1
Does ECB’s monetary policy effect capital inflows to
emerging market economies?
Master thesisAuthor: Stefan van den Berg, 10367004 Email: stefan.vandenberg@student.uva.nl
Supervisor: Kostas Mavromatis Second reader: Naomi Leefmans
Number of words: 11303 14 July 2017
Abstract:
This paper examines the effects of the European Central Bank’s monetary policy on volatile capital inflows toward emerging market economies (EMEs). Using panel data techniques on 10 EMEs and quarterly data, spanning the period between 2000-2015, the empirical research of this study finds evidence for a negative relationship between accommodative monetary policy and portfolio inflows
into the emerging markets. A 1% increase in the proxy for the overall stance of monetary policy in the Eurozone (MCI) leads to a 0.11%-point drop in portfolio inflow as a percentage of GDP for an
individual EME (holding all other variables constant). Furthermore, the results for other types of flows show less relevant results, except for a positive effect the MCI on foreign direct investment
2
Statement of Originality
This document is written by Stefan van den Berg who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for
the supervision of completion of the work, not for the contents.
Table of contents
I. Introduction 3
II. Literature review 5
1. Unconventional Monetary Policy 5
2. Transmission UMP determinants 6
3. Disaggregated capital flows 8
4. Related studies 9
III. Empirical model 10
1. Research methodology 10
2. Basic model presentation 15
3. Data 15
4. Econometric considerations 17
IV. Estimation results 20
1. Test results 21
2. Summary and discussion of empirical findings 23
V. Conclusion 26 VI. References 28 VII. Appendix 32 1. A: Sample countries 32 2. B: Data sources 32 3. C: Descriptive statistics 35 4. D: Econometric tests 36
3
1. Introduction
During the 2008 financial crisis, capital inflows to emerging market economies (EMEs) came to an all-time low, they recovered again in the aftermath of the crisis and the unconventional monetary policies (UMPs) of advanced economies (AEs) were held responsible for it. Over the course of these years, the ECB succeeded in almost doubling its balance sheet to €4 trillion. Although main goal was to be improving conditions in Europe, the program had serious implications for EMEs. Investors were facing depressed returns in Europe and started to seek for alternative investment opportunities. This alternative was found in EMEs as they had enjoyed high growth rates in the past decade (Figure 1), where portfolio inflows where mainly affected. Since the effects of these flows on the recipients’ economy are difficult to measure, this study concentrates on the effect of monetary policy on portfolio inflows towards them. Research on this topic is not new. However, literature mainly focussed on the effects of US policy for regions such as Latin-America and Asia. Therefore, this research augments to the existing literature by focusing on the effects of ECB policy on various EMEs worldwide. Research on this topic could offer outcomes for future economic policy and such for future welfare for EMEs, such that this study conducts its research around the following question:
‘What are the effects of the ECB’s monetary policies on portfolio inflows to emerging market economies?’
After the recovery of capital inflows to EMEs, FDI still dominated total flows, but portfolio and other investment flows have also increased over time, giving policy makers new challenges about how to deal with such flows (Pagliari, 2017). Capital flows to EMEs can foster economic growth by financing investments. However, are also able to induce excessive monetary policy, currency mismatches and distort asset pricing. Therefore, the benefits of international policy cooperation are under debate. The history of Latin America also gives reason for such concern: the major episodes of capital inflows, during the 1920s and 1978-81, were followed by major economic crises and capital outflows, such as in the 1930s and the debt crisis mid-1980s. After the recent financial crisis, the inflow of capital toward EMEs did recover during the period that UMP was applied by AEs (Figure 1). However, that fact alone is insufficient to attribute this recovery to UMP.
4
Figure 1: Disaggregated net capital inflows into various EMEs
Source: Shaghil Ahmed and Andrei Zlate (2014)
To analyse the above main question, this study performs linear regression analysis on a panel database, containing quarterly data from begin 2000 to the end of 2015 for 10 emerging market economies worldwide. Because different types of capital flows are associated with different motivations, they are analysed separately. In earlier literature, the focus was on net capital flows, but recent literature has stressed the need to study gross inflows and outflows as well since both tend to react in different manner (Forbes and Warnock, 2012). Therefore, the dependant variables will consist of four types of capital inflow. While literature suggests that there are different transmission channels through which UMP effects the flow of capital, different channels will be examined. Besides common channels used in literature such as the liquidity, portfolio rebalance and confidence channels, a fourth will be added, which captures the exchange rate channel. Like the model used by Kucharcuková et al (2016), this study will use an indicator that estimates the stance of ECBs monetary policy, namely the monetary conditions index (MCI) and several control variables that are brought forward by Koepke et al (2014) which will be discussed later on. As this paper is interested in variables that vary over time and to allow for differences in country specific characteristics, the tests include fixed effects.
The remainder of this paper will be as follows. Section 2 provides theoretical background on the transmission of UMP. Moreover, determinants of capital flows will be discussed to retrieve relevant variables in further testing. Section 3 will present the empirical model and the dependent-
5 and independent variables for this study. In section 4 estimation results will be presented and discussed. Concluding remarks and further recommendations will be given in section 5.
2 Literature
A wider body of literature has focused on the effects of unconventional monetary policies. These studies include the rationale for UMP in AEs, the particular case of US policy and its cross-border effects on capital flows toward EMEs. Besides providing some relevant determinants of capital flows, this section summarizes different empirical studies that have been applied.
Early 1990s Latin American countries were facing large capital inflows. This was after the 1980s debt-crisis in Latin America, therefore these large inflows of capital could be due to the major economic reforms conducted in the late 1980s. However, in the same period, the US was in trying to recover from recession, accompanied by low interest rates. Calvo, Leiderman and Reinhart (1993) argued that these low interest rates in the US had been driving the capital inflows in the Latin American countries. Results of the principal component analysis they conducted, confirmed this hypothesis. Further support was found by Taylor and Sarno (1997), who also concluded that external factors play a major role in driving capital flows. However, evidence suggesting the contrary was found by Ghosh and Ostry (1993). They found that the domestic economic variables for EMEs were mainly explaining capital flows. With similar method, Chuhan et al. (1998) used panel data for several Asian and Latin American EMEs found a significant effect for domestic factors as well as for external factors affecting monthly gross capital inflows, suggesting both factors to be relevant. There is considerable variation across instruments and also the types of flows considered (net or gross), which will be further elaborated in the coming sections.
2.1 Measuring capital flows
Before assessing some relevant determinants, some clarification is needed on the dependant variable: capital inflows. Current account deficits are typically representing positive net capital flows (Cardarelli et al, 2010). Gross capital measures inward- and outward- investment separately and these types tend to have different factors driving them. For example, the differential in growth rates drive non-residents decisions regarding portfolio investments, while the yield drives the choice of non-residents looking for investments abroad. On the other hand, for other investment flows, interest rate differential influences both residents’ decision to invest abroad as well as non-residents’ decision to invest in the recipient economy. Recent studies have examined the importance of this distinguish
6 between gross- and net capital flows (Forbes and Warnock, 2012; Broner et al, 2013). Furthermore, according to Forbes and Warnock (2012), it is important to make a distinguish between foreign and domestic investors, because they can react differently to shocks. In some cases, their reactions can counterbalance each other, making net flows more stable. In other cases, their actions could in fact magnify each other, making net flows too volatile. The focus in this paper will be on gross capital inflows, supported by Broner et al. (2013), who show that in post-crisis periods gross capital flows tend to react with large volatility and accuracy, especially relative to net capital flows for developing countries. The IMF divides these capital flows into direct investment, portfolio investment, financial derivatives and other investments (IMF BOPS, 2016), which will be presented same way in this study.
