The Impact of Unconventional Monetary Policy on Financial Uncertainty in Emerging Markets

Transcription

1 The Impact of Unconventional Monetary Policy on Financial Uncertainty in Emerging Markets Nathan Converse July 17, 2015 Please do not circulate without the author s permission Abstract This paper studies how the unconventional monetary policies implemented by the Federal Reserve have affected the volatility of financial markets in emerging economies. Since elevated financial uncertainty dampens investment and growth, a significant increase in volatility constitutes an important potential channel through which advanced economy monetary policy affects emerging markets. I examine the conditional volatility of domestic financial conditions and portfolio capital flows in a panel of 18 emerging markets during the two-year period centered on the Fed s announcement of the asset purchase program known as QE3 in late On average, the volatility of bond yields was higher after the Fed resumed large scale asset purchases, as was the volatility of portfolio bond and equity inflows. By constrast, the volatility of equity returns does not appear to have increased. Examining the drivers of financial volatility in detail, I find that while increases in the VIX boosted the financial volatility across the board, the effects of other U.S. and domestic variables vary across equity and bond markets. JEL Classification: F52, G15 Keywords: monetary policy, quantitative easing, emerging markets, volatility, bond yields International Finance Division, Federal Reserve Board; Nathan.L.Converse@frb.gov; The views in this paper are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or any other person associated with the Federal Reserve System. I thank Frank Warnock for his insightful comments and participants at the Fed Conference on Spillovers from Monetary Policy since the Global Financial Crisis for their feedback. Pat Kennedy and Caleb Wroblewski provided valuable research assistance. 1

2 1 Intro The global financial crisis saw capital flows to emerging market economies (EMEs) reverse sharply, from gross inflows of $350 billion in 2007 to gross outflows of $130 billion in While the level of inflows dropped during the crisis, Figure 1 makes clear that their volatility reached unprecedented levels. 1 From mid-2009, foreign capital rapidly moved back into the EMEs. As illustrated in Figure 1, average monthly capital flows to EMEs in the four years after the crisis were indeed higher than the level of the pre-crisis period. The volatility of capital flows to the EMEs, on the other hand, has not reverted to pre-crisis levels. Figure 1: Portfolio Capital Flows to EMEs, Level and Volatility Level of Inflows (US$ Billions) Source: EPFR. Volatility is a rolling 1 year standard deviation Volatility of Inflows Capital Flow Volatility (Right Scale) Monthly Capital Inflows (Left Scale) Annual Average Inflow (Left Scale) Many policymakers in EMEs have drawn a connection between the volatility of international capital flows since the 2008 crisis and the unconventional monetary policies (UMP) pursued by central banks in advanced economies, in particular the U.S. Federal Reserve. Policymakers have voiced concern at the possibility that the tapering of asset purchases might lead to an abrupt reversal in capital flows (as in Rajan (2013) for example) a sudden stop event of the type studied extensively by economists since the Asian financial crisis. At other times during the period, both policymakers and economists have discussed the potential negative effects of surges in capital flows into EMEs. 2 That capital flows both into and out of EMEs 1 Capital flows in Figure 1 are mutual fund flows into EME equity and bond markets. 2 Brazilian President Dilma Rouseff s comments in April 2012 about an impending monetary tsunami are an example. Recent academic work on the topic includes Forbes and Warnock (2012), Reis (2013), Benigno and Fornaro (2014),and Benigno et al. (2015). 2

3 have generated concern in a relatively short period of time highlights their unprecedented volatility during the period in which central banks have resorted to unconventional policy. The rise in capital flow volatility coincides with an increase in economic research on the macroeconomic effects of various types of uncertainty (summarized by Bloom, 2014). 3 Because the increase in the volatility of capital flows is almost entirely the result of larger fluctuations in portfolio capital flows (foreigners purchasing equities and bonds), an obvious channel through which capital flow volatility may affect the real economy is through an increase in uncertainty about future financial conditions. Indeed, recent work has shown how fluctuations in financial uncertainty amplify macroeconomic fluctuations in emerging markets at business cycle frequency (Fernández-Villaverde et al., 2011; Converse, 2014). Moreover, elevated volatility can also depress foreign demand for local currency bonds and thus impede the process of financial market development (Burger and Warnock, 2007). The fact that unconventional monetary policy in advanced economies has coincided with an rise in capital flow volatility thus raises the possibility that such policies negatively affect emerging markets by increasing financial uncertainty. As a first step towards assessing whether this financial uncertainty channel constitutes an important spillover of unconventional monetary policy, this paper examines whether asset purchases by the by the Federal Reserve have resulted in greater financial volatility in emerging economies. I measure financial uncertainty using the volatility of yields on local currency government bonds and local equity market returns, and document how their volatility has evolved since 2010 in a sample of 18 EMEs. To the extent that unconventional monetary policy does affect the volatility of financial conditions in emerging markets, cross border financial flows will act as an important channel of transmission. I therefore also examine the volatility of mutual fund flows to the same set of emerging markets. If financial market conditions in the countries in my sample are more volatile, but fund flows are not, it would suggest that the financial volatility reflects more fundamental uncertainty and not a spillover of quantitative easing via international financial markets. I focus in particular on the two-year period centered on the September 2012 announcement that the Fed would resume large-scale asset purchases with the program known as QE3. This window is particularly suitable for analysis because the Fed paused in its asset purchases after the end of the program known as QE2 in mid-2010 and the beginning of QE3 in September As a result, I am able to study financial uncertainty during one year without an active UMP and a subsequent year during which the Fed made large asset purchases. My focus on financial volatility in emerging markets, rather than the level of financial conditions, motivates my identification strategy. A large literature has showed that economic conditions in the U.S. have significant effects on capital flows to emerging markets and in turn on financial conditions in those economies. Concern about spillovers from unconventional monetary policy stems in part from the belief that such effects are magnified during periods when the Federal Reserve is engaging in large scale asset purchases, due to the large volume of liquidity injected into the financial system. Rather than trying to identify the 3 Throughout this paper, I refer to an increase the second moments of variables as a rise in uncertainty. This is sometimes also called risk, particularly in the finance literature. 3

