝑥 𝑇 − 1 matrix

∆h∆h

. ∆𝜆 is measured by using the equation below:

𝜆 =

∆h∆D

EF-

(8)

The First difference of residuals that are de-factored:

Δ𝑒

i,t =

Δh

i,t

- 𝜆

i

∆𝑁

t

(9)

𝑒

i,t

=

H0$%

∆𝑒

i,j

(10)

The null hypothesis ‘there is no cointegration among variables’ can be verified by testing below equation for µi = 1.

𝑒

i,t

= µ

i

𝑒

i,t-1

+ error

(11)

Another essential estimator to shape Durbin-Hausman test is Kernal Estimator which can be formulated as follow:

𝛺

i2

=

_EF-- K#0$K#

(1 −

_K#L-0

)

EH$0L-

𝑢

i,t

𝑢

i,t-j

(12)

The

𝑢

i,trepresents error term created in the OLS regression from equation (11). Pi is bandwidth parameter. It displays a number of autocovariances of

𝑢

i,tto measure Kernal estimator.

𝛺

i2 is a consistent estimator of

𝛺

i2and assessed variance can be shown as 𝜎i2. Two variance ratios are created

; 𝑍

i

=

QRS TRU

and

𝑍

m

=

QVS TVS S

, where

𝛺

_W%

₌

-W

𝛺

# % W #$-

and

𝜎

W%

=

_W- W#$-

𝜎

#%

(13)

These careful calculations subsequently lead us to find Durbin-Hausman test statistic as:

DHgroup = W 𝑍

#$- i µi - µi)2 EH$%𝑒i,t-1 and DHpanel = 𝑍m= µ- µ)2 W#$- EH$%𝑒i,t-1

VI- REFERENCES

Ahmed, S., Zlate A. (2014), Capital Flows to Emerging Market Economies: A Brave New World?, Journal of International Money and Finance, 48, 221-248.

Barroso, J.B.R., Silva L.A.P., Sales A.S. (2015), Quantitative Easing and Related Capital Flows into Brazil: Measuring its Effects and Transmission Channels Through a Rigorous Counterfactual Evaluation, Journal of International Money and Finance, 9(2), 165-212.

Bauer, M.D., C.J. Neely (2014), International Channels of the Fed's Unconventional Monetary Policy, Journal of International Money and Finance, 44, 24-46. Belke, A., W. Orth, R. Setzer (2010), Liquidity and the Dynamic Pattern of Asset Price

Adjustment: A Global View, Journal of Banking and Finance, 34(8), 1933-1945. Bernanke, B.S., V.R Reinhart, B.P. Sack (2004), Monetary Policy Alternatives at the Zero

Bound: An Empirical Assessment, Brookings Papers on Economic Activity, 2004(2), 1-100.

Bernanke, B.S. (2009), Speech Given at the LSE, 15 January 2009.

Bhattarai, S., Chatterjee, A. and Park, W. (2015), Effects of US Quantitative Easing on Emerging Market Economies, Federal Reserve Bank of Dallas Globalization and

Monetary Policy Institute, Working Paper No. 255, p.28.

Borio, C., & Zhu, H. (2008), Capital Regulation, Risk-Taking and Monetary Policy: A Missing Link in the Transmission Mechanism, Working Paper of Bank for

International Settlements, 1(1).

Bowman, D., J.M. Londono, H. Sapriza (2015b), US Unconventional Monetary Policy and Transmission to Emerging Market Economies, Journal of International Money and

Finance, 55, 27-59.

Brana, S., M.L. Djigbenou, S. Prat (2012), Global Excess Liquidity and Asset Prices in Emerging Countries: A PVAR Approach, Emerging Markets Review, 13(3), 256- 267.

Bruno, G. S. (2005), Estimation and Inference in Dynamic Unbalanced Panel-Data Models with a Small Number of Individuals, Stata Journal, 5(4), 473.

Cagala, T., U. Glogowsky (2014), Panel Vector Autoregressions for Stata (xtvar).

Carrera, C., F.P. Forero, N. Ramirez-Rondan (2015), Effects of the US Quantitative Easing on Latin American Economies, Peruvian Economic Association, No.2015-35. Chen, H., V. Cúrdia, A. Ferrero (2012a), The Macroeconomic Effects of Large-scale Asset

Chen, Q., A.J. Filardo, D. He, F. Zhu (2012b), International Spillovers of Central Bank Balance Sheet Policies, BIS Working Papers Series, No.66, 230-74.

Chen, J., T.M. Griffoli, R. Sahay (2014), Spillovers from United States Monetary Policy on Emerging Markets: Different This Time?, IMF Working Paper, No. 14/240. Chen, Q., A. Filardo, D. He, F. Zhu (2016), Financial Crisis, US Unconventional Monetary

Policy and International Spillovers, Journal of International Money and Finance, Volume 67, October 2016, 62–81.

Dudley, William C. (2010), The Outlook, Policy Choices and Our Mandate, Remarks at the

Society of American Business Editors and Writers Fall Conference, City University

of New York.

Fic, T. (2013), The Spillover Effects of Unconventional Monetary Policies in Major Developed Countries on Developing Countries, DESA Working Paper, No. 131.

Fratzscher, M., M.L. Duca, R. Straub (2013), On the International Spillovers of US Quantitative Easing, ECB Working Paper, No. 1557.

Friedman, M., (1997), Rx for Japan: Back to the Future, Wall Street Journal, December 17, A22.

Friedman, B.M. (1978), Crowding Out or Crowding In? The Economic Consequences of Financing Government Deficits, NBER Working Paper, No.284.