2.2 Transmission channels of monetary policy
Literature on the drivers of capital flows is based on theory where expected returns, risk, and risk preferences matter (Ahmed and Zlate, 2014). The drivers are usually divided into global ‘push’ factors and domestic ‘pull’ factors. The push factors represent external conditions that attracts agents to invest in a country, while the pull factors represent the domestic conditions that affect the risk and return of those agents. Key insight from the literature on this topic brings that push factors seem to influence whether capital inflows occur, while pull factors determine where these flows go and with which magnitude (e.g. Ghosh et al, 2014 and Ahmed and Zlate, 2014). Some major determinants of capital flows are described by Koepke et al. (2014). Risk aversion, AE interest rates and AE output growth tend to be most relevant push factors. Most important pull factors are found to be EME output growth, asset returns and domestic risk. These factors are also supported by Ahmed and Zlate (2014). Their research is conducted around a regression of panel data on net capital inflows since 2002. They find evidence that growth- and interest rate differentials and global risk appetite are significant in explaining capital flows. Where especially the portfolio rebalancing channel is considered driving the effect of UMP on capital flows. In addition, Lim et al. (2014) examined gross financial inflows to EMEs between 2000 and 2013 for the US. They divide the transmission of UMP into three channels: liquidity, portfolio balancing, and confidence. Which is also supported by Chen et al. (2012) and Lavigne et al. (2014). Where the latter also finds support for the existence of an exchange rate channel. Therefore, this study will perform tests on the relevance of these four channels.
7 Liquidity channel
The purchasing of long term assets done by the ECB increased the reserves on the balance sheets of banks. Since these reserves are more liquid, banks are enabled to extend more credit to investors. This reduces the cost of borrowing and increases bank lending, including lending to EMEs and provides the rationale for existence of the liquidity channel (Gagnon et al., 2011). Bruno and Hyun Song Shin (2014) conducted VAR exercises to shed light on the domestic and international transmission channels of monetary policy through the bank-to-bank lending. They found supportive evidence that increasing lenient conditions on local banks provides more lenient conditions to the recipient economy. Thus, a rise in availability of credit is being transmitted through the interactions of global banks.
Portfolio rebalancing channel
UMP is effecting the portfolio composition of existing portfolios. These now bear lower risk and thus lower returns. Unmet risk appetite will result in increased demand for riskier investments. EMEs are typically rich of these kind of investments, which makes the portfolio rebalancing channel relevant. This channel seems to be most relevant in assessing changes in capital movements due to risk and return factors. In the study of Lim et al. (2014) the portfolio rebalancing channel seems to drive results, while the effect of FDI appears to be insignificant. The primary measures indicating this are the yield curve and the interest differential. The yield curve is a global push variable that represents the term structure of interest rates. In their study the interest rate differential between EMEs and the US is used to capture country specific reasons for investors to invest. Besides these primary measures Nier et al., (2014) find that high AE growth rates are found to be negatively related with capital inflow, although this does not has to be true since high growth comes along with more wealth, which would then indeed could lead to risings in foreign investments (Baek, 2006). Because capital movements where notices in response to policy changes, lowering returns in Europe, changing the portfolio compositions, this study’s focus will be on the portfolio rebalancing channel.
Confidence channel
The famous speech of president Draghi: ‘Within our mandate, the ECB is ready to do whatever it takes to preserve the euro. And believe me, it will be enough’ shows the managing of expectations through the confidence channel (ECB media, 2012). To assess the impact of ECB policy shocks on the euro area, Kucharcuková et al (2016) used the MCI and its sub-components for VAR on capital flows toward EMEs. They found the management of inflation expectations and hence the confidence channel as an important measure for the ECB in engaging UMP. Such signalling can lower deflation risk and provide an environment for more investment domestically as well as abroad (Bauer & Rudebusch, 2013).
8 Exchange rate channel
Bouraoui (2015) concludes that the real exchange rate (RER) of various EMEs depreciated significantly relative to the US dollar, driven by the tapering announcements of the FED. The relevance of the exchange rate as an independent variable is also recognized by Prema-Chandra Athukorala and Sarath Rajapatirana (2003). They examined the relation between the RER and capital inflows for Asian and Latin American EMEs from 1985 to 2000. The degree of appreciation was much higher in the Latin American countries than in the Asian countries. This is explained by the fact that non-FDI capital flows tend to drive this RER appreciation and the Latin American countries saw especially this kind in of inflow. This could be due the fact that Asian countries used more fiscal contractions and adjusted their nominal exchange rate to prevent the RER from appreciating. As a rationale behind including the exchange rate as an independent variable in this study, the credit risk of a borrower is the ‘home’ country is considered. An appreciating currency will raise this borrower’s balance sheet through effecting the value of its foreign holdings. Now loans of this agent can be repaid easier, such will lead to declining financing constraints in the ‘home’ country, which tend to attract more capital (Nier et al. 2014). However, this channel can suffer from reversed causality, since capital inflow effects the RER as well (Tillman, 2014).
2.3 Disaggregated capital flows
Changes in monetary policy has different effects on the aggregated capital flows. The types of capital flow used in this study represent different types of investment decisions which obviously have different rationale, where durations, risk and motivation are relevant. This section will give the rationale for direct investment, portfolio investment, financial derivatives and other investments, which can be relevant in interpreting results.
Foreign direct investment (FDI) is widely considered to be the most stable type of capital flow. It consists mainly of fixed assets, is illiquid and not very attractive in crisis since it involves a factor of ownership for the asset invested in. Yet, in crisis this type of inflow remains fairly constant. A depreciation that often accompanies a crisis can increase the profitability of many types of direct investments. FDI is influenced by long-term profitability expectations regarding a country's fundamentals rather than by speculative forces and interest rate differentials. In addition, the distinction between portfolio flows and FDI can be somewhat arbitrary, according to the International Monetary Fund's (IMF) classification, an equity investment above 10% is considered FDI. FDI not only capital is transferred, but also types of knowledge and technology such that FDI tends to incorporate
9 mostly more permanent (longer lasting) investments. The factors influencing these long term investments thus are likely to give long term prospects about the state of economy.
When a purchase of equity is too small, then the purchase is classified as a portfolio flow. Portfolio investments do not entail active management. Rather, the purpose of the investment is solely financial gain, in contrast to FDI, since it does not incorporate a degree of managerial control. Portfolio flows consist of both bond and equity investments. This allows this type to more reactive to short term changes in factors. Portfolio flows tend to be driven by external factors, such that this type of flow is expected to be reacting with high volatility. This is explained by Baek (2006), he calls this type of flow ‘hot’ since it reacts with greater volatility and has a speculative nature.