4 sensitivity of emerging market bond yields or equity returns to specific shocks during normal times and then testing whether that sensitivity changed during periods of quantitative easing, I compare the variance of financial market conditions during the two periods, remaining agnostic about the underlying shocks. To use Brazilian President Rouseff s metaphor, during normal times a pebble thrown into a pond will generate only small ripples. But when liquidity is abundant due to unconventional monetary policy, the ripples may build into a tsunami. Rather than watching for pebbles thrown into the bond and measuring the impact of waves hitting the shore, I seek to measure the amplitude of the waves on the surface of the pond. I find that the volatility of bond yields increased significantly in two-thirds of the EMEs during the first year of QE3. By contrast, the volatility of equity returns was generally lower during the QE3 period than during the previous 12 months. Looking at the volatility of the corresponding capital flows into EME financial markets, I find that bond flows were more volatile once QE3 began than during the previous year. The picture for equity flows is more mixed, with volatility increasing in just over half of cases. In order to determine why financial volatility shifted after September 2012, I regress volatility on real and financial conditions in the U.S. and in the EMEs. Even once I condition on these numerous potential covariates, the average level of bond yield volatility in the QE3 period remains significantly higher. By contrast, the lower volatility of equity returns during the QE3 period appears to be driven by factors other than the presence of unconventional monetary policy. The regression results indicate that both bond and equity flows to the countries in my sample were more volatile under QE3 than during the preceding year, and that this higher volatility cannot be accounted for by the many other factors for which I control. This elevated international capital flow volatility indicates that the greater volatility of bond yields reflects international financial conditions. As a second step towards identifying the channel through which unconventional monetary policy affects financial uncertainty in emerging markets, I test whether the relationship between financial volatility on the one hand and U.S. and domestic economic conditions on the other changed after the announcement of QE3. I find that the impact of changes in U.S. economic conditions on EME bond market volatility increased following the resumption of asset purchases in September By contrast the effects of most domestic economic variables, including those highlighted in the literature as indicators of EME vulnerability, were generally unchanged. This paper therefore makes two contributions. First it shows that unconventional monetary policy was associated with increased bond yield volatility in emerging markets. This suggests that greater financial uncertainty does constitute an important channel for spillovers from UMP. However, the relationship between monetary policy and the volatility of equity markets is much less clear, with factors other than monetary policy dominating. Thus spillovers to financial volatility may appear in some asset markets but not in others. Second, the paper finds that greater uncertainty associated with the onset of QE3 was a consequence of an increase in the importance of so-called global push factors for financial uncertainty in emerging markets. Recent research has shown how global push factors are key determinants 4

5 of the level of capital inflows (Forbes and Warnock, 2012), domestic financial conditions (González-Rozada and Levy-Yeyati, 2008; Rey, 2013), and real variables in emerging markets (Akinci, 2013). This paper contributes to this literature by providing evidence that the effect of these global factors on the volatility of domestic financial conditions (previously documented by Bekaert and Harvey, 1997) increased during the period of unconventional monetary policy. Section 2 of the paper describes the data used to measure financial volatility in EMEs and documents its movements since Section 3 examines to what extent factors other than the onset of QE3 can explain the rise in volatility during the program s first year. Section 4 concludes and discusses avenues for further research. Related Literature This paper contributes to the literature on financial market volatility in EMEs, as well as to the recent literature on the impacts of unconventional monetary policy. Research on the drivers of financial volatility in emerging markets is relatively rare and has not previously examined how wealthy countries monetary policy affects the variance of financial variables. My methodology is adapted from the approach taken by Bekaert and Harvey (1997), who study the determinants of equity return volatility in 20 emerging markets. This paper focuses its analysis on unconventional monetary policy, however, whereas they are primarily concerned with the relative role of local and global factors, a topic which I also address. Rather than attempting to uncover the variables that drive shocks to interest rate volatility, Edwards and Susmel (2003) use a Markov switching GARCH (SWARCH) model to analyze the jumps in volatility observed in Asia and Latin America during crisis episodes in the 1990s. Perhaps closest to this paper is a study of financial volatility during the emerging market crises of the late 1990s by Hayo and Kutan (2005). They analyze how volatility was affected by news regarding the IMF s involvement in the crisis countries. While they look at equity returns rather than bonds, like this paper they focus on the impact of particular policy interventions on volatility. In addition this paper contributes to the large and rapidly growing literature on how U.S. monetary policy, including unconventional policy, affects financial conditions and real variables in emerging markets. My contribution is to examine how unconventional monetary policy affects the second moment of bond yields and equity returns in EMEs, as opposed to their level. Here I discuss only a handful of recent papers most related to my work. Looking at monetary policy when interest rates are away from the zero lower bound, Edwards (2012) shows that changes in the U.S. Fed Funds rate during the period from 2000 to 2008 had significant effects on interest rates in a sample of seven Asian and Latin American economies. Bowman et al. (2014) show that the magnitude of the response of exchange rates, equity prices, and bond yields in EMEs did not increase to monetary policy in the U.S. is not larger when policy is implemented using unconventional measures. Rather regardless of the Fed s monetary policy, the response of financial variables in the EMEs reflect domestic economic conditions. Their approach of interacting U.S. financial shocks with EME domestic variables 5

