WPS7363 Policy Research Working Paper 7363 Global Liquidity and External Bond Issuance in Emerging Markets and Developing Economies Erik Feyen Swati Ghosh Katie Kibuuka Subika Farazi Finance and Markets Global Practice Group July 2015 Policy Research Working Paper 7363 Abstract Using the universe of all externally issued bonds by corpo- industry, and riskiness). Additionally, we document support rates and sovereigns in emerging and developing economies that the risk-taking channel of exchange rate appreciation during 2000–14, this paper analyzes various issuance trends, also operates for external bond issuance. Moreover, while including the unprecedented post-crisis surge. The paper the paper finds that country pull factors affect the impact focuses on external issuance at the country-industry and of global factors, it does not find consistent evidence for individual bond levels and finds that global factors matter this across the board. This result suggests that, during loose greatly for emerging and developing economies issuance. A global funding conditions, flows are mostly driven by push decrease in U.S. expected equity market (or interest rate) factors and do not systematically discriminate between volatility, U.S. corporate credit spreads, and U.S. interbank emerging and developing economies. Taken together, the funding costs and an increase in the Federal Reserve’s bal- findings suggest that although issuers might be able to ance sheet (i) raise the odds that the monthly issuance benefit from benign international funding conditions, the volume of a country-industry is above its historical aver- large issuance volumes, currency risks, and high exposure to age; (ii) decrease individual bond yields and spreads; and global factors could pose external and domestic challenges (iii) raise bond maturities, after controlling for country pull for policy makers, particularly when global cycles reverse. factors, bond characteristics (for example, type of issuer, This paper is a product of the Finance and Markets Global Practice Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The corresponding author may be contacted at efeijen@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Global Liquidity and External Bond Issuance in Emerging Markets and Developing Economies Erik Feyen* Swati Ghosh Katie Kibuuka Subika Farazi World Bank Washington, D.C. Key words: Global Liquidity, Bonds, Primary Market, Capital Flows, International Finance. JEL Classification: F21, F32, F36, G12, G15. * Corresponding author: Erik Feyen is a Lead Financial Sector Economist in the Finance and Markets Global Practice (efeijen@worldbank.org). Swati Ghosh is an Adviser in the Macro and Fiscal Management Global Practice. Katie Kibuuka and Subika Farazi are both Financial Sector Specialists in the Finance and Markets Global Practice. We thank Pierre-Laurent Chatain, Stijn Claessens, Mario Guadamillas, Roberto Rocha, Hyun Shin, and participants of the Analytical Group on Vulnerabilities of the Financial Stability Board for comments and Diego Sourrouille for excellent research assistance. I. Introduction The global financial crisis has given new impetus to the debate on the global financial liquidity cycle, which is primarily brought forth by monetary policy, risk appetite, and leverage in “financial center countries” and transmitted through loose funding conditions to the rest of the world (e.g. Rey (2013)). Emerging and developing economies (EMDEs) benefited from the pre-crisis upturn in this global cycle mainly through internationally active banks (e.g. Bruno and Shin (2015a, 2015b)). However, the process had also led to a build-up in global imbalances and financial fragilities which came to the fore when global banks deleveraged to strengthen their balance sheets and to comply with the new global regulatory architecture. In the wake of the crisis, various developed economies embarked on unprecedented, extraordinary monetary policies (EMPs) to rekindle domestic economic growth and battle disinflationary pressures mainly via (promises of future) ultra-low policy rates and large-scale asset purchasing programs (LSAP) that aimed to bring down long-term interest rates. Since 2009-10, EMPs in the United States in particular have produced a prolonged episode of ultra-low global interest rates as well as extremely low volatility in financial markets. This in turn has contributed to a revival of ample global funding conditions and widespread financial risk taking as developed market investors searched for yield to meet targeted returns. The spillover effects of these EMPs on EMDEs have been profound—EMDEs have experienced an unparalleled surge in total gross capital inflows from an annual average of $0.5 trillion during 2000-2007 to $1.1 trillion during 2010-2013. As a result, portfolio investors in developed markets currently allocate over $4 trillion or 13 percent of their investments to EMDEs. Moreover, bonds funds allocations from developed markets to EMDEs have grown by 375% to $385 billion since 2009 (Figure 1), equities funds allocations have expanded by 70% to $985 billion (Figure 2), and foreign participation in some local bond markets has increased up to 26 percent of volume outstanding (Figure 3). These massive capital inflows can set a in motion a feedback loop in EMDEs that consists of: i) ample domestic liquidity and loosening lending conditions, ii) increasing leverage, iii) rising asset prices and stronger domestic balance sheets through local currency appreciation, and iv) an improving growth and fiscal outlook. And as long as the cycle is virtuous, it attracts even more inflows which reinforces the cycle. Yet, while producing short-term growth, boosting investor optimism, and potentially extending debt maturities, these flows also provide challenges for EMDE policy makers, as they can put pressure on currencies and foreign reserves management, interfere with the local credit cycle and monetary policy, produce shadow banking risks, distort asset prices, and reduce incentives for structural reform. In this context, the inevitable exit from EMPs and the normalization of global interest rates could prove disruptive for EMDEs (e.g. Rajan (2013), Turner (2014), IOSCO (2014)). The “Taper Tantrum”1 episode is instructive in this respect and shows that market expectations regarding 1 In May 2013, the Federal Reserve hinted it might start scaling down its LSAP triggering virulent bouts of volatility in EMDE currencies, equities, and capital inflows. 2 EMPs matter greatly (e.g. Eichengreen and Gupta (2014)). Now, more than 6.5 years later since their launch, EMPs appear to have contributed to the nascent economic recovery in the United States, and the Federal Reserve has finally discontinued its LSAP series of mortgage-backed securities and Treasury bonds purchases and is preparing to raise the policy rate for the first time in a decade. In contrast, the European Central Bank has recently launched its own LSAP in addition to other EMPs. These measures have helped drive down yield curves in Europe to record lows, suggesting the impacts of EMPs on global financial markets and its contribution to global liquidity will endure. This paper focuses on the impact of global liquidity factors on a subset of capital inflows to EMDEs which has grown dramatically: the external issuance of bonds by corporates and sovereigns. Bond markets have become a major transmission channel of global liquidity (e.g. Shin (2013), Avdjiev et al (2014)). During 2009-14, corporates and sovereigns in EMDEs cumulatively issued $1.5 trillion in external bonds—overwhelmingly in foreign currencies—representing almost a tripling from $520 billion in 2002-07. This surge is not driven by a single region or country, but reflects a broad-based trend since the cumulative issuance to GDP ratio is 6.7% for the median EMDE, up from 4.3% in the pre-crisis period. For example, various countries issued externally for the first time during the period, including Angola, Armenia, Botswana, Ghana, Kenya, Laos, Namibia, and Mozambique. Against this backdrop, this paper seeks to answer five main research questions:  Question 1: What are the main trends in external issuance by EMDE entities during the 2000-14 period (e.g. volumes, stocks, currencies, maturities, yields)?  Question 2: What is the impact of global factors—proxied by financial conditions in the United States—on the propensity to issue external bonds by an EMDE country-industry compared to its historical issuance average?  Question 3: What is the impact of these global factors on two important bond characteristics at the time of issuance: its yield (and spread) and maturity?  Question 4: Do country characteristics interact and amplify or dampen the impact of global factors?  Question 5: Does the risk-taking channel through exchange rate appreciation as described and tested in Bruno and Shin (2015a, 2015b) also operate in a similar fashion for external bond issuance by increasing the propensity for country-industries to issue externally? In our setting, this channel hypothesizes that local currency appreciation strengthens local borrowers’ balance sheets and their external bond issuance capacity which triggers higher cross-border flows by international investors who are willing to take on more risk. Our paper makes four contributions. First, we compile external bond issuance data sets, which cover the universe of external bond issuances by EMDEs during 2000-14. We use these data sets to document recent trends in bond flows, stocks, pricing, and maturities across EMDEs. Second, to our knowledge, this paper is the first to study the impact of global factors on primary activity of EMDE entities in international bond markets since the start of new millennium. Third, we find support for the risk-taking channel of exchange rate appreciation for external bond issuance. Fourth, we undertake the analysis on the country-industry or bond tranche level which allows us to account for industry-specific and deal-specific factors (e.g. currency, bond riskiness, bond size). 