Financial derivatives cover financial instruments that are linked to other specific financial instruments, indicators or commodities. Through financial derivatives, financial risk (such as interest rate risk and credit risk) can in their own right be traded on financial markets.
Other investments cover all investments that do not fit elsewhere. However, these tend to consists largely of bank loans (bank to bank and private), for instance in order to allow for liquidity used for trade (Rey, 2013).
2.4 Related empirical studies
From related literature is becomes clear that the way of modelling depends on the variables of interest. Quite straightforward is the difference in modelling between domestic and international effects of monetary policy. However, there is not a clear consensus in how to model monetary policy when studying the determinants of international spill overs. Policy interest rates are often used as measure indicating monetary, but since the zero lower bound is reached this does not seem to be an appropriate measure anymore (Wu and Xia, 2016).
More recently, economies are practising unconventional monetary policy. Literature predominantly uses event studies to examine the effects of UMP on financial market variables. This is done with time variables indicating the tapering announcements of central banks (Joyce et al., 2011). Studies that aimed to analyse the drivers of capital flows, more often used regression analysis including fixed effects (e.g. Chuhan et al., 1998; Ahmed and Zlate, 2014). Since this paper conducts its study around the determinants effecting capital flows, the latter, linear regression on panel data including fixed effects seems more appropriate. Therefore, this study follows the approach by Ahmed and Zlate, 2014, but is parting from it in several ways. They conducted their study using only portfolio inflows as the dependent variable, where this study will not only focus on this type of flow, but also on the other types of capital flows as suggested by the IMF, namely: direct investment, financial derivatives and
10 other investments. This study includes various variables suggested by Koepke (2014) (which will be presented later on), but will also focus on defining a way of measuring UMP. Instead of using a dummy variable or tapering announcements, this study will use an indicator for the overall stance of UMP from the ECB. As did Kucharcuková et al (2016), who composed this proxy measuring the overall instance of UMP in the euro area. This section presented several approaches and methodologies, in the next chapter the modelling of this study will be explained in further detail.
3 Empirical model
The main focus of this study is to investigate the effects of ECB monetary policy on capital flows to EMEs and is therefore related to the effects of monetary policy and the determinants of capital flows internationally. The previous chapters summarized the theory on this areas of interest and the aim of this chapter is to provide an empirical model, by considering the different research methods used in previous literature and addressing potential econometric issues. Following the IMF’s definition of capital flows, four types will be examined: direct investment, portfolio investment, financial derivatives and other investments. Besides the MCI, control variables are included to represent standard determinants of capital inflows. In order to come to conclusions, the relationship between capital inflows and some of the above mentioned determinants is tested.
3.1 Research methodology
In order to estimate the effects of the ECB QE on capital flows toward EMEs, a linear regression will be performed. Based on panel data for 10 EMEs (as can be seen from table 1), this brings a broad scope of interest. These countries represent EMES from various regions, namely: Asia, Latin America, and Europe. Since capital flows are separated into four groups, each group is separately examined as a dependant variable and regressed on all relevant variables. Because these tests include data on different countries, individualities have to be accounted for, such that the regressions will be based on fixed effects models. The model will be based on the model presented by Ahmed and Zlate (2014), will include variables suggested by Koepke (2014) and the MCI variable measuring UMP as suggested by Kucharcuková et al (2016). In the next section the basic model will be presented, followed by the representation of all variables. Below a table of all EMEs examined in this study is provided:
11
Table 1: Sample countries, 23 EMEs
Asia-Pacific Europe Middle East, and Africa Western Hemisphere
Indonesia Czech Republic Brazil
South Korea Poland Colombia
Thailand Turkey Peru
South Africa
Capital flows
The types of capital flows will be examined separately since each type is effected for different reasons. As proposed by Ahmed and Zlate (2014), this study divides capital inflows into direct investment, portfolio investment, financial derivatives and other investments. The IMF defines net capital flows as the difference between assets and liabilities. However, capital flows can be distinguished into purchases by resident agents and those of non-resident agents (Forbes & Warnock, 2012). The determinants influencing the decisions of residents might be different from those affecting non-residents. One could think of the latter being more affected by global shocks, where residents tend to have more knowledge and might be more influenced by domestic shocks. This captures the relevance of gross capital inflows, for examining investment behaviour. Looking back at the research question, it also becomes obvious that the behaviour of global investors is of interest and therefore disaggregated gross capital flows into the recipient country represent the dependent variables. In addition, all flows will be divided by the individual country’s gross domestic product (GDP), such that differences in flow magnitudes are accounted for, these could arise due to differences in the size of the recipient economy (Ahmed & Zlate, 2014).
12
Figure 2: Gross disaggregated capital inflow as a percentage of GDP for all 10 EMEs used in this study
Source: Based on data from IMF BOPS (2017)
As can be seen from figure 2, gross capital inflows have increased between 2000 and the present. Representing almost 3% of total GDP before the crisis in 2008 Q1. Then in the crisis, capital inflows dropped and even led to disinvestments until the beginning of 2009. Total capital inflows recovered in the aftermath of the crisis, but did not reach the 3% heights as before. In this period flows seem to be fairly stable, ranging (on average) between 1%-2% of total GDP. At last, differences between types of flows can be observed. FDI, portfolio and other flows accounted for a large share of total inflows before the crisis hit. FDI flows were positive before, in and after the crisis and seems to be unaffected. In the crisis all other types of flows dropped significantly and even became negative during some quarters. Especially financial flows (who was almost non existing) had strong negative impact on total flows in the crisis. After the crisis the share of portfolio inflow seems to be most affected, relatively representing the largest type of flow after crisis, while before it represented the smallest. Following the literature provided in section 2.3, these statistics seem to show exactly why Baek (2006) talked about ‘cold’ and ‘hot’ flows. FDI is persisting and portfolio flows seem to be more reactive.
Monetary Policy (MCI)
Since the focus of this paper is on the MCI variable, this variable will be discussed in further detail. Various different indicators have been used in estimating monetary policy over time. Using monetary aggregates as an indicator of overall policy stance seems to be part of the past, due to the endogeneity of money and the unstable relationship between money and income (Kucharčuková et al., 2016). In academia the interest rate set by the central bank become known as the best mean of gauging the
13 monetary policy stance (Bernanke and Blinder, 1992). However, since economies are passing the ZLB, the interest rate lost its relevance as a good indicator. Credit provisioning, large-scale asset purchases, forward guidance and maturity extensions are some examples of what is not longer reflected by official rates anymore. Therefore, central banks need to be able to rely on something else.