6 informs my analysis of the channels through which UMP has affected financial volatility. Studies by staff at the IMF also find that while UMP did affect asset prices in non-ump countries, particularly during QE3, the magnitude of these effects was not significantly different than that associated with conventional monetary policy (International Monetary Fund (2013) provides an overview of this work). Again, whereas these and other studies (e.g. Fratzscher et al., 2013) examine how U.S. monetary policy affects affects the first moment of financial variables in EMEs, this paper studies its impact on their second moments. 2 Financial Uncertainty in Emerging Markets 2.1 Data Sources In order to study domestic financial conditions, I compile data on domestic-currency government bond yields and equity returns in a panel of 18 emerging markets between 2010 and Although the original obtained data are daily, I construct a weekly time series using weekly averages. I use weekly data on portfolio bond and equity flows from EPFR. The EPFR data measure flows into mutual funds and ETFs investing in emerging markets, not only dedicated country funds but also global and regional funds. 4 Tables 1 and 2 provide summary statistics for the resulting data set, while further detail on data sources can be found in the appendix. Bond yields range from over 13 percent in Turkey and Brazil to 1.3 percent in Singapore. Malaysia, Thailand, and Singapore had the most stable yields with standard deviations of 0.5 percentage points or less, while the Philippines, Indonesia, and Colombia showed the most volatility. Thus in terms of volatility, the vast majority of countries in the sample exhibit substantially more volatile interest rates than the U.S. Average week-on-week equity returns ranged from a high of 0.35 percent 20 percent on an annualized basis in the Philippines to a low of percent negative nine percent annualized in Slovenia. While the both the average level and variance of bond flows (measured as a share of the recipient economy s GDP) were largest in Malaysia and Hungary, equity flows were large and variable in South Africa, Thailand, and Korea. The data are thus consistent with the the well-known finding that countries with more developed financial markets receive larger capital inflows (e.g. Portes and Rey, 2005). I begin by examining the time series properties of bond yields, equity returns, and capital flows in my sample, estimating an AR(1) model and conducting a Dickey-Fuller test separately for each country (see appendix Table A-3 for the results of these tests). For bond yields, the series are nonstationary in nearly all cases. Only in the case of India does the Dickey-Fuller test reject the null of a unit root. In the remainder of the paper I therefore analyze the first difference of the bond yields series, which Dickey-Fuller tests confirm to be stationary in all cases. The change in yield exhibits significant autocorrelation in half of 4 For more detail on EPFR data, see Miao and Pant (2012). 6

7 the 18 economies, with Mexico, India and the Philippines showing significant and negative AR(1) parameters. The test shows that although equity returns, bond flows, and equity flows all exhibit significant autocorrelation, these series are nonetheless stationary. Table 1: Summary Statistics: Financial Conditions in Emerging Markets Week-on-Week Yield (percent per annum) Week-on-Week Equity Return (percentage points) Mean St.Dev. Min Max Mean St.Dev. Min Max Mean St.Dev. Min Max Mean St.Dev. Min Max Turkey South Africa Brazil Chile Colombia Mexico India Indonesia Korea, Republic of Malaysia Philippines Singapore Thailand Russian Federation Czech Republic Hungary Slovenia Poland Sources: Bloomberg 2.2 Measuring Financial Uncertainty in Emerging Markets I use the volatility of bond yields and equity returns in order to measure the degree of uncertainty about financial conditions in the near future in the countries in my sample. In analyzing advanced economies, a standard measure of uncertainty about an asset s future price is the volatility implied by the prices of options that have the asset as an underlying. However, since options prices are not available for many of the bonds and equities in the sample of emerging markets, I instead turn to measures of realized volatility to provide an indicator of agents uncertainty about future financial conditions. Given the evidence in Table A-3 that that the week-on-week change in yields in several of the countries in the sample follows and AR process, any measure of uncertainty regarding future yields will make use of the residuals from an autoregressive model. LM tests of these squares of AR(1) residuals reject the null of no persistence in 16 of the 18 economies in the sample only for Hungary and Slovenia is the null of autoregressive conditional heteroskedasitity 7

8 Table 2: Summary Statistics: Capital Flows to Emerging Markets Equity Flows Bond Flows (percent of GDP) (percentage of GDP) Mean St.Dev. Min Max Mean St.Dev. Min Max Mean St.Dev. Min Max Mean St.Dev. Min Max Turkey South Africa Brazil Chile Colombia Mexico India Indonesia Korea, Republic of Malaysia Philippines Singapore Thailand Russian Federation Czech Republic Hungary Slovenia Poland Sources: EPFR rejected (tests statistics are given in Table A-5 in the appendix). Given the evidence that the change in bond yields follows an AR(1) process in some EMEs and exhibits conditional heteroskedasticity in nearly all the countries in the sample, I estimate the following ARCH(2) model separately for each country: r t = ρ 0 +ρ 1 r t 1 +ε t (1) σ 2 t = α 0 +α 1 ε 2 t 1 +α 2 ε 2 t 2. (2) Table 3 presents the resulting estimates of the ARCH coefficients (α 1 and α 2 ) for four variables I analyze. For bond yields, the ARCH(1) term is significant in all but one of the EMEs in the sample, with only the Philippines showing no evidence of conditional heteroskedasticity. The second ARCH term is also significant in 13 of the 18 economies. Both equity returns and equity flows exhibit conditional heteroskedasticy in all but one country, Indonesia. All countries debt flows have significant heteroskedasticity. Figures 2 and 3 plot realized (unconditional) and conditional volatility from the model in equations (1) and (1) for all four variables for the two of the countries in the sample (see the appendix for plots of volatility for all 18 countries). While Turkey show a notable increase in financial volatility following the announcement of renewed asset purchases by the Federal Reserve, in Chile the volatility of bond yields and equity returns actually appears to have 8