3 This ameliorates bias due to compositional and selection effects which are present in aggregated capital flows data which are typically the focus of inquiry in the literature. The remainder of this paper is structured as follows. Section II provides a brief overview of the literature. Section III documents external issuance trends to answer Question 1. Section IV discusses the data. Section V lays out the methodology to address Questions 2, 3, and 4 and section VI discusses the empirical results. Section VII concludes. II. Literature Overview Global liquidity The pronounced simultaneous resurgence in capital flows to EMDEs seen since 2009-10 is not a new phenomenon. An extensive literature dating to the 1990s (e.g. Calvo et al (1993)) has emphasized the importance of global push factors, notably real interest rates and growth in advanced economies. Indeed, capital flows to EMDEs have long tended to exhibit strong co- movements suggesting that common drivers in the global environment are at play—both across types of flows (with the exception of FDI flows) and across geographical regions. This observation is corroborated by Koepke (2015), who, while summarizing relevant empirical literature, concludes that global push factors matter relatively more than country pull factors for portfolio flows. He finds that country pull factors matter more for banking flows. Bekaert, Hoerova and Lo Duca (2013) show that a lower Federal Funds rate triggers a subsequent reduction in measures of uncertainty and risk aversion, proxied, for example, by the VIX index, which measures the 30-day ahead expected volatility derived from S&P 500 index options (Figure 16). And Forbes and Warnock (2012) show that a lower VIX is associated with a surge in capital flows. Rey (2013) finds that capital inflows are negatively correlated with the VIX even at a geographically disaggregated level, and that this pattern holds even when conditioned by other global factors such as the real interest rate and world growth rate. Bruno and Shin (2015a, 2015b) highlight banks as a channel of transmission. Bruno and Shin (2015a) provide and empirically test a model of risk-taking through currency appreciation. They show that the leverage cycle of international banks is associated with higher cross-border bank flows. This triggers currency appreciation in the recipient countries which strengthens local balance sheets allowing banks to lend more. In a VAR framework, Bruno and Shin (2015b) find that a lower VIX entices globally active banks to take on additional leverage, arguably because they target a certain value-at-risk (VaR) measure which mechanically allows for higher leverage when uncertainty measures fall (e.g. Adrian and Shin (2010)). They also find that the U.S. Dollar depreciates as VIX decreases, which results in a loosening of Dollar lending conditions in international funding markets. Rey (2013) and Bruno and Shin (2015b) also provide evidence that higher leverage leads to a subsequent fall in risk aversion measures, giving rise to a positive feedback loop. The mechanism is as follows: when balance sheets expand in response to lower uncertainty (VIX) through increased collateralized lending and borrowing by financial intermediaries, the newly released funding resources chase available assets for purchase. If this leads to a generalized increases in asset prices in the financial system, it then affects future risk appetite (leads to a fall in risk aversion). 4 However, recently, bond markets have taken over as a transmission channel. For example, Shin (2013) documents the impact of the VIX on portfolio bond flows. He argues that since 2010, “reaching for yield” by investors in developed economies has contributed to the decline in risk premiums for debt securities and a surge in issuance of international debt securities. In particular, Shin (2013) discusses the increase in offshore issuance of international debt securities by non- financial firms that operate across borders. Co-movement of capital flows also translates to co-movement of asset prices. For example, Miranda-Agrippino and Rey (2012) look at equity markets and show that about 25% of the variance of a large cross section of prices of risky assets is explained by a single global factor (the VIX). That is, they find that irrespective of the geographical location of the market in which the assets are traded or the specific asset class they belong to, risky returns load to a large extent on this global factor. Of course, domestic factors still matter for capital inflows as well. For instance, Ghosh et al (2010) look at aggregate capital inflows and find that, conditional on a surge of capital inflows occurring (which is determined by global factors), whether or not a particular country receives any flows depends on its domestic macroeconomic and institutional factors. Similarly, Fratzscher (2011) documents that common shocks exert a large effect on portfolio flows, but also finds the effects are highly heterogeneous across countries, with a large part of this heterogeneity explained by differences in the quality of institutions, country risk and the strength of domestic fundamentals. Cerutti et al. (2014) who analyze cross-border banking flows in particular, find that, while U.S. financial conditions (VIX and term premia) are important, recipient country characteristics affect both the level of country specific flows as well as the cyclical impact of global liquidity on the domestic economies. EMPs EMPs mainly operate through various channels to affect investor portfolio decisions and contribute to the global liquidity cycle, with the attendant domestic and international consequences, including:  The portfolio balance channel: To the extent that assets are not perfectly substitutable, the central bank’s purchase of a security such as a U.S. Treasury, affects the available supply of this asset to investors and reduces its yield, pushing investors into holding other assets.  The expectations channel: If the markets interpret the central bank’s announcements or operations as signaling lower future policy rates than they had previously expected, bond yields may decline via a lower risk neutral component of interest rates.  The confidence channel: The central bank’s actions may also provide new information about the current state of the economy –which in turn could affect the portfolio decisions and asset prices by changing investors’ risk appetites.   The liquidity channel: Assets purchased through LSAP operations boost the reserves of commercial banks held at the central bank which can more easily be traded on secondary markets than can long term securities. As a result, the liquidity premium declines, which helps unclog funding markets, lower borrowing costs, and boost bank lending (Joyce et al. 2012).  5 Recent research has also looked specifically into the effects of EMPs in the United States on both capital flows and asset prices in EMDEs. Fratzscher, Lo Duca and Straub (2013) find that the first LSAP or quantitative easing (QE1) in the United States (which focused on providing liquidity to financial institutions to repair markets) triggered a reversal of flows back to the United States as investor anxiety over U.S. conditions subsided. In contrast, subsequent LSAPs (QE2 and QE3, which focused on asset purchases) had the opposite effect and induced a portfolio rebalancing out of U.S. equities and bonds and partly into EMDEs. These effects occurred both at the time of announcement of the program as well as during actual asset purchases. Burns et al (2014) find that 13% of the total variation in capital flows from developed economies to EMDEs can be specifically attributed to a QE effect in the Unites States. Jointly, financial conditions in the United States and domestic pull factors in EMDEs account for 60% and 40% of the variation, respectively. Expectations regarding EMPs in the Unites States particularly matter for flows to EMDEs. Koepke (2014) finds that a one percentage point increase in market expectations for the Federal Funds rate three years forward was associated with a short-term decrease of $6-7 billion on bond flows to EMDEs and $1.2-6.5 billion on equity flows. The cumulative, long-term effect might be twice as large. The effect also appears to be asymmetric as a shift towards expectations of monetary tightening is much larger than a shift towards expectations of easing. III. The Evolution of EMDE Activity in International Primary Bond Markets This section addresses the first research question. Since 2000, external issuance of corporate and sovereign entities in EMDEs has shown various trends. We discuss i) issuance volume, ii) outstanding stocks, iii) currencies, iv) issuing industries and use of proceeds, v) maturities and yields at issuance, vi) maturing profile, and vii) quality of issuance. Panel A in Appendix 2 provides summary statistics on bond issuance by year. Issuance volume trends 1. External bond issuance increased steadily before the global financial crisis and accelerated rapidly after the crisis reaching unprecedented levels (Figure 4). Total annual issuance rose from around $64 billion in 2000 to $400 billion in 2014. For the pre- crisis years (2000-07), annual average issuance was about $80 billion and grew at an average annual rate of 6%. The global financial crisis negatively affected external issuance across all the regions. Subsequently, total external issuance dropped to $48 billion in 2008 compared to $100 billion a year before. However, issuance resumed quickly and during the post-crisis period (2009-14) average annual issuance was about $250 billion and grew by an average 24% annually. South Asia (SAR), Africa (AFR) and the Middle East (MNA) regions have been the smallest external issuers, and, although in recent years absolute volumes have increased, they are still among the lowest. Of particular interest is China’s issuance, which grew rapidly since 2009 in the wake of the major credit stimulus driven by banks and real estate developers, and surpassed Latin America (LAC) in 2014. 2. Pre-crisis external issuance was mostly driven by sovereigns whereas post-crisis issuance was dominated by corporates (Figure 5). Issuance by sovereigns and corporates has been increasing on average since 2000 at 5% and 23% annually, respectively. However, the pace of issuance accelerated in the post crisis period, especially for corporates which posted a total issuance of around $300 billion in 2014, compared to $14 billion in 2000. 6 EMDE sovereigns experienced a much more moderate increase in their external issuance, issuing $99 billion in 2014 compared to $50 billion in 2000. 3. Cumulative post-crisis issuance is large relative to country GDP and grew much faster for the poorest countries (Figure 20). For all EMDEs combined, the median cumulative external issuance to GDP ratio was 6.7% in 2009-14, a significant increase from 4.3% in 2002-07. Richer EMDEs are the main issuers, accounting for 85% of total issuance during this period. Yet, the median ratio for the poorest country group (LMIC)2 is 6.2% of GDP, up from 1.9% in 2002-07. This dramatic increase has important implications for sovereign and corporate liability structures in these countries. 4. External issuance of oil exporting EMDEs has also increased and might pose additional risks given recent oil price and U.S. Dollar developments (Figure 20). Total volumes by this group has increased from $68 billion in 2002-07 to more than $100 billion by 2009-14. Cumulatively, external bond issuance in 2009-14 was 3.8% of GDP for the median oil exporter, up from 1.2% in 2002-07. A strong Dollar, current oil price trends, and tightening of international funding conditions all raise financial risks for this group. Outstanding stock trends 5. External debt stocks in absolute terms and relative to the size of the economy have risen to unprecedented levels post-crisis. This is a widespread phenomenon and is not driven by a single country or region (Figure 6). For March 2015, we estimate poorer EMDEs (LMICs) have about $280 billion outstanding while the corresponding figure for richer EMDEs (non-LMICs) is $1.4 trillion. We find that the median ratio of outstanding external bonds issued since 2000 to GDP has risen across all regions. Most of the increase across regions has taken place since 2009 and 2011 when LSAPs in the United States were fully operational and the long-term refinancing operations (LTROs) of the European Central Bank were launched, respectively. In February 2015, the median ratio was largest in LAC with 12.6%, up from 7.5% in 2007. It is also high in Eastern Europe (ECA) and East Asia (EAP, excluding China) standing at 9.2 and 7.8%, respectively. The ratio almost quadrupled in ECA from a crisis nadir of 2.3% in 2008. Similarly, the ratio tripled for MNA to over 6% currently. Currency trends 6. External issuance is still mostly denominated in foreign currencies. As such, the recent trend of a strong U.S. Dollar raises financial vulnerabilities. Local currency issuance has increased, driven by Dim Sum bonds (Figure 7). External issuance has mostly occurred in foreign currencies though the share of local currencies has been increasing gradually. In 2000, around 1% ($327 million) of total issuance by EMDEs was in local currencies and this has increased to 15% ($60 billion) in 2014. A key contributor to the trend are Dim Sum bonds issued offshore by Chinese entities which are denominated in renminbi. Industry and use of proceeds trends 2 These are countries with a GNI per capita of $4,125 or lower, according to World Bank Income group definitions. 