The literature on unconventional monetary policy had come with several solutions to this problem. One solution is using a proxy, for example governmental bond spreads. These are able to take on negative values, solving the ZLB issue. This has been mostly done for the US and not for Europe and this not without its reasons. According to Kucharčuková et al, 2016 the relative low usage of these kind of bonds in the euro area makes this approach not suitable in examining the actions of the ECB. Besides, Lombardi & Zhu, 2014 recognize that these kind of spreads are likely to be distorted by many other factors, giving this approach a major measuring error. Since unconventional policies tend to increase the size of the balance sheet, changes in the central bank’s assets could represent another proxy. However, as central banks tend to communicate these kinds of transactions long before they actually are conducted, the change in size of the balance sheet seems to have the wrong timing. As Wright (2012) explains, in efficient market the impact occurs at the moment some transaction is announced. This brings forward the rationale for measuring the forward guidance of the central bank. In literature this is mostly done with event studies. Focusing on the announcement effect of unconventional policies (Eser & Schwaab, 2016). This method is useful in estimating effects on the financial market and highly dependent on the availability of high frequency data. Furthermore, event studies have the disadvantage of focusing only on certain single measures of unconventional monetary policy, where a measure representing various indicators would be more relevant.
Other papers have attempted to compute a so-called shadow rate. Such a rate should capture the movements of the policy interest rates in times when the ZLB is not binding and also give a good impression of changes in the overall policy stance at the ZLB. Such synthetic indicators have been conducted before such as Wu and Xia (2014), who employed an approximation that makes a nonlinear term structure model and show that such a model offers a description of the data compared to the benchmark model and can be used to summarize the macroeconomic effects of UMP.
Approaches based solely on interest rates thus seem problematic for the euro zone, due to the limitations in the availability of governmental and corporate bond spreads. The financial crisis effected the way risks are perceived as well within the EMU. This crisis showed the heterogeneity among country risks and thus interest rates. Besides interest rates, using measures on the ECB balance sheet just seems too simple. The ECB implemented a wide variety of instruments and therefore a broader approach is needed in measuring the overall monetary policy stance.
14 This study will use the MCI measure conducted by Kucharčuková et al, 2016. They followed a similar approach as Lombardi and Zhu (2014), who used factor analysis to construct a measure of the overall monetary conditions (MCI) for the US. This index on the overall stance of monetary policy is for the ECB, based on interest rates and spreads, monetary aggregates, selected ECB balance sheet items and additionally the exchange rate and is shown below in Figure 3.
Figure 3: MCI indicator in comparison with the Euribor rate
Source: Based on data from Kucharcukova et al. (2016) and the ECB Statistical Warehouse (2017)
Fig. 3 plots the synthetic indicator of the overall monetary conditions (MCI) in the euro area next to the 3-month Euribor for comparison. Until the start of the financial crisis in 2007, the MCI seems to follow the 3-month Euribor closely. In this pre-crisis period the policy interest rate setting was the major instrument of the ECB in effecting market conditions, the MCI indicator seems to represent overall monetary policy very adequate. However, in 2002 there seems to be some deviation, this seems to relate to the increase in monetary aggregate after the euro cash changeover in January 2002 (Peersman, 2011). After the global financial crisis, the MCI significantly deviates from the short term interest rates. From 2009 onwards, the transmission of monetary policy has changes, unconventional measures were introduced. In 2011 the Securities Market Programme (SMP) and the Long-Term Refinancing Operations (LTRO) programme were introduced and we see an easing in monetary conditions looking at the movement of the MCI. Despite Mario Draghi’s famous speech in 2012, monetary conditions seem to be tightening here. According to Kucharčuková et al (2016), the decrease in the ECB’s balance sheet due to prepayments of LTRO loans induced this decrease. Finally, as from 2014 we observe significant policy easing following the implementation of targeted LTROs.
-2% -1% 0% 1% 2% 3% 4% 5% 6% 3M Euribor MCI
15
3.2 Basic model presentation
Following the literature, the determinants of capital flows will be divided into global ‘push’ and domestic ‘pull’ variables. At first a basic model will be presented in equation (1), representing this distinguish:
𝑦𝑖,𝑡= 𝛽0+ 𝛼𝑖+ 𝛽1𝐺𝑙𝑜𝑏𝑎𝑙𝑡−𝑛+ 𝛽2𝐷𝑜𝑚𝑒𝑠𝑡𝑖𝑐𝑖,𝑡−𝑛+ 𝜖𝑖,𝑡 (1)
The dependent variable 𝑦𝑖,𝑡 represents the type of capital inflow compromising direct investment,
portfolio, financial derivatives and other investments to country i during time period t, as a fraction of country i’s nominal GDP. This is modelled by a function including fixed effects 𝛼𝑖, a vector representing
‘Global’ factors or ‘push’ variables and a vector representing ‘Domestic’ factors or ‘pull’ variables which both depend on country i and time t-n indicating that lagged values have been used for some variables. Table 2 provides the independent variables. Including the overall recognized push and pull factors, some of the pull factors are composed of differentials vis-à-vis the EU’s characteristics (interest- and growth rates), push factors are: global risk aversion and an indicator for the stance of monetary policy. The pull factors consist of: market capitalizations, current accounts and the exchange rates of all individual countries. Together they capture the factors influencing capital flows.
Table 2: Determinants
Push/Pull Push Pull
Growth rate differential Global risk aversion Market capitalisation
Monetary policy Exchange rate
Current account
3.3 Data
As discussed in the literature part four types of capital inflow will represent the dependent variables in this study. Following the existing literature, separate regressions will be employed on relevant ‘push’ and ‘pull’ variables. In this section, the model and variables employed in the present study are discussed in more detail.
The panel database consists of quarterly data from 2000Q1 to 2015Q4, for all 10 EMEs. This time period includes the pre- and post- crisis periods, such that more data points can be used in regressions. The ten EMEs that are chosen are usually used in literature, but since the transmission of
16 the monetary policy of the ECB is examined, some European emerging countries are used as well. The country sample consists of ten countries that are from various regions around the world. Namely, Asia Pacific: (Indonesia, South Korea and Thailand); Europe Middle East, and Africa: (Czech Republic, Poland, Turkey and South Africa); Western Hemisphere: (Brazil, Colombia and Peru).
Data on the dependent variable capital inflows come from the IMF Balance of Payments and International Investment Position Statistics (IMF BOP/IIP, 2017), which presents quarterly data for all countries. Data on the independent variables for this study comes from various sources. With all data being quarterly, lagged values on the following determinants are used in this study: MCI, growth rate differentials, MCAP and the VIX. For exchange rates and current accounts, the normal values are used. Data on the interest rate differential is composed of the short term interest rates of all individual EMEs, minus the policy rate of the ECB (Euribor). Short term interest rates for all countries are mostly retrieved from the OECD (2017) and de remaining (Peru and Thailand, since they are not included in the OECD countries) from the Worldbank (2017) and the Euribor is retrieved from the ECB Statistical Warehouse (2017). The interest rate differential is excluded from regressions later on, which will be explained by the econometric tests later on. The growth rate differential is composed of all individual GDP growth rates taken from the IMF IFS (2017), minus the average growth rate of the G-7 countries’ GDP. Where the nominal GDP values are taken from the Worldbank. Global risk aversion is measures by the VIX, which is retrieved from the CBOE (2017) and data on the MCI variable is obtained from Kucharčuková et al. (2016). Trade openness and market capitalization (MCAP) are also taken from the Worldbank (2017), but data provided from this source is only available on yearly base, such that this data is assumed to be constant throughout the year. Trade openness is also deleted from regressions later on, this will also be explained in this chapter. Data on the current account is obtained from the IMF BOPS (2017) and the quarterly exchange rates of all individual countries are collected from the Federal Reserve Bank of St. Louis (FRED). While most data is obtained on quarterly base, the data taken from the Worldbank is only available on yearly base. For this study these measures are held constant throughout the year. Yearly data is therefore expected to not explain much of the variation in capital flows. However, nominal GDP rates are only used to scale capital inflows and trade openness, so not many variables are affected by this. Yet, growth rates in GDP for all countries are quarterly obtained, since these where available at the OECD. All variables are thus in percentages, except for the exchange rates. Therefore, the logarithms of exchange rates are taken in order to account for unit root. Growth rates, MCAP and exchange rates are also corrected with an HP filter. Overall the panel is very balanced.