9 fallen. Thus at first glance it is far from clear from the graphs that financial conditions and capital flows became more unpredictable with the onset of quantitative easing. The graphs of volatility also show that conditional and unconditional volatility exhibit a significant amount of comovement, and the correlations between the two series range from 0.4 to 0.6. However, conditional volatility derived from the ARCH model exhibits much less persistence than the volatility measure based on the rolling average. This is unsurprising given that the unconditional volatility is calculated using a rolling 12-week window. Table 3: ARCH Coefficients Bond Yields Equity Returns Debt Flows Equity Flows α 1 α 2 α 1 α 2 α 1 α 2 α 1 α 2 South Africa 0.258*** 0.369*** 0.294*** 0.357*** 0.788*** 0.738*** 0.276*** (0.069) (0.064) (0.096) (0.098) (0.135) (0.078) (0.087) (0.04) Brazil 0.321** *** 0.169** 0.256*** 0.588*** 0.241*** 0.265*** (0.129) (0.116) (0.048) (0.083) (0.055) (0.103) (0.087) (0.054) Chile 0.257*** 0.276*** 0.505*** *** 1.01*** 0.186** 0.242*** (0.058) (0.06) (0.076) (0.067) (0.105) (0.117) (0.082) (0.052) Colombia 0.182*** *** 0.199** 0.305*** 0.889*** 0.257*** 0.562*** (0.06) (0.04) (0.065) (0.098) (0.053) (0.111) (0.09) (0.127) Mexico 0.71*** 0.164** 0.319*** 0.434*** 0.386*** 0.637*** 0.342*** 0.17** (0.09) (0.082) (0.075) (0.09) (0.089) (0.098) (0.092) (0.068) India 0.353*** 0.515*** 0.268*** 0.284*** 0.927*** 0.11** *** (0.09) (0.084) (0.077) (0.089) (0.117) (0.054) (0.11) (0.049) Indonesia 0.182** 0.239*** ** 0.552*** (0.076) (0.082) (0.104) (0.105) (0.098) (0.067) (0.095) (0.079) Korea 0.381*** 0.433*** 0.37*** 0.296*** 0.757*** 0.162** 0.188** 0.122** (0.075) (0.062) (0.075) (0.052) (0.114) (0.08) (0.078) (0.059) Malaysia 0.306*** 0.685*** 0.188** 0.344*** 0.574*** 0.208** 0.475*** 0.151*** (0.097) (0.107) (0.084) (0.08) (0.089) (0.083) (0.087) (0.051) Philippines *** 0.324*** 0.482*** 0.083* 0.586*** 0.113*** (0.077) (0.074) (0.092) (0.081) (0.089) (0.047) (0.095) (0.041) Singapore 0.507*** *** 0.152*** 0.454*** 0.081** *** (0.075) (0.059) (0.116) (0.049) (0.081) (0.039) (0.086) (0.047) Thailand 1.045*** 0.137* 0.346*** 0.187** 0.551*** 0.568*** 0.212** 0.181*** (0.131) (0.076) (0.054) (0.081) (0.106) (0.08) (0.084) (0.069) Russia 0.218* 0.214** 0.466*** *** 0.453*** 0.425*** 0.145** (0.116) (0.087) (0.104) (0.084) (0.091) (0.084) (0.064) (0.071) Czech Republic 0.272*** *** 0.287*** *** 0.359*** 1.102*** (0.073) (0.054) (0.081) (0.078) (0.036) (0.078) (0.108) (0.102) Hungary 0.346*** 0.278*** 0.218*** 0.339*** 0.406*** 0.438*** 0.375*** 0.692*** (0.067) (0.057) (0.067) (0.077) (0.065) (0.063) (0.103) (0.082) Slovenia 0.18*** 0.315*** 0.253*** 0.328*** 3.563*** 0.704** 0.346*** 0.884*** (0.052) (0.045) (0.094) (0.102) (0.678) (0.305) (0.078) (0.177) Poland 0.16** 0.244*** 0.263*** 0.259*** 0.49*** 0.343*** 0.188** 0.981*** (0.068) (0.051) (0.071) (0.092) (0.09) (0.051) (0.089) (0.109) Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Sources: Global Financial Data, Bloomberg 9

10 Figure 2: Financial Volatility in Turkey Basis Points Bond Yields Bond Flows Equity Returns Equity Flows Unconditional Volatility (Left Scale) Conditional volatility (Right Scale) Notes: Red line marks Sep. 13, 2012 announcement of QE3 large scale asset purchase program. Unconditional volatility is the rolling 12 week st.dev of AR(1) residuals. Conditional volatility is the square root of residuals from an ARCH(2) model 10

11 Figure 3: Financial Volatility in Chile Basis Points Bond Yields Bond Flows Equity Returns Equity Flows Unconditional Volatility (Left Scale) Conditional volatility (Right Scale) Notes: Red line marks Sep. 13, 2012 announcement of QE3 large scale asset purchase program. Unconditional volatility is the rolling 12 week st.dev of AR(1) residuals. Conditional volatility is the square root of residuals from an ARCH(2) model 3 Impacts of Unconventional Monetary Policy 3.1 Unconditional Means I now examine whether unconventional monetary policy impacted the volatility of financial conditions and capital flows in EMEs by focusing on the two-year period centered on the Fed s announcement of the asset purchase program known as QE3 on September 13, Table 4 compares the average value of conditional volatility in each economy in the year before the announcement of QE3 and the year after. All economies in the sample exhibited significant conditional volatility in both periods. Based on these unconditional means of volatility, the evidence on whether QE3 boosted financial market volatility appears mixed. Bond yield volatility increased in 12 of the 18 EMEs, significantly so in seven cases. Economies often mentioned as vulnerable during the 11