7 7. The largest issuing industries include the Finance and Utilities Sectors (Figure 8). Finance captured the largest share among sectors by second half of 2014. This might be driven by the fact that large internationally active banks started to deleverage in the face of stricter regulatory requirements and market pressures. Utilities and Other sectors (which includes agribusiness, forestry and paper, healthcare, chemicals, closed end funds, defense, and government) are the other two sectors with relatively larger volumes of total issuances. 8. Proceeds have mostly been used to finance general corporate activities and public investment. In the wake of the “Taper Tantrum”, refinancing has become a key use (Figure 9). General corporate activities include capital expenditures, R&D expenditures, and other productive investments. Refinancing of debt surged around the “Taper Tantrum” suggesting EMDE entities issued to make their debt profiles less risky while funding conditions were still benign. Public sector uses which cut across industries are also substantial and primarily used for financing community projects at the sovereign and sub- sovereign levels. Maturities and yields trends 9. Average yields of new external issuances have dropped precipitously since the crisis (Figure 10). In 2007, right before the financial crisis, yields stood at 8.4% and have fallen since to about 5% in 2015. As expected, yields of the poorest countries (LMICs) have been consistently higher than for richer EMDEs (non-LMICs). However, the spread between the two has declined steadily from a peak of 4.4% in 2009 to 1.8% in 2015. Taken together, these findings are consistent with search-for-yield motives. 10. The average maturity of external issuances dropped sharply during the crisis. While maturities have increased since, they remain well below pre-crisis levels (Figure 11). Right before the crisis, volume-weighted average maturities were almost 9 years. The crisis triggered a sharp drop to 7.3 years in 2009. While maturities recovered somewhat since, around the time of the Taper Tantrum, they started falling again, reaching 6.7 years by the end of 2013, when search for yield flows resumed. Currently, the average maturity for new issuances is almost 8 years. Maturities in richer EMDEs (non-LMICs) were particularly affected during the crisis, dropping from almost 9 years in 2007 to 7.3 years in 2009. Since then they have been on an upward trend and currently stand at almost 8 years. Post-crisis volatility of maturities have been high for poorer EMDEs (LMICs), reflecting lower deal volume compared to non-LMICs. With that caveat, since 2014, LMIC maturities have been increasing sharply from 6.6 to 8.6 years. Maturity profile of currently outstanding bonds 11. The majority of the $1.7 trillion currently outstanding external bonds of EMDEs will mature before 2024 with a peak in 2019. Richer EMDEs will experience another peak in 2017 (Figure 12). In March 2015, we estimate the outstanding stock of external bonds for EMDEs to be $1.7 trillion, of which $1.5 trillion will mature by 2035. Of this initial $1.7 trillion stock, the average still outstanding monthly amount of bonds maturing within the next 12 months is highest during 2015-19 when it peaks at $207 billion. During this period, the average monthly amount of maturing bonds is $164 billion ($28 billion and $136 billion for LMICs and non-LMICs, respectively). This monthly maturing amount declines during 2020-24 in which the average drops to $109 billion. Non-LMICs experience two peaks of roughly $150 billion in 2017 and 2019. LMICs will experience a 8 single peak in 2019 when the amount that matures within 12 months reaches $40bn. According to current market expectations, these peaks will occur after the Federal Reserve has raised interest rates. 12. By 2020, all regions will have experienced peaks in which more than 10% of their currently outstanding stocks will mature within 12 months (Figures 13 and 14). China’s peak should occur in 2017 in when almost 20% of its currently outstanding bonds will mature ($333 billion). A significant portion of these bonds however are denominated in renminbi which ameliorates currency risks. South Asia peaks in 2019 with 20% of its current stock ($81 billion). Eastern Europe peaks at almost 15% in 2018 (current stock: $239 billion). East Asia (ex-China) peaks at 12% in 2019 (current stock: $174 billion). Africa, Middle East, and Latin America peak at 15%, 15% and 10% in 2020, respectively (current stocks; $64 billion, $42 billion, and $751 billion). Credit quality 13. The credit quality of post-crisis external issuance has improved significantly (Figure 15). Before the crisis, only 30-40% of issuance was investment grade. Since 2010, this fraction has steadily improved from around 50% to 70%. While this is a positive trend, it is important to keep in mind that ratings can be pro-cyclical. IV. Data We now turn to the description of our two data sets that cover the universe of EMDE external bond issuance in the period 2000-14. Table 1 describes the variable definitions. Our data sets matches three types of data: i) highly granular bond data (i.e. industry or bond deal level), ii) high frequency financial global push factors, and iii) country pull factors. Data on bonds are derived from Dealogic which provides information on borrowers, bond yields and non-pricing terms at origination on the individual deal level, which typically comprises several tranches. Global push factors are from Bloomberg and country pull factors are sourced from the IMF’s World Economic Outlook. A. Bond Deals Country-industry panel dataset First, to analyze the impact of global factors on the propensity to issue external bonds by EMDE entities, we compile a balanced panel data set of monthly total external bond issuance for each industry in 71 emerging and developing countries between 2000 and 2014. There are 7 industrial sectors, which translates to 497 individual country-industries for which we have monthly observations. Note that some of these country-industries have not issued externally at all in our sample. Hence the number of country-industry-month observations in the panel is about 84,000. Our dependent variable is a dummy which denotes for a particular country-industry whether its total volume issuance in a given month is above its historical average over the period 2000-07. In doing so, we essentially control for general issuance patterns for each country-industry and ameliorate bias due to absolute size effects. Bond tranche deal dataset 9 Second, to study the impact of global factors on individual bond yield and maturities, we construct a data set which captures the universe of 6,307 individual bond deals for 71 emerging and developing economies in the 2000-14 period. These bonds are issued by 210 country-industries. The other 289 country-industries never issued externally during the sample period. Appendix 1 provides details of issuance activity on the country level. Bonds often consist of multiple tranches with different characteristics. Therefore the number of observations in this dataset is 6,925 bond tranches. Our two bond variables of interest are yield to maturity (defined as the rate of return on a bond assuming the bond is held until maturity at the time of issuance) and maturity (defined as the number of years for which the bond remains outstanding at the time of issuance). We lose tranche observations due to missing data. As a result, we have yield data for 5,962 bond tranches and maturity data for 6,804 (non-perpetual) bond tranches, respectively. This bond tranche level data set allows us to control for bond-specific characteristics that could influence the two variables of interest. We can therefore account for changes in issuance composition over time. These bond tranche level variables include:  Size of bond tranche issued refers to the total U.S. Dollar volume of the individual tranche of the deal;  Currency is an indicator variable that captures the currency in which the tranche is issued;  Investment grade type is a set of indicator variables that indicates whether the bond tranches are investment grade or not – i.e. a credit rating of BBB- or higher according to S&P or Baa3 or higher according to Moody's. This variable allows us to control for adverse selection issues;  Borrower industry is a set of indicator variables that captures the industrial sector of the issuing entity (Consumer, Finance, Metals, Professional Services, Transportation, Utilities, and Other);  Borrower type distinguishes between three different types of borrowing entities, public- local (local and state/provincial authorities), public-other (central government) and non- public; and  Deal type is a set of indicator variables which reflects the type of bond tranche such as Asset Backed Securities, Corporate Bond-High Yield, or Sovereign (see Table 2 for more details). The grouping is defined by Dealogic. B. Global Push Factors We study the impact on external bond issuance of four global push factors that proxy for global financial conditions: 1. The VIX index (VIX) (Figure 16) captures the options-implied 30-day ahead volatility of the S&P 500 equity index and is the most frequently used indicator as a proxy for global risk appetite, risk, and uncertainty. Higher values of VIX are associated with higher bond yields and lower maturities. Research suggests EMPs have contributed to extremely low volatility. 10 2. The Libor-OIS spread (LIBOR) (Figure 17) is used as a control for risk perception in credit markets. This spread is a measure of inter-bank risk and liquidity in the money market and captures fear of bank insolvency. Higher spreads indicate low liquidity and an unwillingness of banks to lend to each other, and are typically associated with higher bond yields and a decrease in maturities. 