17
3.4 Econometric considerations
Multicollinearity
Multicollinearity is defined as correlation among variables. It does not directly violate OLS assumptions, but can induce problems in retrieving good estimates on all parameters involved. According to Badi Baltagi (2008), using panel data should eliminate this problem. Panel data has the advantage of giving more informative data, more variability, less collinearity among variables, more degrees of freedom and more efficiency. However, interest rates and the MCI are likely to correlate with risk measures (VIX), also trade openness and the current account could be related to each other since both measure some kind of international openness. Multicollinearity will nevertheless be tested for by looking at the correlation matrix of these variables.
Figure 4,5 & 6: Relation between the interest rate differential à-vis the MCI, Trade openness vis-à-vis CA (as % of GDP) and the VIX vis-vis-à-vis MCI
-6% -4% -2% 0% 2% 4% 6%
18 Sources: Based on data from the ECB Statistical Warehouse (2017), Kucharčuková et al (2016) & the CBOE (2017)
As can be seen from figure 4, the interest rate differential seems to have exactly the same shape as the MCI, but mirrored against the x-axis. Now, this seems fairly plausible since the interest differential is measured as the short term interest for an EME minus the short term interest in the Eurozone (Euribor). Even after the implementation of unconventional measures these two variables seem to move in exactly opposite direction. Therefore, multicollinearity is expected and will be further tested for. In figure 5 the trade openness and current account (as % of GDP) are presented. The relationship among these two variables is harder to see directly. However, both tend to be fairly constant over time, with the exception of moving in contradicting directions during the crisis period (as can be seen
-10% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90%
Trade openess Current account (as % of GDP)
-10% 0% 10% 20% 30% 40% 50% MCI VIX
19 from the changes in levels in 2008 and 2009). At last in figure 6, global risk aversion defined by the VIX and monetary policy (MCI) have a similar upward movement during the crisis and both tend to have greater volatility during the latter period. As a result, over the entire sample period there is a negative correlation of the VIX regarding the MCI variable, as also can be seen from the correlation matrix in table 3 below (for the total correlation table see appendix C.2.1). In table 3 some relevant correlations are presented.
Table 3: Correlation matrix for VIX, MCI, Euribor, interest rate differential, trade openness and CA MCI VIX Interest rate
differential Trade openness Current account (as a % of GDP) MCI 1 VIX -0.3237 1 Interest rate differential -0.6109 0.0576 1 Trade openness -0.4408 0.6648 -0.2039 1 Current account (as a % of GDP) 0.0717 0.3341 -0.6031 0.7984 1
From table 3 all underlined values are most likely to give problems in further regressions. As expected the interest differential and the MCI are highly negatively correlated, with a value of -0.6109 and thus likely to suffer from multicollinearity. The interest rate differential is also highly correlated with the CA’s of sample countries (-0.6031). To resolve potential problems with this variable the interest rate differential will be omitted from the regression, since it measures (almost) the same as the MCI does and is not of big importance in this study, where the MCI is. From table 3 it can also be seen that the trade openness of sample countries indeed correlated too much with the VIX and the CA. Therefore, also the trade openness will be omitted from the regressions, such that the standard deviations of all other relevant variables will not be effected by multicollinearity.
Unit root
OLS assumptions should be fulfilled. However, concern could arise about the stationarity of a variable when examining a large time-dimension. The existence of unit root, also called stochastic trend, in the variables is a potential problem with time series data. In case of unit root, the variable is
20 stationary and that particular estimator will have a non-standard distribution for its coefficient, leading to a bias. In this study, all variables will be tested for unit roots, to be precise all first lags will be tested since these are the ones which will be used in regressions. Formal methods will be applied including the Levin–Lin–Chu (LLC, 2002) type, where time trends will be included. Since the LLC test needs strongly balanced data, it is replaced by the Fisher test, the latter is applied for the trade openness, currents account and interest rate differential variables. If a variable seems to be non-stationary the Hoddrick-Prescott filter will be applied in order to eliminate the disturbing trend. The results in appendix D show that for most variables there are no unit roots. However, for exchange rates, trade openness and at a 1% level also the MCAP seem to have unit root. Therefore, Hoddrick-Prescott filter will be applied on these three variables to de-trend the data.
Random versus fixed effects
Including fixed effects assumes that the country-specific dummies are unique and not correlated to others. The use of random-effects versus fixed-effects is tested by a Hausman test. Results show that a fixed effects regression is the appropriate way of modelling.
Heteroscedasticity
Heteroscedasticity in the error terms is an econometric problem common for panel data (Baek, 2006). In this case, Ordinary Least Squares (OLS) remains unbiased but is no longer efficient; the t-statistic and confidence interval are not valid (Stock and Watson, 2007). Scaling some variables that are expected to give this problem would give more confidence in avoiding this econometric issue. Therefore, the disaggregated capital flows are divided by GDP, the same holds for the current accounts. To resolve any potential problems regarding this issue, robust errors can be applied. The presence of heteroscedasticity in a fixed-effects model can be tested by a modified Wald test. The results in appendix D show that heteroskedastic error are indeed present, such that robust errors will be applied.
4 Estimation results
After the completion of the tests for multicollinearity, unit roots fixed effects and heteroscedasticity, the estimation results of the fixed-effects regression on disaggregated capital inflows as percentage of GDP into 10 emerging market economies, from 2000 to 2015 are presented in this section. Dividing the flows by GDP will provide data that is corrected for differences in magnitude. The aim of this study is to determine whether monetary policy in Europe contributes to changes in portfolio inflows toward
21 emerging market economies. In the estimation, four types of capital inflows will represent the dependant variables, where portfolio inflow is seen as the most relevant type for this study. The estimation includes several global ‘push’ factors and domestic ‘pull’ factors. From the previous section it follows that standard robust errors should be applied and some variables should be adjusted in order to prevent unit roots. First the natural logarithm is taken of the exchange rates. Then, interest rates, growth rates and exchange rates are adjusted with an HP filter. The first lagged value is used for the MCI, while monetary policy is expected to have a delayed effect (Kiendrebeogo, 2016). First lagged values are also used of the VIX, growth rates and the MCAP.