12 second half of 2013 all saw an increase in interest rate volatility. At the same time, in Chile and Mexico, portrayed as less vulnerable by the financial press, as well as in several Eastern European economies, volatility actually fell after the announcement of renewed asset purchases. Forequity thepicture isreversed: volatilityfell in13ofthecountries, in8ofthem significantly. For both types of capital flows, volatility was greater during the first 12 months of QE3 in the majority of cases, but the increase was only significant in around a third of the cases. Thus, comparing average conditional volatility before and after the onset of QE3 presents very limited evidence that quantitative easing boosted financial market volatility across the board. 3.2 Methodology In this section I analyze why financial volatility increased in some emerging markets after the announcement of QE3 while falling elsewhere. To do this, I follow Bekaert and Harvey (1997) in modelling the impact of unconventional monetary policy on financial volatility as varying with several features of the international environment and the domestic economy: ln (ˆσ 2 i,t ) = β1 UMP t +β 2 UMP t ( X US t +X i,t 1 ) +γ 1 X US t +γ 2 X i,t 1 +θ i +e i,t (3) Where ˆσ i,t 2 is the fitted value of volatility in country i in week t from the model of equations (1) and (2). I take the natural logarithm the volatility so that coefficients can be interpreted asmeasuring thepercentage change inconditional volatility. The variableump t is adummy variable set equal to one in the 52 weeks after September 13, 2012 when the Fed announced the program of renewed asset purchases referred to as QE3. Variables measuring financial and real economic conditions in the U.S. are gathered in the vector Xt US while X i,t is a vector of variables characterizing domestic conditions in the individual EMEs. The domestic explanatory variables are lagged one period in order to limit concerns about endogeneity. 5 Finally θ i represents a country-specific intercept. This specification tests whether reintroduction of unconventional monetary policy affected the average volatility of bond yields in emerging markets, conditional on other factors affecting volatility, as well as whether the impact of those other factors on volatility shifted with the announcement of QE3. I measure financial conditions in the U.S. using the weekly change in the yield on 10-year Treasury bonds as well as the weekly average level of the VIX index, which measures the implied volatility of options on the S&P500 index and is frequently used as an indicator of risk appetite in the U.S (e.g. Adrian and Shin, 2010; Fratzscher, 2011). As a measure of real economic performance in the U.S. I use the month-on-month change in industrial production. The goal of this exercise is to determine whether changes in the volatility of bond yields, equity returns, and capital flows in EMEs after September 2012 can be explained by factors other than the introduction of unconventional monetary policy. I therefore include 5 All variables were lagged according the their own frequency, so that lagged inflation during the weeks of February 2012 refers to inflation in in January Similarly, the lagged value of GDP growth in May is annualized percent change in real seasonally adjusted GDP in the first quarter. 12

13 Table 4: Difference in Conditional Volatility: Before QE3 vs During QE3 Bond Yields Equity Returns Bond Flows Equity Flows Turkey ** 0.1*** (11.7) (0.5) (0.1) (0.0) South Africa (6.6) (0.2) (0.2) (0.2) Brazil * 0.1 (4.8) (0.5) (0.1) (0.1) Chile -5.7** (2.3) (0.2) (0.1) Colombia 12.6*** -0.6** *** (4.8) (0.3) (0.2) (0.0) Mexico *** (14.1) (0.2) (0.1) (0.1) India 20.6* -0.5* (12.4) (0.3) Indonesia (7.1) (0.4) (0.1) Korea *** *** (1.9) (0.3) (0.1) Malaysia 1.7* (0.9) (0.2) (0.1) Philippines 9.6** *** (3.8) (0.3) (0.2) (0.1) Singapore 6.6** -0.6*** (3.1) (0.2) (0.1) (0.1) Thailand 2.7** ** 0.4*** (1.3) (0.4) (0.1) (0.1) Russia ** 0.2** 0.0 (7.5) (0.4) (0.1) Czech Republic -9.8** -0.9*** 0.1*** 0.0 (4) (0.3) (0.1) Hungary -26.6*** -1.2*** (10.3) (0.4) (0.3) (0.0) Slovenia ** (10.1) (0.3) (0.1) Poland 11.9*** -0.7** ** (4.1) (0.3) (0.1) 1 Fitted values for conditional volatility from an ARCH(2) model, averaged over the 52 weeks before Sep.13, 2012 and the 52 weeks after. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Sources: Bloomberg, EPFR as regressors several variables measuring domestic economic performance in the individual emerging markets, including the quarter-on-quarter change in real seasonally adjusted GDP, the month-on-month change in the seasonally adjusted consumer price index, the seasonally adjusted quarterly government budget balance as a percent of GDP, and the current account balance as a share of GDP. 6 6 See the appendix for details of the sources for these and other variables. Seasonal adjustment was done 13