3. The corporate credit spread (RISK) (Figure 18) tracks the performance of U.S. Dollar denominated investment grade rated corporate debt that is publically issued in the U.S. domestic market. This options-adjusted spread is the difference between U.S. treasury bonds and corporate bonds with a BBB rating or higher. RISK is an indicator of corporate sector health, where wider spreads are associated with deteriorating investor confidence and are expected to increase bond premiums and shorten the duration at which EMs can issue debt. Search-for-yield will exert a downward pressure on this spread. In unreported robustness regressions we use the high-yield corporate debt spread instead with qualitatively similar results. 4. The size of the Federal Reserve Balance Sheet (FED) (Figure 19), calculated as the sum of mortgage- backed securities and U.S. treasuries, is used to gauge spillover effects from U.S. LSAPs.   Panel B in Appendix 2 provides average values of these global push factors around time of each individual bond issue by year. In all our regressions, we also control for the United States 10-year Treasury yield (UST10Y), which is generally considered a pricing benchmark and a proxy for global liquidity conditions as well. Falling U.S. long-term treasury yields are associated with an abundance of capital in the international market and an increased willingness to hold relatively riskier assets, such as emerging and developing market debt. Indeed, the empirical literature has found this global factor to be a key determinant of emerging market bond prices. Notably, an increase in U.S. treasury yields tends to increase emerging market bond yields and spreads while decreasing the probability of bond issuance (e.g. Eichengreen and Mody (1998a) and Eichengreen and Mody (1998b)). Global push factors are all based on daily time series. To best estimate the global financial conditions that impacted bond issuance as well as investor confidence, we incorporate these global factors into our two data sets as follows (See Table 2 for more details on global push factors). For the industry-level dataset, we calculate for each month the average value of each factor for the 6 preceding months. For the bond-level data set, for each individual bond we compute the average value for each factor the 6 months prior to the issuance date. C. Domestic Pull Factors As regards country-specific factors, the analysis controls for five macro-financial variables used to evaluate a country’s development, creditworthiness, and vulnerability. These variables are available on an annual basis and we match the macro variables with the corresponding year for each month in the industry-level panel dataset and the year of the bond issue date in the bond-level dataset:  Real GDP per capita in U.S. Dollars (GDPPC) is used to control for the level of development of a country given its positive correlation with international bond issuance. 11  Real GDP growth rate (GROWTH) is used to proxy for investment opportunities as higher economic growth can potentially drive down bond yields and increase their maturities.  The current account balance expressed as a percent of GDP (CA) is used as larger current accounts can make countries more vulnerable to a slowdown in capital inflows or sudden stops and hence can result in higher yields and shorter maturities on debt issued.  Total external debt as a percentage of GDP (EXT) is used as lower levels of external debt are expected to reduce default risk and boost investor confidence in the economy which can positively impact bond issue prices and maturities.  Total bank credit to the private sector as a percentage of GDP (PCRED) is often used as a proxy of financial depth and development which can enhance resilience to economic and financial shocks, and, in turn, positively impacts bond prices and maturities. While private sector credit is considered a financial variable, it is also an indicator of economic activity – improved economic activity is usually reflected in greater credit growth and potentially in reduced prices and maturities for bonds. D. Descriptive Statistics and Correlations Table 2 provides descriptive statistics. Panel A shows the average yield and maturity at issuance in our universe of bonds during the 2000-14 period was 5.1% and 6 years, respectively. The average bond size was about $123 million. The average propensity for a country-industry to issue above its 2000-07 historical average in any month was 3% (Panel B). All global push factors exhibit very high variation as a result of the pre-crisis boom, the global financial crisis, and the effect of subsequent policy measures, including EMPs, which drove down interest rates, volatility, and risk spreads. Table 3 reports correlations. We document a particularly strong negative unconditional association between individual bond yields in EMDEs and the size of the Fed’s balance sheet around the time of issuance (ρ=-0.58), suggesting that EMPs have contributed to search-for-yield climate to EMDEs. The correlations between bond features and various country characteristics (e.g. PCRED) are also quite high, suggesting pull factors are important as well. Correlations between the global push factors are relatively strong, with the exception of the Fed’s balance sheet. Appendix 2 Panels A, B and C display annual bond issuance characteristics (excluding issuance by Chinese entities) and annual averages of push and pull factors around the time of issuance. Panel D shows the fraction of all country-industries with monthly issuance volume above their historical average by year. A few points are worth highlighting. Panel C shows that the country profile of issuers has changed significantly, with both positive and negative features. Post-crisis, issuing countries are significantly richer than before the crisis as measured by GDP per capita (2010-14: $7,800 vs. 2000-07: $4,300). They also have deeper financial systems as proxied by private credit to GDP (2010-14: 51% vs. 2000-07: 36%). Moreover, they have lower levels of external debt to GDP (2010-14: 39% vs. 2000-07: 48%). However, at the same time the current account and economic 12 growth of issuing countries has deteriorated significantly, particularly during 2011-14 (3.8% and -3% of GDP, respectively). Panel D clearly shows the presence of synchronized external issuance waves on the country- industry level, even after correcting for historical average issuance patterns of individual country- industries. In the run up to the crisis, the average monthly fraction of country-industries with higher issuance than their own average during 2000-07 climbed from 1.59% in 2002 to 3.35% in 2006. This fraction fell to 1.14% during the height of the crisis in 2008. However, the fraction has increased again since 2010 to record levels from 3.67% in 2010 to 5.30% in 2013. V. Methodology This section describes our econometric approach to analyze research questions 2, 3, and 4 of this paper. A. Modeling the Propensity to Issue Externally on the Country-Industry Level To address the first research question, we fit logistic regressions on our industry-level panel data set to test the impact of our global factors on the tendency of country-industries in EMDEs to issue external bonds above their own historical average. By comparing monthly issuance of a country- industry to its own historical average issuance volume, we effectively control for country-industry level issuance trends. In all regressions, we cluster standard errors on the country-industry level to allow for within industry correlation. We estimate the issuance propensity for a particular country- industry as: _ _ 1 F (1) where _ _ is an indicator variable which assumes a value of 1 if total issuance volume in industry s in country i during month t is above the pre-crisis historical monthly average of industry s during 2000-07 and 0 otherwise. F(·) denotes the cumulative logistic distribution. denote vectors of time-varying explanatory variables that contain global push factors (INT) and domestic pull factors (DOM). The vector of global factors consists of , where ∈ , , , . In other words, we always control for the United States 10 year Treasury rate. The vector of domestic factors is defined as: , , , , . Importantly, we include a battery of fixed effects. We account for time-invariant country factors such as the overall institutional environment, the macro-financial framework, and the level of development of the country which influences investment opportunities and investor appetite. We include year factors to capture the overall impact of global conditions such as trade and crisis effects. As such, we exploit within-year variation and avoid drawing false inference due to general cyclical or time trends. Finally, we include industry factors to capture intrinsic differences between industries in terms of their need for and use of external bond finance. 13 B. Modeling Yields and Maturities on the Bond Tranche Level We estimate pooled OLS regressions on the bond tranche-level dataset to evaluate the impact of global factors on the pricing and maturity of bonds. Again, in all regressions, we cluster standard errors on the country-industry level to allow for within industry correlation. The model can be written as: (2) where denotes the yield to maturity3 or the maturity of bond tranche b. The first two vectors capture global push factors (INT) and domestic pull factors (DOM) around the time bond b was issued, as described above. is a vector of bond-specific characteristics: Size of bond issued, Currency, Investment grade, Borrower industry, and Deal type. For yield to maturity regressions, we also include Maturity in . Importantly, allows us to isolate the impact of issuance composition and bias effects (e.g. differences in bond risk, bond size or industry) so we can make much stronger inference than is possible at higher levels of aggregation where such information is lost. We also incorporate two sets of indicator variables that capture general global conditions such as global trade and general crisis effects in the year which bond b was issued ( ) as well as time-invariant factors associated with the country in which bond b was issued ( ). Missing data in limits the sample size. Therefore, in robustness regressions we substitute and the country fixed effects for country-period fixed effects. Our model becomes: (3) VI. Empirical results This section addresses research questions two through five. It summarizes and discusses the main empirical results for the impact of global push factors on external bond issuance in EMDEs. Given that the global factors are relatively highly correlated, we estimate their effects in separate regressions. A. Impact of Global Factors on the Propensity to Issue Externally on the Country-Industry Level Table 4 shows the results of the logistic regressions that estimate Equation (1) on the country- industry-month level in EMDEs during the 2000-14 period. All four global push factors (VIX, RISK, FED, and LIBOR) are highly statistically significant on the 1-percent level with the expected sign. This finding supports the notion that external issuance across EMDEs is highly synchronized with the global financial cycle which triggers capital flows out of developed markets in search for yield in EMDEs. 3 Note that because the regression controls for the 10-yr U.S. government yield, the results can also be interpreted as if the dependent variable were a “spread”. 