4.1 Test results
All regression results from the model equation (1) are presented in table 4, with disaggregated capital inflow as percentage of GDP as the dependent variable, adding all different explanatory variables to the regressions. In general, results are significant and the monetary policy index has the expected negative coefficient. The R-squared is fairly low for all regressions, indicating that the relevant variables used in this study only explain a small proportion of total variation in capital flow toward EMEs worldwide. The MCI is significant at explaining the portfolio inflows and the capital inflows labelled FDI, yet it seems not to be an important driver of financial derivatives and other investments. Portfolio inflow is negatively correlated with the MCI, yet the other disaggregated flows are positively correlated. However, the magnitude of the effect among the disaggregated flows seems to be roughly the same. A one percentage point increase in the MCI variable is affiliated with a 0.1125 percentage point decrease in portfolio inflow as a percentage of GDP (significant at the 1 percent level). This seems to be very small, but certainly is not. The average of portfolio inflow accounts for 0.41% of total GDP over the entire sample, such that a one percent increase in the MCI induces on average a 27.44% drop in total portfolio inflow for an EME (holding all other factors constant). Portfolio inflow seems to be also affected by domestic factors; the exchange rate and the current account (as a % of GDP) seem both important drivers of portfolio inflow behaviour. The sign of the exchange rate is positive and indicates that an appreciation is followed by portfolio inflow. In contrast the current account coefficient has a negative sign, indicating the need for finance for the recipient country.
22
Table 4: Estimated effect of ECB monetary policy on disaggregated capital flows toward EMEs
(1) (2) (3) (4)
Dependant variable FDI Portfolio
investment Financial derivatives Other investments Push factors: MCI t-1 0.0609** -0.1125*** 0.0558 0.0048 (0.0244) (0.02850) (0.0360) (0.0403) VIX t-1 -0.0041 -0.0072 -0.0014 -0.0053 (0.0032) (0.0047) (0.0026) (0.0042) Push/pull factors: Growth differential t-1 0.00095 0.0159 -0.0069 0.0132 (0.0192) (0.0091) (0.0071) (0.0191) Pull factors: MCAP t-1 0.0034** 0.0017 0.0004 0.0055*** (0.0014) (0.0013) (0.0007) (0.0016)
Log exchange rate -0.1132 1.1312** 0.8206 1.6912*
(0.2745) (0.4851) (0.9160) (0.8379) Current account -0.0381** -0.0389*** 0.0077 -0.0733*** (0.0151) (0.0133) (0.0133) (0.0078) Control: Constant 0.378** 0.3325 -0.2761* -0.2057 (0.135) (0.1835) (0.1460) (0.1886) 𝑹𝟐 0.0727 0.1197 0.0627 0.1398 F 3.55 9.15 0.9 36.82 Obs 599 599 599 599
Country FE Yes Yes Yes Yes
Robust standard errors in parentheses * significant at p<0.1, ** significant at p<0.05 and *** significant at p<0.01. The dependent variable is defined as aggregated capital inflows as % of GDP, all independent variables are defined in percentages or in percentages of GDP, excepts for the exchange rate which is a natural logarithm. See appendix B for a complete overview.
As can be seen from table 4 monetary policy measured by the MCI is positively correlated with the remaining types of capital inflow. FDI, financial derivatives and other investments seem to increase with 0.0609, 0.0558 and 0.0048 percentage point respectively accompanied by a one percent point increase of the lagged MCI, thus portfolio flows are more reactive to changes in monetary policy, which was expected. These results do not have the expected sign and have to be treated with caution. The coefficients for the MCI on financial derivatives and other investments are insignificant, such that
23 these will not be given further interpretation. However, for FDI the sign is significant at the 5% level. Indicating that tightening policy in Europe decreases FDI flows toward EMEs.
In general, the results of this study support previous empirical literate, also regarding the characteristics of the disaggregated capital flows. Where FDI is affected not only by the MCI, but also by domestic factors (pull factors) such as the market capitalisations and the currents account of the individual EMEs. External factors (besides monetary policy in the Eurozone) not seem to affect FDI inflows. The effects of MCAP on FDI flows seems intuitive, FDI transactions involve big purchases and a way of ownership, such that well a well-developed market increases the attractiveness of such investments. FDI flows were very stable during the entire sample period, including during the financial crisis when risk aversion was high. The VIX does not seem to affect FDI, so results indeed support this view. As a matter of fact, the VIX is insignificant in explaining all other types of inflow. Financial derivatives do not seem to be affected by any factor used for this study, where it would be expected that the MCAP and VIX were able to explain some of these financial flows. The policy setting by the ECB seems to not affect other investments. However, this type of flow is affected by domestic factors; the exchange rate, MCAP and the CA seem important drivers of this kind of inflow. This sign of the current account coefficient for other investments is negative, seems to capture the loss in willingness for an AE to finance the recipient country when the latter becomes a larger debt holder, thus riskier to invest in.
4.2 Summary and discussion of empirical model findings
Several tests were performed before the actual model was estimated. This has been done to come to a reliable model in explaining capital inflows to emerging market economies. All types of flows are divided by GDP to account for differences in magnitudes. From the results in chapter 3.4 and the related tests presented in appendix D, it is concluded that the combination of push and pull factors interest differential, growth rate differential, global risk aversion, monetary policy, market capitalization, exchange rate and currents account is restricted to the difference in GDP growth between the G-7 and the EMEs chosen for this study, VIX, MCI, MCAP, the natural logarithm of exchange rates and the current account as a % of GDP. All determinants are defined in percentages such that no logs are taken there, except for exchange rates. Because initial regressions suffered quite heavily from multicollinearity, the interest rate differential and the trade openness were omitted from the regressions. Statistics on this problem are presented in figure 4,5 and 6 and also in table 3. Unit root tests were applied on all variables, for most variables the LLC test was applied, but the current account was not balanced enough such that the Fisher test was more appropriate to use. These tests
24 pointed out that HP filter had to be applied on the exchange rates and the market capitalization because these suffered from unit roots (as can be seen from appendix D). Then the appropriateness of a fixed effects regression was tested with a Hausman test. Results in appendix D show that the fixed effects regression was indeed appropriate. At last the model was tested for heteroscedasticity, the modified Wald test results from appendix D show that the regression indeed suffers from this phenomenon and to deal with this, robust standard errors are applied in all regressions.