14 I also include as regressors several variables which, although time varying, capture structural features of the economy rather than its cyclical performance. As is standard I use equity market capitalization as a share of GDP as a proxy for the level of financial development. Trade (the sum of imports and exports) as a share of GDP captures openness to trade, while financial openness is measured with the Chinn-Ito index. Finally, I include a dummy set to one in countries with a floating exchange rate, a variable I obtain from Ilzetzki et al. (2010) and which does not vary over time during the period. In order to ascertain whether the exchange rate regime does affect interest rate volatility, I therefore I estimate the model using a random effects specification that assumes θ i is uncorrelated with the other regressors, as well as a fixed effects specification and present results for both. 3.3 Regression Results I begin by imposing that β 2 in equation (3) be equal to zero, to first test whether the relationship between UMP and conditional volatility in Table 4 remains once I control for other factors that the literature has found to affect financial conditions broadly speaking in EMEs. Table 8 therefore presents the OLS estimates of equation (3) when the interaction terms are excluded. My preferred specification, in columns (1) and (2), takes as the dependent variable the estimates of conditional volatility obtained in the previous section; however, I also present results from using unconditional volatility as the dependent variable. I do this in order to make clear the extent to which my results may vary depending on the particular volatility measure used. Even controlling for numerous other covariates, the period in which the Fed was undertaking unconventional monetary policy remains significantly associated with higher EME bond yield conditional volatility. For equity returns, conditional volatility is not significantly different across the two periods. Although the unconditional volatility is significantly smaller under QE3, the coefficient is relatively small. Quantitative easing is associated with higher volatility for both bond and equity flows. For all four variables and across all specifications the coefficient on the VIX is always significant and positive, implying that periods of low risk appetite in the U.S. are associated with elevated financial volatility in emerging markets. Changes in yields on Treasury bonds are associated with significantly higher equity yield volatility as well as more volatile capital flows. Output growth in the domestic economy does appear related to bond market or capital flow volatility, but is positively associated with equity return variance. Unsurprisingly, domestic inflation is significantly related to the volatility of bond yield and equity returns. On the other hand, domestic growth and inflation are not significantly related to capital flow volatility. The current account balance is negatively related to the volatility equity returns, which is logical given that a larger current account deficit means a country relies more on external financing. using the TRAMO-SEATS algorithm implemented in the Demetra+ software available from Eurostat. 14

15 Table 5: Regression Results: UMP and Interest Rate Volatility Volatility Measure: Conditional Unconditional (1) (2) (3) (4) UMP Dummy 0.161*** 0.090** 0.129*** (0.033) (0.036) (0.028) (0.029) VIX 0.018*** 0.021*** 0.025*** 0.030*** (0.003) (0.003) (0.002) (0.002) US 10-year Yield *** 0.359*** (0.121) (0.118) (0.101) (0.097) US Industrial Production *** *** (0.009) (0.009) (0.007) (0.007) Growth ** (0.002) (0.002) (0.002) (0.002) Inflation 0.010*** 0.006* 0.006** 0.009*** (0.003) (0.003) (0.003) (0.003) Current Account (%GDP) *** *** * ** (0.008) (0.009) (0.008) (0.008) Fiscal Balance (%GDP) *** ** *** (0.004) (0.005) (0.004) (0.004) Equity Market Cap (%GDP) *** *** (0.001) (0.006) (0.001) (0.005) Trade (% GDP) *** *** 0.073*** (0.000) (0.009) (0.001) (0.007) Financial Openness 0.084*** *** (0.024) (0.113) (0.047) (0.094) Floating ER Dummy ** ** (0.082) (0.179) Country Fixed Effects? No Yes No Yes Observations Standard errors in parentheses; ** p<0.01, ** p<0.05, * p<0.1. All domestic regressors lagged one period. Unconditional volatility is the 12-week rolling standard deviation of the change in bond yields. Conditional volatility is the vector of fitted values produced from estimating the model in Section 2. See the appendix for data sources. 15

16 Table 6: Regression Results: UMP and Equity Return Volatility Volatility Measure: Conditional Unconditional (1) (2) (3) (4) UMP Dummy ** *** (0.013) (0.014) (0.018) (0.02) VIX 0.011*** 0.012*** 0.033*** 0.035*** (0.001) (0.001) (0.001) (0.002) US 10-year Yield 0.001** 0.001** 0.003*** 0.003*** (0.000) (0.000) (0.001) (0.001) US Industrial Production (0.003) (0.003) (0.005) (0.005) Growth 0.002*** 0.002** 0.006*** 0.005*** (0.001) (0.001) (0.001) (0.001) Inflation 0.003** 0.003** 0.014*** 0.016*** (0.001) (0.001) (0.002) (0.002) Current Account (%GDP) 0.007** 0.007* 0.032*** 0.030*** (0.003) (0.004) (0.005) (0.005) Fiscal Balance (%GDP) ** *** *** (0.002) (0.002) (0.003) (0.003) Equity Market Cap (%GDP) ** 0.005** ** (0.000) (0.002) (0.001) (0.003) Trade (% GDP) ** *** (0.000) (0.004) (0.001) (0.005) Financial Openness ** (0.011) (0.046) (0.045) (0.083) Floating ER Dummy (0.039) (0.176) Country Fixed Effects? No Yes No Yes Observations Standard errors in parentheses; ** p<0.01, ** p<0.05, * p<0.1. All domestic regressors lagged one period. Unconditional volatility is the 12-week rolling standard deviation of the change in bond yields. Conditional volatility is the vector of fitted values produced from estimating the model in Section 2. See the appendix for data sources. Equity market capitalization is associated with higher bond and equity market volatility in all the fixed effects specifications. This indicates that it is countries undergoing particularly rapid financial development in which volatility is elevated, rather than variation in financial development across countries driving the result. The coefficients on the openness measures are significant and positive for bond return volatility, but not for the other three dependent variables. Finally, consistent with theory, countries with floating exchange regimes have on average lower domestic bond yield volatility but higher capital flow volatility. I now estimate equation (3) with the full compliment of interaction terms. Results are presented in Table 9, which for the sake of space focuses on the results for conditional volatility and employs a country fixed effects specification only. The relationship between the UMP dummy and conditional volatility for bond yields and flows is now negative. This is evidence that the higher average EME bond market volatility during QE3 was the result of an amplification of the effects of other determinants of volatility, rather than QE itself. In 16