14 Model 1 shows that industries are less likely to issue above their historical 2000-07 average if VIX increases even after controlling for the UST10Y, and time-varying and time-invariant (e.g. country fixed effects) country pull factors. The result is also economically significant. A 10% increase in the VIX leads to a decline in the odds an industry will issue above its average by almost 6% (1.1^(- 0.63)-1). Model 2 shows that a decrease in the BBB U.S. corporate credit spread (RISK) lowers the odds of above-average issuance even more than for the VIX. These odds drop by 10% for a 10% increase in RISK. Model 3 indicates that an increase in the size of the Fed’s balance sheet (FED) boosts the odds of above-average issuance. The coefficient suggests that a 10% increase in the Fed balance sheet increases the odds by 8%. Finally, Model 4 shows that lower interbank risk increases the above-average issuance odds. A 10% decline in LIBOR increases the odds by 5.5%. Table 4 also consistently shows that GDP per capita (GDPPC), GDP growth (GROWTH), and the current account (CA) are the most important country pull factors. Industries in countries with higher GDPPC and GROWTH are more likely to issue above their historical 2000-07 average volume in a given month. This could reflect both demand and supply factors: industries in more developed or faster growing countries could have a higher need for external finance while investors have more appetite to supply it given lower risks. Similarly, industries in countries with current account surpluses are less likely to issue above average, perhaps since countries with surpluses are net exporters of capital. We don’t find evidence that other macro pull factors such as external debt (EXT) or financial development (PCRED) of the country contain additional information. In unreported robustness regressions we use the MOVE index and obtain qualitatively similar results as for the VIX in Model 1. The MOVE Index captures expected U.S. Treasury volatility and acts as a proxy for interest rate uncertainty. Higher values indicate greater uncertainty. More specifically, the Merrill lynch Option Volatility Estimate Index is a yield curve weighted index of the normalized implied volatility on 1-month Treasury options which are based on the 2, 5, 10, and 30 year contracts. Intuitively, MOVE is similar to VIX for the government bond market. In another set of unreported robustness regressions, we assess the impact of the U.S. Economic Policy Uncertainty (EPU) Index as developed in Baker et al (2015), but do not find any statistically significant results. The EPU Index is based on three types of underlying components:4 “One component quantifies newspaper coverage of policy-related economic uncertainty. A second component reflects the number of federal tax code provisions set to expire in future years. The third component uses disagreement among economic forecasters as a proxy for uncertainty.” B. Impact of Global Factors on Yields of External Bonds at Time of Issuance Table 5 presents bond tranche-level OLS regressions which document the impact of the global factors on individual bond yields in EMDEs during the 2000-14 period. For each global push factor we present two models to estimate Equations (2) and (3), respectively. We exclude Chinese 4 For details, see http://www.policyuncertainty.com/. 15 issuance in the second model to avoid a possible China bias since 2,945 bonds in the sample (consisting of 3,143 tranches) are issued by Chinese entities. A consistent picture emerges in which favorable global conditions bring down bond yields across EMDEs in a synchronized manner. Since the regressions control for the 10-year U.S. treasury yield (UST10Y), the results also imply that the “spread” (see footnote 3) relative to U.S. treasuries falls when global factors are benign. Except for one model, all results are significant at the 5 percent level at least. Models 1 and 2 demonstrate that a decrease in the VIX is associated with lower bond “spreads” across EMDEs. A 10% decrease in the VIX decreases the EMDE bond “spread” by 6 to 12 basis points. Model 2 excludes Chinese issuance and adds data for 11 countries by dropping time- varying country factors and produces a result that is significant on the 1 percent level and doubles in magnitude. Models 3 and 4 show that the impact of RISK is strong and highly significant as well. A 10% decrease in RISK decreases EMDE bond “spreads” by 12-13 basis points. Models 5 and 6 are also highly significant and indicate that a 10% increase in the Fed’s balance sheet size brings down EMDE bond “spreads” by 8-9 basis points. Finally, a 10% fall in the LIBOR-OIS spread is significantly associated with a reduction in EMDE bond spreads by 3-6 basis points. As expected, we find that the UST10Y and bond maturity have a consistent positive impact on the yield. The size of the bond does not contain additional explanatory power. Unreported regressions show the level of economic development (GDPPC) as well as economic growth (GROWTH) are significantly negatively associated with spreads, as expected. However, after inclusion of year fixed effects the GDPPC coefficient switches sign and GROWTH is no longer significant. This suggests global factors play a more significant role. Again, in unreported robustness regressions we use the MOVE index and obtain qualitatively similar results as for the VIX in Models 1 and 2. In another set of unreported robustness regressions assessing the impact of the U.S. Economic Policy Uncertainty (EPU) Index, we again do not find any statistically significant results. C. Impact of Global Factors on Maturities of External Bonds at Time of Issuance Table 6 documents bond tranche-level OLS regression results which show the impact of global factors on maturities of non-perpetual external bonds issued during 2000-14. The standard errors are clustered on the country-industry level. As in Table 5, for each global push factor we present two models to estimate Equations (2) and (3), respectively. Overall, we find that favorable global factors are associated with a maturity extension across EMDEs. This result is consistent with a willingness of investors to extend maturities when global liquidity is ample and search for yield effects are strong. However, the results are somewhat weaker in terms of statistical significance, compared to the impact on yields. Models 1 and 2 suggest that a 10% fall in VIX extends bond maturities by 16-17 weeks, although Model 2 is only significant at the 10-percent level. The results in Models 3 and 4 are statistically strongest, both at the 1 percent level, and suggest that a 10% fall in RISK boosts maturities by 17- 16 24 weeks. Only Model 5 is statistically significant and suggests a 10% increase in the Fed’s balance sheet increases maturities by 14 weeks. We do not find strong evidence of a significant impact of a lower LIBOR-OIS spread although the coefficient has the expected sign, suggesting that a lower spread has a positive impact on maturities. Across regressions we also find evidence that larger bonds typically carry longer maturities. This is in line with expectations since larger issuers are typically able to issue at longer maturities. As regards country characteristics, in unreported regressions we find that economic growth (GROWTH) has a strong significant positive impact on maturities, consistent with expectations as well. However, after inclusion of year fixed effects, GROWTH is no longer significant. In unreported robustness regressions we use the MOVE index and obtain similar results as for the VIX in Model 1 and 2. The impact of the U.S. Economic Policy Uncertainty (EPU) Index is again statistically insignificant. D. Interaction of Country Characteristics with Global Factors Appendix 3 contains 60 additional regressions in which we investigate whether country characteristics amplify or dampen the impact of our four global factors ∈ , , , . In doing so, we modify Equations (1) and (2) by sequentially adding an interaction between a global factor and a country variable from , , , , . This strategy produces 4*5=20 additional regressions for each independent variable. For all 40 bond tranche level regressions (i.e. Equation (2)), we omit Chinese bonds to avoid a China bias. While some of these interactions are statistically significant, we do not find consistent evidence across the board that country variables amplify the effect of global factors. This suggests search- for-yield flows during loose global funding conditions do not strongly discriminate between EMDEs but are primarily driven by global factors. In that context, in our 20 additional augmented Equation (1) regressions we highlight that the interaction with PCRED is significant for VIX and FED at the 1-percent level and RISK at the 10- percent level. This suggests higher financial development could amplify benign global factors and raise the odds that a country-industry will issue above its historical average. In addition, for our 20 additional Equation (2) regressions to explain individual bond yields (and “spreads”), we document that the interaction with GROWTH is significant for VIX and RISK at the 5-percent level and LIBOR at the 1-percent level. We also find significant interactions for EXT with RISK and LIBOR at the 5-percent level. These findings provide some support for the notion that country growth and external debt can amplify the impact of these global factors on individual bond yields and spreads. We do not find any strong results in our 20 additional Equation (2) regressions to explain individual bond maturities, indicating that maturities are not significantly differently affected by global factors across EMDEs with different domestic characteristics. 17 E. The Risk-Taking Channel of Exchange Rate Appreciation Following Borio and Zhu (2012) and Bruno and Shin (2015a, 2015b), in this section, we briefly explore the risk-taking channel of financial conditions and monetary policy in developed countries via exchange rate appreciation. As described earlier, Bruno and Shin (2015b) argue that looser financial conditions are associated with an increase in cross-border capital flows intermediated through higher leverage in the international banking system. The mechanism operates via stronger local borrower balance sheets as a result of local currency appreciation, allowing banks to lend them more and take on more risk. We test whether this risk-taking channel is active for international investors and external bond issuance as well. Under that hypothesis we would expect U.S. Dollar depreciation/local currency appreciation to be associated with a higher propensity for country-industries to be able to issue higher external bond volumes, all else equal. We use two exchange rate variables as global push factors, following Bruno and Shin (2015a, 2015b):  The 6-month log difference of the U.S. real effective exchange rate (USREER), along the lines of the VAR framework in Bruno and Shin (2015b). USREER is a trade-weighted Dollar index. Higher values imply a real depreciation of trade partner currencies (appreciation of the U.S. Dollar).  