The results from this study’s analysis suggest that expansive monetary policy, global risk aversion and the relative size of a country’s currents account with respect to its GDP are negatively associated with portfolio inflows into emerging market economies while the growth differential, market capitalization and the exchange rate are positively associated with portfolio inflows. The finding that monetary policy seems to play an important role as a determinant of capital inflows is consistent with the literature provided in chapter 2. The direction and magnitude of the coefficient also seem to be quite appropriate. The global risk aversion does have the expected positive sign, but loses its relevance due to a lack of significance in explaining portfolio inflows. In other literature the VIX actually was an important driver of capital inflows like in the research of Ahmed and Zlate (2014), but this is not brought forward by the present study. Yet, the level of the current account as a percentage of GDP does seem to be a relevant factor in explaining portfolio capital inflows into EMEs, with the expected negative relationship towards portfolio capital inflow. Now the growth differential does not seem to be relevant in the present study, but was expected to be so since in most literature this determinant seemed to be an important factor. The positive sigh however was expected even though this determinant does not seem important in this study. The sign of market capitalization is positive, which seems to be logical following relevant literature and seems to be only relevant in explaining FDI and other investments into EMEs. The quarterly data on this variable was assumed to be constant throughout the year since the World bank any presents this data on yearly base. From chapter 2 we can conclude that transactions for FDI mainly driver by longer term investment decisions, where the portfolio inflows tend to be far more reactive to changes in factors. This could explain the relevance of the ‘yearly constant’ market capitalization for FDI relevant to portfolio inflows. The positive relationship of exchange rates with portfolio inflows follows the empirical findings in the literature part, but could still suffer from reversed causality. The appreciating currencies could indeed be attractive to foreign investors, but the investing of those foreign investors could also induce this appreciation. In this study no clear distinction is made between these theories, which would be interesting to examine, but is not that relevant for this study, such that this is let open for further discussion. In table 5 the results are summarized with respect to their relevant effects.
25
Table 5: Effects of determinants on aggregated capital inflows to 10 EMEs
Type Determinant FDI PF FD OI
Push: MCI + - ? ? VIX ? ? ? ? Growth diff. ? ? ? ? Pull: MCAP + ? ? + Exchange rate ? + ? + CA - - ? -
Note: FDI stands for foreign direct investment, PF for portfolio inflow, FD for financial derivative inflow and OI for other investments. ‘+’ indicates evidence for positive relationship, ‘?’ an unclear relationship, ‘-‘ evidence for a negative relationship.
With an average 𝑅2 of 0.1197 for portfolio inflows the determinants used for analysis do not seem to explain that much, yet this study tried to find the importance of monetary policy as a driver of portfolio capital inflows into emerging markets, such that the key finding as of more relevance than the overall quality of the model. Namely, the key finding is that there indeed is evidence of spillovers of accommodative monetary policy in the Eurozone in the form of portfolio flows into emerging market economies in the period 2000 to end 2015. It should be noted that the ECB was not the only central bank that turned to more accommodative monetary policy after the financial crisis. The gross of literature for the effects of monetary policy of other central banks as for instance the FED find similar results, such that this study augments these studies by adding the central bank of the Eurozone as an important player in the field of international spillovers. Jaime Caruana, General Manager of the Bank for International Settlements stated that spillovers of accommodative monetary policy are ‘widely recognised’ in his speech at the bank’s general annual meeting in Basel, 2015. This claim is (at least for the policy of the ECB) confirmed by this study.
26
5. Conclusion
With emerging market economies (EMEs) criticizing advanced economies’ (AEs) accommodative monetary policy in reaction to the financial crisis and the resumption of capital inflows into EMEs during that same period, debate set in whether these policies could have exerted negative externalities. Several events in history justify the fear of a reversal in capital inflows followed by major economic crisis for the initial recipient country. Economic theory as well as empirical findings from relevant literature confirm that the monetary policy of AEs indeed can have effects on international capital flows, specifically towards EMEs. However, the majority of empirical research has been focussing on the effects of US unconventional measures toward emerging markets from either Latin America or Asia.
This study focussed instead on the effects of monetary policy exercised by the ECB on the more volatile capital inflows towards EMEs worldwide. In addition, this study uses a measure for the overall instance of monetary policy for the ECB conducted by Kucharčuková et al. (2016), instead of using a dummy or shadow rate as is regularly used in other empirical research. Panel data techniques describe quarterly data on aggregated capital inflows for a set of 10 EMEs worldwide and all relevant push and pull factors driving those flows in the period 2000 to 2015. Portfolio flows are of main interest of the analysis as these flows are historically most volatile, thus will be most reactive to changes in external factors and bring most challenges regarding potential reversals of inflows.
Portfolio inflows as a percentage of GDP into 10 EMEs were regressed on the lagged value of the MCI as a proxy for the overall monetary policy stance in Europe, controlling for several domestic ‘pull’ factors and global ‘push’ factors including the lagged VIX, growth differential, market capitalization and the values for exchange rates and current accounts of all individual countries.
The key finding of this empirical research for a 64-quarter period suggest that, for a given EME, portfolio inflows as a percentage of GDP are negatively associated with accommodative monetary policy in Europe. However, the effect on other (less volatile) type of flows differs in magnitude and direction. After the conduction of multiple robustness checks, it seems that accommodative policy by the ECB only has a positive significant relationship with FDI, where the effect on the other types of capital inflows stay remains unclear. Furthermore, results show that the VIX is negatively associated with all types of inflows, but only seems to be relevant for the ‘other investments’ and is not significant in explaining portfolio capital inflows. There is no empirical evidence for the growth differential to be relevant for any type of inflow. The market capitalization has been found to have positive effects on FDI and other investments, the exchange rate of an individual EME is found to be positively associated with both portfolio and other investment inflow and the current account of an emerging market is
27 negatively related to all type of inflows except for financial derivatives. These results suggest that the accommodative monetary policy exercised by the ECB was (also) responsible for the inflow of capital toward EMEs, in particular portfolio inflow seems to be the most important type of flow.
A few limitations of this study are important to notice. First of all, the period ranges from 2000 to the end of 2015 since quarterly data on capital flows was only available to this date. After 2015 the ECB expended their asset purchase program, such that it would be interesting to examine which international effects these brought. Some data used in this study could only be retrieved in very low frequency, but very later on omitted from the regressions. However, the rest of the data was measures quarterly, yet portfolio flows are able to react very responsive to short term changes in external factors, such that even quarterly data could be too low-frequency. Thirdly, the international spillover effects of monetary policy are transmitted not by capital inflows only. However, to make this discussion on capital inflows more extensive, research is needed on the effects of capital surges on EMEs their real economies and financial stability.
These results provide a call for attention to policy makers in emerging market economies. Improving institutions and deepening the financial markets to protect their countries against large capital in- and outflows. It is complicated to study this subject in more detail since data is limitedly available. Further research is needed to determine whether the current dynamics of capital inflows are going to continue and what kind of implications these will have for the financial stability and economies of the recipient countries.
28
References
Ahmed, S. & Zlate, A., (2014). Capital flows to emerging market economies: A brave new world? Journal of International Money and Finance 48
Athukorala, P. C., & Rajapatirana, S. (2003). Capital inflows and the real exchange rate: a comparative study of Asia and Latin America. The World Economy, 26(4)
Athukorala, P. C., & Rajapatirana, S. (2003). Capital inflows and the real exchange rate: a comparative study of Asia and Latin America. The World Economy
Baek, I.-M., (2006). Portfolio investment flows to Asia and Latin America: Pull, push or market sentiment? Journal of Asian Economics
Baltagi, B. (2008). Econometric analysis of panel data. John Wiley & Sons.
Bauer, M. D., & Rudebusch, G. D. (2013). The signalling channel for Federal Reserve bond purchases.