17 Table 7: Regression Results: UMP and Bond Flow Volatility Volatility Measure: Conditional Unconditional (1) (2) (3) (4) UMP Dummy 0.616*** 0.573*** 0.281*** 0.164*** (0.067) (0.073) (0.032) (0.034) VIX 0.061*** 0.061*** 0.041*** 0.045*** (0.005) (0.006) (0.003) (0.003) US 10-year Yield 0.010*** 0.010*** 0.006*** 0.005*** (0.002) (0.002) (0.001) (0.001) US Industrial Production *** *** 0.021*** 0.024*** (0.017) (0.017) (0.008) (0.008) Growth * ** (0.004) (0.005) (0.002) (0.002) Inflation *** 0.016*** (0.007) (0.007) (0.004) (0.003) Current Account (%GDP) *** *** *** *** (0.018) (0.019) (0.009) (0.009) Fiscal Balance (%GDP) (0.01) (0.011) (0.005) (0.005) Equity Market Cap (%GDP) * *** (0.003) (0.012) (0.002) (0.006) Trade (% GDP) *** 0.093*** (0.002) (0.019) (0.002) (0.011) Financial Openness * (0.099) (0.234) (0.064) (0.105) Floating ER Dummy 1.606*** 1.628*** (0.361) (0.256) Country Fixed Effects? No Yes No Yes Standard errors in parentheses; ** p<0.01, ** p<0.05, * p<0.1. All domestic regressors lagged one period. Unconditional volatility is the 12-week rolling standard deviation of the change in bond yields. Conditional volatility is the vector of fitted values produced from estimating the model in Section 2. See the appendix for data sources. 17

18 Table 8: Regression Results: UMP and Equity Flow Volatility Volatility Measure: Conditional Unconditional (1) (2) (3) (4) UMP Dummy 0.252*** 0.236*** 0.252*** 0.194*** (0.038) (0.042) (0.023) (0.025) VIX 0.006* 0.006* 0.015*** 0.017*** (0.003) (0.003) (0.002) (0.002) US 10-year Yield 0.005*** 0.005*** 0.002** 0.002** (0.001) (0.001) (0.001) (0.001) US Industrial Production *** 0.023*** (0.01) (0.01) (0.006) (0.006) Growth (0.003) (0.003) (0.002) (0.002) Inflation ** (0.004) (0.004) (0.003) (0.003) Current Account (%GDP) *** *** (0.011) (0.011) (0.006) (0.006) Fiscal Balance (%GDP) 0.012** 0.011* 0.009** (0.006) (0.006) (0.004) (0.004) Equity Market Cap (%GDP) 0.007*** *** 0.024*** (0.002) (0.007) (0.002) (0.004) Trade (% GDP) *** (0.001) (0.011) (0.002) (0.006) Financial Openness *** (0.067) (0.133) (0.063) (0.077) Floating ER Dummy ** (0.256) (0.356) Country Fixed Effects? No Yes No Yes Observations Standard errors in parentheses; ** p<0.01, ** p<0.05, * p<0.1. All domestic regressors lagged one period. Unconditional volatility is the 12-week rolling standard deviation of the change in bond yields. Conditional volatility is the vector of fitted values produced from estimating the model in Section 2. See the appendix for data sources. 18

19 particular, the estimates suggest that the marginal effect of an increase in the VIX on bond yields and flows increased dramatically when the Fed resumed its asset purchases. For the equity variables, the reverse is true. The positive impact of the VIX on equity return and flow volatility is muted in the QE3 period. The U.S. 10-year yield also had less of an effect on volatility once unconventional monetary policy resumed, while better U.S. real economic performance (as measured by industrial production) was associated with lower volatility during the QE3 period, but not before. Domestic GDP growth in EMEs was associated with slightly lower volatility in the pre-qe3 period, but on net amplified volatility once asset purchases resumed. The interaction between other domestic variables and the QE3 dummy are not significant (and thus not shown in Table 9 for the sake of space), and the estimated coefficients on the variables themselves are generally unchanged from those in Table 8. Thus the relationship between domestic factors other than growth from growth and inflation does not appear to have changed with the onset of QE3 in September