The 6-month log difference of the real U.S. Dollar – Local currency exchange rate (XRATE), which is similar to the panel regression setting in Bruno and Shin (2015a). XRATE reflects the real bilateral exchange rate where higher values indicate a real depreciation of the local currency (appreciation of the U.S. Dollar). We use the 6-month log difference of the real exchange rate which is calculated as the log of the nominal exchange rate multiplied by the U.S. CPI and divided by the local CPI. Table 7 presents the results. Analogous to Table 4, Models 1 and 2 estimate Equation (1) and provide strong support for the risk-taking channel of exchange rate appreciation hypothesis: the coefficients on USREER and XRATE are negative and highly statistically significant5. This indicates that the propensity to issue bonds externally above historical average volumes for a particular country-industry is significantly higher when the U.S. Dollar depreciates in real terms in the 6 months prior. In other words, when the local currency appreciates, local borrowers’ balance sheets strengthen. This in turn increases their external borrowing capacity which triggers higher cross-border flows by international investors who are willing to take on more risk. These results are closely tied to our findings in Table 4 of the impact of the VIX on external bond issuance volume discussed in Section A. Particularly, the results point to a channel through which the VIX operates since Bruno and Shin (2015b) document a link between the VIX and USREER. 5 We are aware that there may be potential endogeneity issues in that the local currency appreciation could also be the result of capital inflows. While the use of the 6-month prior exchange rate difference should help to address this issue, we leave it to future work to examine it further. 18 VII. Conclusion and Policy Implications Using the universe of all externally issued bonds during the 2000-14 period, this paper shows the post-crisis period has seen an unprecedented surge in external bond issuance and stocks across emerging and developing economies (EMDEs). Bond yields (and spreads) at the time of issuance have fallen to record lows, in part as a result of loose global funding conditions produced by extraordinary monetary policies (EMPs) in developed economies. In particular, the volume of bonds issued in the six post-crisis years tripled to $1.5 trillion compared to the six years before the crisis, overwhelmingly denominated in foreign currencies and driven by corporate issuance. This surge is not driven by a single region or country, but reflects a broad-based trend, since the 2009-14 cumulative external issuance to GDP ratio is 6.7% for the median EMDE, up from 4.3% in the pre-crisis period. The trend is also present at the country- industry level across EMDEs even after we correct for their own historical issuance average. Under such benign conditions, many EMDEs issued externally for the first time, including Armenia, Angola, Ghana, Laos, and Tanzania. Contrasting the pre- and post-crisis periods, we find that countries of external issuers currently are on average richer, have deeper financial systems, and lower external debt. The fraction of issuance that is rated investment grade has also improved. However, these countries currently also have much slower GDP growth and larger current account deficits which can weaken debt servicing capacity and raise external vulnerabilities. This paper also finds that global factors have a powerful impact on primary activity in international bond market by corporates and sovereigns EMDEs. Controlling for United States interest rates, a battery of country pull factors, and year fixed effects to account for the overall impacts of major global conditions and time trends, we find that a decrease in i) expected U.S. equity market (or interest rate) volatility, ii) U.S. corporate credit spreads, iii) U.S. interbank funding costs and iv) an increase in the Federal Reserve’s balance sheet: 1. Raise the odds that a country-industry’s monthly external issuance volume is above its own historical average. For example, a doubling (halving) of the Fed’s balance sheet increases these odds by about 75% (-43%); 2. Lower the yield-to-maturity spread of external bonds at the time of issuance, even after accounting for individual bond characteristics (e.g. volume, currency, riskiness, industry, type of issuer). For example, a doubling (halving) of the Fed’s balance sheet lowers (increases) a bond’s spread by 63 basis points; and 3. Increase the maturity of non-perpetual external EMDE bonds at the time of issuance, again after accounting for individual bond characteristics. For example, a doubling (halving) of the Fed’s balance sheet is associated with a maturity lengthening (shortening) of 48 weeks. We also find empirical support that the risk-taking channel of exchange rate appreciation (e.g. Bruno and Shin (2015b)) also operates for external bond issuance: real depreciation of the U.S. Dollar is associated with a higher propensity for country-industries to issue externally above their historical average volume. More specifically, when the local currency appreciates, local borrowers’ balance sheets strengthen. This in turn increases their external borrowing capacity which triggers higher cross-border flows by international investors who are willing to take on more risk. This process can be self-sustaining, at least for a while. 19 In addition, in line with the literature, we find evidence that some country characteristics such as the level of financial development can affect the impact of global factors. However, the results are not consistently statistically significant implying that the global cycle is mostly driven by push factors and does not structurally discriminate between EMDEs. Taken together, our findings provide strong support for synchronized primary issuance flows across EMDEs driven mostly by global factors. As a result, both sovereigns and corporates in EMDEs have collectively been able to take advantage of ample international liquidity by lowering their borrowing costs and extending maturities which can improve risk profiles, although in the wake of the crisis, maturities in EMDEs remain below pre-crisis levels. The massive and widespread external issuance in EMDEs raises important questions regarding the impact of procyclical investor behavior once the global cycle winds down, or if global shocks materialize, with potential systemic implications for EMDEs. Moreover, while issuance at lower cost and maturity extension can help lower individual borrower risk profiles, large foreign currency exposures raise risks, particularly for unhedged issuers. The recent trend of a rapidly strengthening U.S. Dollar against most EMDE currencies further heightens currency risks. In this context, the inevitable exit from EMPs will tighten international funding conditions, which could prove disruptive for currencies, balance sheets, and funding capacity in EMDEs. Additionally, fragility in EMDEs can be further compounded by their shallow local financial markets and a lack of strong institutions, supervisory and surveillance capacity, and technical experience. As such, in terms of financial sector policies, there is a continued need for, inter alia: i) creating vibrant local currency (corporate) bond markets and an active, diverse domestic investor base; ii) building macroprudential tools and monitoring capacity to deal with synchronized foreign investor activity to prevent or manage a situation where certain flows create a variety of risks which jeopardize undoing financial and (socio-)economic progress made; and iii) strengthening data collection efforts, particularly regarding foreign currency exposures and hedges. References Adrian, T., and H. S. Shin (2010). “Liquidity and leverage,” Journal of Financial Intermediation, 19(3), pp. 418-437. Avdjiev, S., M. Chui, and H.S. Shin (2014). “Non-Financial Corporations from Emerging Market Economies and Capital Flows”, BIS Quarterly Review, December 2014. Baker, S., N. Bloom, and S. Davis (2015). “Measuring Economic Policy Uncertainty”, Stanford University Working Paper. Bekaert, G., M. Hoerova, and M. Lo Duca (2013). “Risk, uncertainty and monetary policy” Journal of Monetary Economics 60, pp. 771-788. Borio, C. and H. Zhu (2012). “Capital regulation, risk-taking and monetary policy: a missing link in the transmission mechanism?” Journal of Financial Stability, 8(4), pp. 236-251. Bruno, V. and H.S. Shin (2015a). “Cross-Border Banking and Global Liquidity”. Review of Economic Studies 82, pp. 535-564. 20 Bruno, V and H.S. Shin (2015b). “Capital Flows and the Risk Taking Channel of Monetary Policy”. Journal of Monetary Economics 71, pp. 119-132. Burns, A., M. Kida, J. Lim, S. Mohapatra, M. Stocker (2014). Unconventional Monetary Policy Normalization in High-Income Countries Implications for Emerging Market Capital Flows and Crisis Risks. World Bank Policy Research Working Paper 6830. Calvo, G., L. Leiderman, L. and Reinhart, C. (1993). “Capital Inflows and Real Exchange Rate Appreciation in Latin America: the Role of External Factors.” IMF Staff Papers: 108-151. Cerruti, E., Claessens, S. and Ratnovski (2014). “Global Liquidity and Drivers of Cross Border Bank Flows” IMF Working Paper 14/69. Fratzscher, M., M. Lo Duca, and R. Straub (2014). “On the International Spillovers of U.S. Quantitative Easing”, European Central Bank Working Paper 1557. Eichengreen, B. and P. Gupta (2013). “Tapering Talk: The Impact of Expectations of Reduced Federal Reserve Purchases on Emerging Markets”. World Bank Policy Research Working Paper 6754. Eichengreen, B., and A. Mody (1998a). "Interest Rates in the North and Capital Flows to the South: Is There a Missing Link?", International Finance, Vol. 1, No. 1, pp. 35-57. Eichengreen, B., and A. Mody (1998b). "What Explains Changing Spreads on Emerging-Market Debt: Fundamentals or Market Sentiment?," NBER Working Paper 6408. Fratzscher M. (2011). “Capital Flows Push Versus Pull Factors and the Global Financial Crisis”. ECB Working Paper No. 1364. Fratzscher, M., M. Lo Duca, and R. Straub (2013). “On the International Spillovers of U.S. Quantitative Easing”, European Central Bank Working Paper 1557. Ghosh A. Kim, J. Qureshi, M. and J. Zalduendo (2012). “Surges”. IMF Working Paper WP/12/22. International Organization of Securities Commissions (IOSCO) (2014). “Securities Markets Risk Outlook 2014-2015”, October 2014. Koepke, R. (2014). “Fed Policy Expectations and Portfolio Flows to Emerging Markets”, Institute of International Finance Working Paper. Koepke, R. (2015). “What Drives Capital Flows to Emerging Markets? A Survey of the Empirical Literature” Institute of International Finance Working Paper. Miranda-Agrippino S. and H. Rey (2012). “World asset Markets and Global Liquidity” presented at the Frankfurt ECB BIS Conference, London Business School, mimeo, February. Rajan, R. (2013).”A Step in the Dark: Unconventional Monetary Policy after the Crisis”. Bank for International Settlements, Andre Crockett Memorial Lecture, June 2014. 21 Rey, H. (2013). “Dilemma not Trilemma: The global financial cycle and monetary independence”. Federal Reserve Jackson Hole Symposium. Shin, H. S. (2013). “The Second Phase of Global Liquidity and Its Impact on Emerging Economies”, mimeo Princeton University. Turner, P. (2014) “The global long-term interest rate, financial risks and policy choices in EMEs”. BIS Working Paper 441. 