Beirne, J. (2012). The EONIA spread before and during the crisis of 2007–2009: The role of liquidity and credit risk. Journal of International Money and Finance
Bouraoui, T. (2015). “The effect of reducing quantitative easing on emerging markets” Applied Economics
Broner, F., Didier, T., Erce, A., & Schmukler, S. L. (2013). Gross capital flows: Dynamics and crises. Journal of Monetary Economics
Bruno, V., & Shin, H. S. (2015). Capital flows and the risk-taking channel of monetary policy. Journal of Monetary Economics
Calvo, G. A., Leiderman, L. & Reinhart, C. M., (1993). Capital Inflows and Real Exchange Rate Appreciation in Latin America: The Role of External Factors. Palgrave Macmillan Journals on behalf of the IMF
29 Cardarelli, R., Elekdag, S., & Kose, M. A. (2010). Capital inflows: Macroeconomic implications and
policy responses. Economic Systems.
CBOE, (2017). VIX Options and Futures Historical Data. Available at: http://www.cboe.com/micro/vix/historical.aspx. , Accessed on 02/06/2017
Chen, Q., Filardo, A., He, D., and Zhu, F. (2012). “International spillovers of central bank policies” Bank of International Settlements Papers No. 66
Chuhan, P., Claessens, S., & Mamingi, N. (1998). Equity and bond flows to Latin America and Asia: the role of global and country factors. Journal of Development Economics
Caruana, J. (2013), Making the Most of Borrowed Time, at the ‘Annual General Meeting’ of the Bank for International Settlements, Basel, 23 June 2013
ECB Media Speeches (2012). Speech by Mario Draghi, President of the European Central Bank at the Global Investment Conference in London. Retrieved from:
https://www.ecb.europa.eu/press/key/date/2012/html/sp120726.en.html
ECB SDW, (2017). Statistical Data Warehouse. Available at: http://sdw.ecb.europa.eu/ Accessed on 10/06/2017
Eser, F. & Schwaab, B., (2016). Evaluating the impact of unconventional monetary policy measures: Empirical evidence from the ECB's Securities Markets Programme. Journal of Financial Economics 119
Forbes, K. J. & Warnock, F. E., (2012). Capital flow waves: Surges, stops, flight, and retrenchment. Journal of International Economics 88
FRED, (2017). Economic data. Available at: https://research.stlouisfed.org/fred2/. Accessed on 02/05/2017
Gagnon, J., Raskin, M., Remache, J. & Sack, B., (2011). The financial market effects of the federal reserve's large-scale asset purchases. International Journal of Central Banking
30 Ghosh, A. R., & Ostry, J. D. (1993). Do capital flows reflect economic fundamentals in developing
countries?
IMF BOPS, (2017). Balance of Payments and International Investment Position Statistics. Available at:
http://data.imf.org/?sk=7A51304B-6426-40C0-83DD-CA473CA1FD52. Accessed on
15/05/2017
IMF IFS, (2017). International Financial Statistics. Available at:
http://data.imf.org/?sk=5DABAFF2-C5AD-4D27-A175-1253419C02D1. Accessed on 10/05/2017
Joyce, M., Miles, D., Scott, A. & Vayanos, D., (2012). Quantitative easing and unconventional monetary policy. The Economic Journal 122
Kiendrebeogo, Y., (2016). Unconventional monetary policy and capital flows. Economic Modelling 54
Koepke, R., (2015). What Drives Capital Flows to Emerging Markets? A Survey of the Empirical Literature. IIF Working Paper.
Kucharčuková, O. B., Claeys, P. & Vašiček, B., (2016). Spillover of the ECB's monetary policy outside the euro area: How different is conventional from unconventional policy? Journal of Policy
Lavigne, R., Sarker, S. and Vasishtha, G. (2014). “Spillover Effects of Quantitative Easing on Emerging-Market Economies” Bank of Canada Review Autumn 2014
Lim, J. J., Mohapatra, S. & Stocker, M., (2014). Tinker, Taper, QE, Bye? The Effect of Quantitative Easing on Financial Flows to Developing Countries. World Bank Policy Research Paper 6820.
Lombardi, M. & Zhu, F., (2014). A shadow policy rate to calibrate US monetary policy at the zero lower bound. BIS Working Paper No 452
Morgan, P.J. (2011). “Impact of US Quantitative Easing Policy on Emerging Asia” Asian Development Bank Institute Working Paper Series No. 321
31 Nier, E., Sedik, T.S., and Mondino, T. (2014). Gross Private Capital Flows to Emerging Markets: Can
the Global Financial Cycle Be Tamed? International Monetary Fund Working Paper 14/196
OECD, (2017). OECD.Stat. Available at: http://stats.oecd.org/ Accessed on 25/06/2017
Pagliari, M., & Ahmed, S. (2017). The Volatility of Capital Flows in Emerging Markets: Measures and Determinants
Peersman, G. (2011). Macroeconomic effects of unconventional monetary policy in the euro area.
Rey, H., (2013). Dilemma not Trilemma: The Global Financial Cycle and Monetary Policy Independence. Jackson Hole Paper
Taylor, M. P., & Sarno, L. (1997). Capital flows to developing countries: long-and short-term determinants. The World Bank Economic Review
Tillmann, P. (2014). Unconventional monetary policy shocks and the spillovers to emerging markets.
World Bank, (2017). World Development Indicators. Available at: http://data.worldbank.org/data-catalog/world-development-indicators. Accessed on 25/06/2017
Wright, J. H., (2012). What does monetary policy do to long-term interest rates at the zero lower bound? The Economic Journal 122
Wu, J. C. & Xia, F. D., (2016). Measuring the Macroeconomic Impact of Monetary Policy at the Zero Lower Bound. Journal of Money, Credit and Banking
32
Appendix
Appendix A: Sample countries
The sample of countries consists of 10 emerging market economies from the MSCI emerging market index. Chosen such that there is an even distribution between regions and for the countries in Europe it is taken in consideration that countries are not in the Eurozone.
Region Country
Africa South Africa
Asia Indonesia
South Korea Thailand
Latin America Brazil
Colombia Peru
Europe Czech Republic
Poland Turkey
Appendix B: Data sources
Variable Description Frequency Source Dependent variables Disaggregated capital inflows Foreign direct investment, portfolio investment, financial derivatives and other investments inflow in billion USD divided by nominal GDP in billion
33 USD in percentage
points.
Global factors
MCI A proxy of the overall
monetary policy stance of the ECB, an excel sheet with relevant data was provided by Oxana Kucharčuková who works at the Czech National Bank, Prague.
Q Kucharčuková et al.,
2016
VIX A proxy of global risk
global risk aversion measured by the CBOE Volatility Index
Q CBOE, 2017
AE growth Average growth rate
of GDP for the G-7 countries: Canada, France, Germany, Italy, Japan, the United Kingdom and the United States (in percentages)
Q OECD, 2017
Euribor Interbank lending rate in Europe, set by the ECB
Q ECB Statistical Data
Warehouse, 2017 Domestic factors Short-term interest rates EMEs Money market interest rates (3 month) Q OECD 2017 & Central Banks of individual countries (for non-OECD)