20 Table 9: Regression Results: UMP and Changing Drivers of Financial Volatility Conditional Volatility of: Bond Equity Bond Equity Yields Returns Flows Flows (1) (2) (3) (4) UMP Dummy ** 0.228** *** (0.223) (0.101) (0.504) (0.289) VIX 0.017*** 0.012*** 0.055*** 0.007** (0.003) (0.001) (0.006) (0.003) US 10-year Yield *** 0.013*** 0.008*** (0.001) (0.001) (0.003) (0.002) US Industrial Production (0.01) (0.004) (0.022) (0.013) Growth * 0.002** ** (0.002) (0.001) (0.005) (0.003) Inflation (0.004) (0.002) (0.01) (0.006) Fiscal Balance (%GDP) ** *** * (0.01) (0.004) (0.022) (0.012) Current Account (%GDP) *** * ** (0.006) (0.003) (0.013) (0.007) Equity Market Cap (%GDP) 0.019*** (0.006) (0.003) (0.014) (0.008) Trade (% GDP) 0.041*** 0.011** 0.055** 0.034** (0.011) (0.005) (0.026) (0.015) Financial Openness ** (0.103) (0.047) (0.234) (0.134) UMP * VIX 0.026*** ** 0.113*** 0.01 (0.01) (0.004) (0.022) (0.012) UMP * US 10yr * *** (0.002) (0.001) (0.005) (0.003) UMP*US Ind.Production ** *** *** (0.015) (0.007) (0.034) (0.02) UMP*Growth 0.021*** *** 0.021*** (0.005) (0.002) (0.012) (0.007) UMP * Financial Openness ** *** (0.026) (0.012) (0.059) (0.034) Country Fixed Effects? Yes Yes Yes Yes Observations Standard errors in parentheses; ** p<0.01, ** p<0.05, * p<0.1. All regressors lagged one period. Unconditional volatility is the 12-week rolling standard deviation of the change in bond yields. Conditional volatility is the vector of fitted values produced from estimating the model in Section 2. See the appendix for data sources. 4 Conclusions This paper has examined how unconventional monetary policy pursued by the Federal Reserve has affected financial market volatility in a set of 18 emerging markets since I 20

21 estimated the conditional volatility of bond yields using an ARCH(2) model and showed that on average volatility was higher in the first year of QE3 than in the year prior to the beginning of the asset purchase program. In panel data regressions, I found that the higher volatility after September 2013 results from greater sensitivity to changes in U.S. real and financial market conditions. While the impact of domestic inflation on bond yield volatility does increase, all other domestic variables either have an unchanged or a lower impact during the QE3 period. These findings suggest that financial market uncertainty constitutes a channel through which unconventional monetary policy in advanced economies spills over to EMEs. They also suggest that the global factors that much recent research has shown to affect the level of real and financial variables in emerging markets also have important effect on the volatility of bond yields and thus on uncertainty regarding future financial conditions. References Adrian, T. and Shin, H. S. (2010). Liquidity and leverage. Journal of Financial Intermediation, 19(3): Risk Transfer Mechanisms and Financial Stability. Akinci, O. (2013). Global financial conditions, country spreads and macroeconomic fluctuations in emerging countries. Working Paper 1085, Board of Governors of thefederal Reserve System. Bekaert, G. and Harvey, C. R. (1997). Emerging equity market volatility. Journal of Financial Economics, 43(1): Benigno, G., Converse, N., and Fornaro, L. (2015). Large capital inflows, sectoral allocation, and economic performance. Journal of International Money and Finance, 55: Macroeconomic and financial challenges facing Latin America and the Caribbean after the crisis. Benigno, G. and Fornaro, L. (2014). The financial resource curse. The Scandinavian Journal of Economics, 116(1): Bloom, N. (2014). Fluctuations in uncertainty. Journal of Economic Perspectives, 28(2): Bowman, D., Londono, J. M., and Sapriza, H. (2014). U.s. unconventional monetary policy and transmission to emerging market economies. Working paper, Board of Governors of the Federal Reserve. Burger, J. D. and Warnock, F. E. (2007). Foreign participation in local currency bond markets. Review of Financial Economics, 16(3): ce:title Exchange Rates and International Financial Assets: A Special Issue in Honor of Stanley W. Black /ce:title. Converse, N. (2014). Uncertainty, capital flows, and maturity mismatch. mimeo. 21

22 Edwards, S. (2012). The federal reserve, the emerging markets, and capital controls: A highfrequency empirical investigation. Journal of Money, Credit and Banking, 44: Edwards, S. and Susmel, R. (2003). Interest-rate volatility in emerging markets. The Review of Economics and Statistics, 85(2):pp Fernández-Villaverde, J., Guerrón-Quintana, P. A., Rubio-Ramírez, J., and Uribe, M. (2011). Risk matters: The real effects of volatility shocks. American Economic Review, 101(6): Forbes, K. J. and Warnock, F. E. (2012). Capital flow waves: Surges, stops, flight, and retrenchment. Journal of International Economics, 88(2): Fratzscher, M. (2011). Capital flows, push versus pull factors and the global financial crisis. Working Paper 17357, National Bureau of Economic Research. Fratzscher, M., Lo Duca, M., and Straub, R. (2013). On the international spillovers of US quantitative easing. Working Paper Series 1557, European Central Bank. González-Rozada, M. and Levy-Yeyati, E. (2008). Global factors and emerging market spreads. The Economic Journal, 118(533): Hayo, B. and Kutan, A. M. (2005). Imf-related news and emerging financial markets. Journal of International Money and Finance, 24(7): Ilzetzki, E., Reinhart, C. M., and Rogoff, K. S. (2010). Exchange rate arrangements entering the 21st century: Which anchor will hold? mimeo. International Monetary Fund (2013). Global impact and challenges of unconventional monetary policies. Imf policy paper, International Monetary Fund. Miao, Y. and Pant, M. (2012). Coincident indicators of capital flows. Working Paper 12/55, International Monetary Fund. Portes, R. and Rey, H. (2005). The determinants of cross-border equity flows. Journal of International Economics, 65(2): Rajan, R. (2013). A step in the dark: Unconventional monetary policy after the crisis. Andrew Crockett Memorial Lecture. BIS. Reis, R. (2013). The portuguese slump and crash and the euro-crisis. Brookings Papers on Economic Activity. Rey, H. (2013). Dilemma not trilemma: The global financial cycle and monetary policy independence. Paper presented at the 25th Jackson Hole Symposium. Federal Reserve Bank of Kansas City. 22