22 Figure 1: Bonds Funds Allocations ($ billions) 80 200 AFR ECA MNA 70 180 SAR EAP (RHS) LCR (RHS) 160 60 140 50 120 40 100 30 80 60 20 40 10 20 0 0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Source: EPFR; Author’s calculations Figure 2: Equities Funds Allocations ($ billions) 80 600 AFR ECA 70 MNA EAP (RHS) 500 LCR (RHS) SAR (RHS) 60 400 50 40 300 30 200 20 100 10 0 0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Source: EPFR; Author’s calculations Figure 3: Foreign participation in local currency government bond markets (%) 50 Share of total (current; percent) 40 Share of total (7‐year average; percent) 30 20 10 0 Turkey Africa Mexico Colombia Thailand Hungary Romania Indonesia Poland South Brazil Malaysia Source: IMF Global Financial Stability Report (2014) 23 Figure 4: Total External Volume Issued by EMDEs Figure 5: Total External Volume Issued by EMDEs, by (billions USD) borrower type (billions USD) 16 35 14 EAP ECA LCR MNA 30 Corporate 12 SAR AFR 25 Sovereign 10 China 20 8 6 15 4 10 2 5 0 0 Figure 6: Outstanding External Bonds as % of GDP by Figure 7: Total External Volume Issued by EMDEs, by EMDEs - Medians currency (billions USD) 35 14 EAP ECA Local Currency Local Currency‐China 30 12 LCR MNA SAR AFR 25 Non‐Local Currency 10 China Billion, USD 20 8 15 6 10 4 5 2 0 0 Figure 8: Total External Volume Issued by EMDEs, by Figure 9: Total External Volume Issued by EMDEs, by use industry (billions USD) of proceeds (billions USD) 12 CONS FIN 8 ACQ OTH META OTH 20 10 PROFS TRAN PFIN RFIN UTIL 6 WCAP Unknown 16 8 PUBF GCP (rhs) 12 6 4 4 8 2 2 4 0 0 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 24 Figure 10: Yields of External Issuance by EMDEs, by Figure 11: Maturities of New External Issuance by income level EMDEs, by income level 11 7 12 LMICs 10 WB Countries 9 10 Non‐LMICs (rhs) 6 9 8 8 8 5 7 6 6 LMICs 7 4 4 WB Countries 5 Non‐LMICs (rhs) 4 3 2 6 Figure 12: Maturity Profile Outstanding External Bonds, Figure 13:Maturity Profile Outstanding External Bonds, by Income Group (billions USD) per March 2015 by Regions (billions USD) 40 160 90 LMIC EAP 35 140 80 ECA Non‐LMICs (rhs) 70 LCR 30 120 MNA 60 SAR 25 100 AFR 50 China 20 80 40 15 60 30 10 40 20 5 20 10 0 0 0 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 Figure 14: Maturity Profile Outstanding External Bonds, Figure 15: Credit Quality of External Issuance (billions % of stock in March 2015 USD) 20% 35 EAP ECA Investment Grade China 18% LCR MNA 30 Investment Grade Other 16% SAR AFR EMDEs 25 14% Non‐Investment Grade China Billions, USD 12% 20 10% 15 8% 6% 10 4% 5 2% 0 0% 2015m3 2016m3 2017m3 2018m3 2019m3 2020m3 2021m3 2022m3 2023m3 2024m3 2025m3 2026m3 2027m3 2028m3 2029m3 2030m3 2031m3 2032m3 2033m3 2034m3 2035m3 25 Figure 16: VIX Index (% per annum) Figure 17: Libor-OIS Spread (bps) 90 400 80 350 70 300 60 250 50 200 40 150 30 20 100 10 50 0 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Figure 18: BofA Merrill Lynch U.S. Corporate BBB Figure 19: Federal Reserve Balance Sheet Size (billions Index OAS (%) USD) 9 5,000 8 4,500 7 4,000 3,500 6 3,000 5 2,500 4 2,000 3 1,500 2 1,000 1 500 0 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Figure 20: Cumulative total bond issuance by emerging and developing economies 6 pre-crisis years (2002-2007) 6 post-crisis years (2009-2014) Issuance in Issuance in Total foreign Total foreign Issuance Median currencies Issuance Median currencies (Bil USD) Issuance/GDP (Bil USD) (Bil USD) Issuance/GDP (Bil USD) EMDEs 519 4.3 498 1492 6.7 1323 Low and Low-Middle 87 1.9 82 246 6.2 238 Income Other 432 6.2 416 1246 9.1 1084 Oil Exporters 68 1.2 64 110 3.8 102 26 Table 1: Definition of Variables Variable Definition Source Bond Variables Fixed Yield-to-Maturity Rate of return on security assuming it is held until maturity at Dealogic time of issuance, weighted by deal volume. Maturity of Bond issued Duration (Years) of bonds weighted by deal volume Dealogic Log of Size of Bond Issued Log of total proceeds of bond deal (U.S. Dollars) Dealogic Currency Denotes the currency in which the bond issue is priced, either: Dealogic U.S. Dollar, Euro, British Pound Sterling, Japanese Yen, Australian Dollar, Canadian Dollar, or Other Investment Grade Indicator value with value 1 if a bond tranche is investment Dealogic grade rated and 0 otherwise (credit rating is BBB- or higher according to S&P or Baa3 or higher according to Moody's) Borrower Industry Type of industry: Consumer, Finance, Metals, Professional Dealogic Services, Transportation, Utilities, and Other Borrower Type Type of the borrowing entity, either: Local or State/Provincial Dealogic Authority, Central Government, Non-Public Deal Type Type of security offered, either of the following product Dealogic types: Asset Backed Securities, Corporate Bond-High Yield, Corporate Bond-Investment-Grade, Covered Bond, Medium- Term Note, Money Market, Mortgage-Backed Security, Non- U.S. Agency, Preferred Share, Short-term Debt, Sovereign, Local Authority UST10Y 6 month trailing average of 10Y U.S. Treasury Constant Bloomberg Maturity Rate Global Push Factors VIX Log of 6 month trailing average of VIX index. VIX captures Bloomberg the implied 30-day ahead market volatility derived from S&P 500 index options. RISK Log of 6 month trailing average of U.S. Corporate BBB Bloomberg Option Adjusted Spread. FED Log of 6 month trailing average of Fed Balance Sheet (Sum Bloomberg of Mortgage Backed Securities and U.S. treasuries) LIBOR Log of 6 month trailing average of 3 Month Libor-OIS Spread Bloomberg (3 Month Libor less 3 Month USD Overnight Indexed Swap) MOVE Log of 6 month trailing average of the MOVE index. The Bloomberg Merrill lynch Option Volatility Estimate (MOVE) Index is a yield curve weighted index of the normalized implied volatility on 1-month Treasury options which are weighted on the 2, 5, 10, and 30 year contracts. 27 Variable Definition Source USREER 6 month log difference of U.S. Real Effective Exchange rate Bloomberg from BIS. Base year is 2010 (Weighted basket of foreign currencies vs USD) XRATE 6 month log difference of real exchange rates of each EMDE Bloomberg country in the sample (USD to country local currency) Domestic Pull Factors GDPPC Real GDP per capita in U.S. Dollars IMF World Economic Outlook GROWTH Year-on-year percentage changes in real GDP IMF World Economic Outlook EXT Total debt owed to nonresidents repayable in currency, goods, IMF World or services as a percent of GDP Economic Outlook CA Current account balance as a percent of GDP IMF World Economic Outlook PCRED Total domestic private credit to the real sector by deposit IMF money banks as a percent of GDP International Financial Statistics Country-Industry Issuance ABOVE_AVG_ISSUANCE Indicators variable which assumes value 1 for a given month Author’s in which a country-industry’s total external bond issuance is calculations above its monthly 2000-07 average and 0 otherwise 28 Table 2. Descriptive Statistics of Variables Panel A. Bond tranche data Std. Variable Obs. Mean Min Max Dev. Bond Variables Fixed Yield-to-Maturity 5962 5.06 3.25 0.20 12.31 Maturity of Bond issued 6804 6.00 7.31 0.10 100.08 Log of Size of Bond Issued 6925 18.63 1.49 11.51 22.63 Global Push Factors VIX 6925 2.86 0.28 2.41 3.95 RISK 6925 0.69 0.29 0.15 1.97 FED 6925 14.44 0.67 13.08 15.23 LIBOR 6573 2.84 0.49 1.92 5.09 MOVE 6925 4.38 0.24 4.03 5.22 USREER 6925 -0.002 0.026 -0.074 0.098 XRATE 6883 -0.015 0.057 -0.319 1.449 UST10Y 6925 3.08 1.14 1.66 6.36 Domestic Pull Factors GDPPC 6894 8.62 0.59 6.10 9.63 GROWTH 6897 6.02 3.06 -14.80 34.50 EXT 6918 27.63 29.28 1.30 203.70 CA 6922 -0.15 4.48 -39.50 35.50 PCRED 6905 84.56 48.16 2.23 135.76 29 Panel B. Country-Industry data Std. Variable Obs. Mean Min Max Dev. Bond Variables ABOVE_AVG_ISSUANCE 89957 0.03 0.17 0 1 Global Push Factors VIX 89957 2.99 0.33 2.42 3.95 RISK 89957 0.73 0.41 0.15 1.97 FED 72065 14.00 0.68 13.08 15.24 LIBOR 78029 2.91 0.75 1.92 5.08 MOVE 89957 4.54 0.27 4.04 5.20 USREER 89957 -0.003 0.034 -0.074 0.098 XRATE 84707 -0.007 0.084 -0.570 1.503 UST10Y 89957 3.88 1.17 1.66 6.36 Domestic Pull Factors GDPPC 87696 7.79 0.99 4.69 9.64 GROWTH 87780 4.82 4.41 -14.80 59.74 EXT 88452 49.74 35.39 1.30 282.90 CA 88788 -3.82 9.17 -49.80 35.50 PCRED 83328 38.00 26.29 1.97 135.76 30 Table 3. Correlations between Key Variables Fixed Yield- Maturity of Log of Size of to-Maturity Bond issued Bond Issued Global push factors VIX 0.1891 0.0251 0.0945 RISK -0.0677 -0.087 -0.0006 FED -0.5751 -0.2587 -0.2324 LIBOR -0.1391 -0.0951 -0.0471 MOVE 0.3311 0.1099 0.1641 USREER -0.1492 -0.1159 -0.0979 XRATE -0.0962 -0.0397 -0.0575 UST10Y 0.5469 0.2319 0.2039 Domestic Pull Factors GDPPC -0.2249 -0.0394 -0.0429 GROWTH -0.358 -0.2328 -0.3234 EXT 0.3692 0.2112 0.2807 CA -0.2592 -0.1892 -0.2212 PCRED -0.6565 -0.3912 -0.4685 Global push factors VIX RISK FED LIBOR MOVE USREER XRATE RISK 0.7864 FED -0.2948 0.1282 LIBOR 0.6738 0.8203 0.1469 MOVE 0.7908 0.5221 -0.5604 0.3982 USREER -0.0118 0.0171 0.2012 0.0792 -0.2156 XRATE -0.0885 -0.0361 0.1398 0.0065 -0.146 0.3598 UST10Y 0.0983 -0.3783 -0.836 -0.4811 0.408 0.8279 0.0249 Domestic Pull Factors EXT CA PCRED GDPPC -0.0936 -0.0028 0.2108 GROWTH -0.3165 0.3445 0.5031 EXT -0.5258 -0.3839 CA -0.5258 0.3815 PCRED -0.3839 0.3815 31 Table 4. Bond Issuance Logit Regression Results Dependent variable: Country-industry monthly issuance above 2000-07 average (1=Yes, 0=No) (1) (2) (3) (4) VIX -0.603*** (0.185) RISK -0.789*** (0.199) FED 0.728*** (0.193) LIBOR -0.494*** (0.0833) GDPPC 1.543*** 1.544*** 1.435*** 1.431*** (0.384) (0.384) (0.431) (0.443) GROWTH 0.0294** 0.0294** 0.0339** 0.0375*** (0.0133) (0.0133) (0.0156) (0.0144) EXT 0.00520 0.00520 0.00630 0.00709* (0.00358) (0.00358) (0.00433) (0.00411) CA -0.0329*** -0.0329*** -0.0315** -0.0289** (0.0121) (0.0121) (0.0124) (0.0127) PCRED 0.000204 0.000194 0.00255 0.00227 (0.00757) (0.00758) (0.00831) (0.00791) UST10Y -0.0143 -0.0400 -0.0858 -0.196** (0.100) (0.0960) (0.0990) (0.0962) Country fixed effects Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Observations 79,464 79,464 61,824 69,048 No. of Countries 64 64 62 64 No. of Country-Industries 448 448 434 448 Pseudo R-squared 0.359 0.360 0.352 0.356 Robust standard errors clustered on the country-industry level in parentheses. *** p<0.01, ** p<0.05, * p<0.1 32 Table 5. Bond Pricing OLS Regression Results Dependent Variable: Fixed yield-to-maturity of Bond Tranche (1) (2) (3) (4) (5) (6) (7) (8) VIX 0.550* 1.141*** (0.297) (0.300) RISK 1.219*** 1.266*** (0.299) (0.376) FED -1.125*** -0.837*** (0.295) (0.308) LIBOR 0.335** 0.556*** (0.146) (0.145) GDPPC 0.777** 0.742** 0.740** 1.227*** (0.320) (0.323) (0.325) (0.238) GROWTH -0.0187 -0.0206 -0.0200 -0.0284 (0.0229) (0.0231) (0.0219) (0.0299) EXT (0.00754) (0.00761) (0.00761) (0.00950) 0.0225 0.0217 0.0213 0.0236 CA (0.0215) (0.0215) (0.0214) (0.0218) 0.00867 0.00857 0.00835 0.0139 PCRED -0.000675 -0.000986 -0.000631 0.000119 (0.00571) (0.00576) (0.00569) (0.00666) Log of Size of Bond Issued -0.0441 -0.0666 -0.0495 -0.0684 -0.0506 -0.0683 -0.0318 -0.0479 (0.0479) (0.0706) (0.0472) (0.0706) (0.0473) (0.0718) (0.0471) (0.0759) Maturity of Bond issued 0.0537** 0.0344*** 0.0542** 0.0347*** 0.0536** 0.0341*** 0.0548** 0.0326*** (0.0205) (0.00766) (0.0207) (0.00761) (0.0203) (0.00774) (0.0215) (0.00764) UST10Y 0.370* 0.850*** 0.378* 0.848*** 0.432** 0.822*** 0.416* 0.932*** (0.201) (0.175) (0.202) (0.174) (0.187) (0.163) (0.218) (0.190) Bond