Gagnon, J., M. Raskin, J. Remache, B. Sack (2011), The Financial Market Effects of the Federal Reserve’s Large-Scale Asset Purchases, International Journal of Central

Banking, 7(1), 3-43.

Gambacorta, L., Hofmann B., Peersman G. (2014), The Effectiveness of Unconventional Monetary Policy at the Zero Lower Bound: A Cross-Country Analysis, Journal of

Money, Credit and Banking, 46(4), 615-642.

Georgiadis, G. (2016), Determinants of Global Spillovers from US Monetary Policy, Journal

of International Money and Finance, Volume 67, 41–61.

Georgiadis, G., Gräb J. (2016), Global Financial Market Impact of the Announcement of the ECB's Extended Asset Purchase Programme, Journal of Financial Stability, .257-265,26.

Glick, R., Leduc S. (2013), The Effects of Unconventional and Conventional US Monetary Policy on the Dollar, Federal Reserve Bank of San Francisco, Working Paper, No. 2013-11.

Hamilton, J.D., Wu J.C. (2012), The Effectiveness of Alternative Monetary Policy Tools in a Zero Lower Bound Environment, Journal of Money, Credit and Banking, 44(s1), 3-46.

Joyce, M., Lasaosa A., Stevens I., Tong M. (2011), The Financial Market Impact of Quantitative Easing in the United Kingdom, International Journal of Central

Banking, 7(3), 113-161.

Kawai, M. (2015), International Spillovers of Monetary Policy: US Federal Reserve's Quantitative Easing and Bank of Japan's Quantitative and Qualitative Easing, Asian Development Bank Institute, No. 512.

Keynes, J.M. (1936), The General Theory of Interest, Employment and Money, Palgrave Macmillan.

Kiendrebeogo, Y. (2016), Unconventional Monetary Policy and Capital Flows, Economic

Modelling, 54, 412-424.

Krugman, P. (2005), Is Fiscal Policy Poised for a Comeback?, Oxford Review of Economic

Policy, Volume 21, Issue 4, 515-523.

Lavigne, R., Sarker S., Vasishtha G. (2014), Spillover Effects of Quantitative Easing on Emerging-Market Economies, Bank of Canada Review, 2014(Autumn), 23-33. Lim, J.J., S. Mohapatra (2016), Quantitative Easing and the Post-Crisis Surge in Financial

Flows to Developing Countries, Journal of International Money and Finance, Volume 68, 331–357.

Maddala, G.S., S. Wu (1999), A Comparative Study of Unit Root Tests with Panel Data and a New Simple Test, Oxford Bulletin of Economics and Statistics,61(S1), 631- 652.

Meinusch, A., Tillmann, P. (2016), Unconventional Monetary Policy and the Spillovers to Emerging Markets, Journal of International Money and Finance, 66, 136-156. Moore, J., S. Nam, M. Suh, A. Tepper (2013), Estimating the Impacts of US LSAP's on

Emerging Market Economies' Local Currency Bond Markets, Federal Reserve Bank

of New York Staff Report, No. 595.

Mundell, R.A. (1963), Capital Mobility and Stabilization Policy Under Fixed and Flexible Exchange Rates, The Canadian Journal of Economics and Political Science, 29(04), 475-485.


Neely, C.J. (2015), Unconventional Monetary Policy Had Large International Effects, Journal of Banking & Finance, 52, 101-111.

Pesaran, M.H. (2003), A simple panel unit root test in the presence of cross section dependence, Working Paper, Trinity College, Cambridge.

Phillips, Peter C.B., Perron P. (1986), Testing for a Unit Root in Time Series Regression, Cowles Foundation Discussion Papers, Cowles Foundation for Research in

Economics, Yale University, revised Sep 1987.

Rosa, C. (2012), How Unconventional Are Large-Scale Asset Purchases? The Impact of Monetary Policy on Asset Prices, FRB of New York Staff Report, No. 560.

Sun, Gié. (2015), Spillover Effect of U.S. Quantitative Easing from the Emerging Markets Perspective, Research Master Thesis, Louvain School of Management.

Sims, C.A. (1980), Macroeconomics and Reality, Econometrica, 48(1), 1-48.

Sims, C.A., J.H Stock, M.W Watson (1990), Inference in Linear Time Series Models with Some Unit Roots, Econometrica, 58(1), 113-144.

Sousa, J., A. Zaghini (2008), Monetary Policy Shocks in the Euro Area and Global Liquidity Spillovers, International Journal of Finance & Economics, 13(3), 205-218.

Tobin, J. (1961), Money, Capital and Other Stores of Value, The American Economic

Review, 51(2), 26-37.

Tobin, J. (1963), An Essay on the Principles of Debt Management, Cowles Foundation for

Research in Economics, Yale University.

Tobin, J. (1969), A General Equilibrium Approach to Monetary Theory, Journal of Money,

Credit and Banking, 1(1), 15-29.

Turkay, M. (2016), International Transmission Mechanism of Unconventional Monetary Policy, Hacettepe University Graduate School of Social Sciences Department of

Economics, 49-63.

Uribe, M., & Schmitt-Grohe, S. (2017), Open economy macroeconomics, Princeton, NJ:

Princeton University Press, pp.27

Westerlund, J. (2008), Panel Cointegration Tests of the Fisher Effect, Journal of Applied

Econometrics, 23(2), 193-233.

Woodford, M. (2012), Methods of Policy Accommodation at the Interest-Rate Lower Bound,

Presented at the Jackson Hole Symposium, “The Changing Policy Landscape,” August 31-September 1, 2012.