Tranche fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Country fixed effects Yes No Yes No Yes No Yes No Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Country-period fixed effects No Yes No Yes No Yes No Yes Includes China issuance Yes No Yes No Yes No Yes No Observations 5,881 3,153 5,881 3,153 5,881 3,153 5,593 2,863 R-squared 0.805 0.703 0.805 0.702 0.805 0.704 0.795 0.687 No. of Countries 63 70 63 70 63 70 63 70 No. of Industries 187 192 187 192 187 192 187 191 No. of Bonds 5437 2865 5437 2865 5437 2865 5176 2602 Robust standard errors clustered on the country-industry level in parentheses. *** p<0.01, ** p<0.05, * p<0.1 33 Table 6. Bond Maturity OLS Regression Results Dependent Variable: Maturity of Bond Tranche (1) (2) (3) (4) (5) (6) (7) (8) VIX -2.951** -3.076* (1.245) (1.749) RISK -3.301*** -3.201*** (0.983) (1.157) FED 1.186 2.822** (1.824) (1.385) LIBOR -1.364* -1.176 (0.738) (0.735) GDPPC -2.046** -1.988** -2.031** -2.213* (0.836) (0.847) (0.844) (1.121) GROWTH 0.0968 0.0996 0.0929 0.0974 (0.0652) (0.0657) (0.0658) (0.0763) EXT -0.0355* -0.0351* -0.0346* -0.0419* (0.0185) (0.0187) (0.0186) (0.0229) CA -0.0553 -0.0549 -0.0563 -0.0273 (0.0714) (0.0719) (0.0721) (0.0654) PCRED 0.0528* 0.0542* 0.0532 0.0609 (0.0316) (0.0320) (0.0324) (0.0387) Log of Size of Bond Issued 0.524** 0.717** 0.527** 0.713** 0.525** 0.722** 0.445* 0.573 (0.250) (0.320) (0.246) (0.317) (0.253) (0.311) (0.238) (0.352) UST10Y 0.00287 -0.361 0.0999 -0.339 0.160 -0.299 -0.0613 -0.425 (0.265) (0.533) (0.257) (0.501) (0.301) (0.519) (0.261) (0.537) Bond Tranche fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Country fixed effects Yes No Yes No Yes No Yes No Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Country-period fixed effects No Yes No Yes No Yes No Yes Includes China issuance Yes No Yes No Yes No Yes No Observations 6,749 3,684 6,749 3,684 6,749 3,684 6,406 3,347 R-squared 0.393 0.298 0.393 0.298 0.391 0.297 0.401 0.301 No. of Countries 64 71 64 71 64 71 64 71 No. of Industries 198 203 198 203 198 203 197 202 No. of Bonds 6144 3268 6144 3268 6144 3268 5840 2969 Robust standard errors clustered on the country-industry level in parentheses. *** p<0.01, ** p<0.05, * p<0.1 34 Table 7. The Risk-Taking Channel and the Exchange Rate Country-industry logit regressions Dependent variable: Country-industry monthly issuance above 2000-07 average (1=Yes, 0=No) (1) (2) USREER -4.474*** (0.840) XRATE -2.399*** (0.478) Controls As in Table 4 As in Table 4 Observations 79,464 76,664 No. of Countries 64 63 No. of Country-Industries 448 441 Pseudo R-squared 0.360 0.361 Robust standard errors clustered on the country-industry level in parentheses. *** p<0.01, ** p<0.05, * p<0.1 35 Appendix 1. Country Sample and Issuance Activity by EMDE Entities Pre-crisis Crisis Post-crisis (2000-2006) (2007-2010) (2011-2014) Total Total Total No. of Volume No. of Volume No. of Volume Country Bonds ($ mln) Bonds ($ mln) Bonds ($ mln) Albania 1 405 Angola 1 1,000 Argentina 70 40,105 36 6,807 25 6,414 Armenia 1 690 Azerbaijan 2 5 5 279 6 3,444 Bangladesh 1 297 Belarus 3 3 6 1,350 1 800 Belize 2 223 Bolivia 2 989 Bosnia and Herzegovina 1 110 Botswana 1 80 Brazil 331 103,668 190 89,006 260 175,420 Bulgaria 12 3,040 1 291 5 4,583 China 48 16,802 132 29,783 2765 374,001 Colombia 42 15,852 18 12,380 43 33,492 Congo, Democratic Republic of 1 478 Costa Rica 8 1,600 8 5,496 Cote d'Ivoire (Ivory Coast) 1 2,332 1 736 Dominican Republic 10 2,271 3 1,180 9 5,160 Ecuador 1 596 2 2,981 Egypt 7 3,664 4 3,439 7 5,261 El Salvador 12 3,540 2 1,244 4 2,558 Ethiopia 3 741 Fiji 1 149 1 250 Gabon 1 1,000 2 2,109 Georgia 4 982 5 1,645 Ghana 1 750 2 1,985 Grenada 1 99 Guatemala 5 1,205 1 85 9 3,770 Honduras 2 1,000 Hungary 44 22,675 25 15,872 22 21,385 India 46 10,277 53 22,235 151 53,640 Indonesia 44 14,380 37 22,487 60 43,058 Iran 2 993 Iraq 1 2,700 Jamaica 22 4,604 10 4,482 9 6,060 Jordan 8 407 1 742 2 2,250 Kazakhstan 67 15,634 33 19,558 21 14,681 Kenya 2 2,794 36 Pre-crisis Crisis Post-crisis (2000-2006) (2007-2010) (2011-2014) Total Total Total No. of Volume No. of Volume No. of Volume Country Bonds ($ mln) Bonds ($ mln) Bonds ($ mln) Laos 4 348 Lebanon 41 19,899 17 9,796 10 9,791 Macedonia 1 176 1 243 1 666 Malaysia 41 17,650 11 8,612 75 21,765 Marshall Islands 1 230 Mexico 117 78,882 103 60,935 163 131,257 Mongolia 3 249 7 3,404 Montenegro 1 253 3 744 Morocco 1 453 2 2,007 6 5,674 Mozambique 2 810 Namibia 1 491 Nigeria 2 522 14 5,971 Pakistan 3 1,900 1 750 2 3,000 Panama 15 6,855 7 2,365 13 5,242 Paraguay 6 2,398 Peru 15 6,168 29 14,393 61 22,056 Philippines 76 28,983 25 13,968 35 14,382 Romania 11 3,502 3 2,569 12 17,082 Rwanda 1 393 Senegal 1 196 2 988 Serbia 1 1,018 7 6,109 Seychelles 1 199 1 30 South Africa 28 11,632 25 13,204 54 23,902 Sri Lanka 1 100 3 2,000 10 5,775 Tanzania 1 600 Thailand 21 6,218 10 3,531 23 15,954 Togo 1 248 Tunisia 9 2,949 1 253 5 2,017 Turkey 78 45,091 30 24,270 128 58,555 Ukraine 28 8,993 27 10,729 32 21,746 Venezuela 23 16,464 10 31,785 4 12,444 Vietnam 1 737 3 1,173 4 1,532 Zambia 2 1,728 37 Appendix 2. Annualized External Bond Issuance Statistics Averages are not weighted. Statistics exclude issuance by Chinese entities. Panel A. Bond issuance characteristics Total Volume Number of Avg. Maturity Investment Year Avg. Yield (%) ($mln) bond tranches (years) grade (%) 2000 6.33E+10 189 9.5 7.2 21% 2001 5.73E+10 170 8.8 6.7 33% 2002 4.39E+10 125 9.0 8.5 36% 2003 5.76E+10 167 8.0 7.9 41% 2004 7.71E+10 200 7.4 8.3 36% 2005 1.11E+11 249 7.3 9.7 40% 2006 9.56E+10 294 8.0 9.3 32% 2007 9.94E+10 297 8.1 12.2 45% 2008 4.55E+10 103 7.4 8.3 46% 2009 1.07E+11 159 7.7 9.2 48% 2010 1.60E+11 300 7.1 9.4 53% 2011 1.56E+11 284 6.8 9.4 53% 2012 2.05E+11 400 5.3 8.9 70% 2013 2.24E+11 438 5.3 7.8 64% 2014 2.16E+11 407 5.1 8.7 69% Panel B. Global push factors around time of issuance Avg. LIBOR Avg. FED Avg. UST10Y Year Avg. VIX Avg. RISK (%) (bps) ($mln) (%) 2000 23.05 1.85 - 5.90E+05 6.20 2001 25.01 2.35 23.33 6.09E+05 5.24 2002 26.12 2.59 14.68 6.63E+05 4.80 2003 24.69 2.25 14.67 6.52E+05 3.93 2004 16.57 1.35 11.85 6.75E+05 4.31 2005 13.13 1.28 9.56 7.19E+05 4.23 2006 13.10 1.23 7.93 7.54E+05 4.73 2007 13.53 1.27 13.87 7.79E+05 4.71 2008 24.17 2.84 70.84 6.67E+05 3.89 2009 33.88 5.08 66.03 9.77E+05 3.30 2010 23.67 2.48 16.28 1.78E+06 3.35 2011 20.41 2.15 15.23 2.16E+06 3.05 2012 21.27 2.68 33.14 2.51E+06 1.88 2013 14.92 2.05 16.66 2.84E+06 2.04 2014 13.80 1.67 14.57 3.84E+06 2.67 38 Panel C. Domestic pull factors around time of issuance Avg. GDPPC Avg. Avg. PCRED Year Avg. CA (%) Avg. EXT (%) (real US$) GROWTH (%) (%) 2000 4706 3.64 -3.14 45.39 31.89 2001 4077 0.88 -3.03 52.08 34.40 2002 2980 3.50 -0.87 54.66 36.99 2003 3282 3.00 0.53 48.83 31.53 2004 3983 6.50 -0.01 51.60 36.24 2005 4662 5.36 -0.71 42.64 35.98 2006 4946 6.43 -0.88 47.95 39.65 2007 5799 6.63 -2.62 42.53 42.72 2008 7848 4.28 -1.85 41.81 46.30 2009 6401 0.00 -1.11 41.28 40.34 2010 7507 6.87 -1.62 33.97 41.86 2011 8293 4.64 -2.45 40.36 50.92 2012 7856 3.51 -3.15 38.16 54.65 2013 7611 3.66 -3.40 39.96 53.76 2014 7669 3.43 -3.22 40.77 53.34 Panel D. Fraction of country-industries with monthly issuance volume above historical average % of country- industries that Year issue above their historical average 2000 1.98% 2001 1.83% 2002 1.59% 2003 1.91% 2004 2.53% 2005 3.02% 2006 3.35% 2007 3.10% 2008 1.14% 2009 1.96% 2010 3.67% 2011 3.79% 2012 4.63% 2013 5.30% 2014 5.26% 39 Appendix 3. Interactions between Country Characteristics and Global Factors Table 1. Country-Industry Level Logit Regressions (Equation (1)) Dependent Variable: Issuance Country-industry monthly issuance above 2000-07 average (1=Yes, 0=No) Push Inter- Push Inter- Push Push Inter- Push Factor GDPPC action Factor GROWTH action Factor EXT Interaction Factor CAD action Factor PCRED Interaction VIX -1.710 -0.710*** -0.735*** -0.713*** -0.351* (1.269) (0.135) (0.154) (0.103) (0.184) Pull Factor 0.504 0.0126 0.00351 -0.0334 0.0197** (0.464) (0.0631) (0.00719) (0.0381) (0.00918) Interaction 0.119 -0.000720 0.000504 0.000276 -0.00724*** (0.150) (0.0200) (0.00231) (0.0132) (0.00270) Obs. #C #Id Obs. #C #Id Obs. #C #Id Obs. #C #Id Obs. #C #Id 70,644 61 427 70,644 61 427 70,644 61 427 70,644 61 427 70,644 61 427 RISK -0.535 -0.642*** -0.575*** -0.628*** -0.462*** (0.873) (0.120) (0.128) (0.0912) (0.154) Pull Factor 1.034*** -0.000531 0.00636* -0.0322*** 0.00104 (0.159) (0.0201) (0.00329) (0.0118) (0.00511) Interaction -0.0111 0.00310 -0.00126 -0.000528 -0.00343* (0.105) (0.0154) (0.00173) (0.00939) (0.00208) Obs. #C #Id Obs. #C #Id Obs. #C #Id Obs. #C #Id Obs. #C #Id 70,644 61 427 70,644 61 427 70,644 61 427 70,644 61 427 70,644 61 427 FED 2.540*** 0.353*** 0.894*** 0.569*** 0.178 (0.939) (0.101) (0.101) (0.0812) (0.142) Pull Factor 3.183** -0.704** 0.112*** -0.212 -0.111*** (1.376) (0.328) (0.0274) (0.182) (0.0368) Interaction -0.229** 0.0540** -0.00789*** 0.0135 0.00774*** (0.106) (0.0236) (0.00194) (0.0129) (0.00255) Obs. #C #Id Obs. #C #Id Obs. #C #Id Obs. #C #Id Obs. #C #Id 53,928 59 413 53,928 59 413 53,928 59 413 53,928 59 413 53,928 59 413 LIBOR -1.018** -0.473*** -0.452*** -0.462*** -0.408*** (0.442) (0.0631) (0.0684) (0.0495) (0.0871) Pull Factor 0.921*** -0.000968 0.00742* -0.0288 0.00225 (0.190) (0.0294) (0.00392) (0.0178) (0.00584) Interaction 0.0654 0.00266 -0.000242 8.13e-05 -0.00105 (0.0510) (0.00923) (0.00115) (0.00579) (0.00117) Obs. #C #Id Obs. #C #Id Obs. #C #Id Obs. #C #Id Obs. #C #Id 60,732 61 427 60,732 61 427 60,732 61 427 60,732 61 427 60,732 61 427 Robust standard errors clustered on the country-industry level in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Note: Obs. = Observations, #C = Number of Countries, #Id = Number of Industries 40 Table 2. Bond-Level Pricing OLS Regressions (Equation (2), excluding Chinese issuance) Dependent Variable: Fixed yield-to-maturity of Bond Tranche Push Inter- Push Inter- Push Inter- Push Inter- Push Factor GDPPC action Factor GROWTH action Factor EXT action Factor CAD action Factor PCRED Interaction VIX -0.907 0.654* 0.860** 0.946*** 1.390** (1.654) (0.334) (0.340) (0.328) (0.523) Pull Factor -0.423 -0.317** -0.00223 0.103 0.0153 (0.792) (0.137) (0.00719) (0.108) (0.0233) Interaction 0.224 0.0945** 0.00371 -0.0340 -0.00814 (0.188) (0.0412) (0.00232) (0.0362) (0.00805) Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd 2,738 58 173 2,502 2,738 58 173 2,502 2,738 58 173 2,502 2,738 58 173 2,502 2,738 58 173 2,502 RISK -0.179 0.893** 0.937** 1.076*** 1.277*** (2.163) (0.381) (0.390) (0.375) (0.446) Pull Factor 0.155 -0.0874** 0.00516 0.0264 -0.00645 (0.456) (0.0387) (0.00751) (0.0266) (0.00557) Interaction 0.154 0.0713** 0.00419** -0.0355 -0.00396 (0.251) (0.0277) (0.00204) (0.0264) (0.00655) Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd 2,738 58 173 2,502 2,738 58 173 2,502 2,738 58 173 2,502 2,738 58 173 2,502 2,738 58 173 2,502 FED -0.832 -0.640* -0.624* -0.694* -0.655* (1.844) (0.357) (0.362) (0.382) (0.389) Pull Factor 0.0478 0.457 0.0292 -0.0677 0.00858 (2.794) (0.449) (0.0401) (0.355) (0.0408) Interaction 0.0141 -0.0354 -0.00152 0.00531 -0.00125 (0.207) (0.0327) (0.00284) (0.0258) (0.00300) Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd 2,738 58 173 2,502 2,738 58 173 2,502 2,738 58 173 2,502 2,738 58 173 2,502 2,738 58 173 2,502 LIBOR 0.311 0.294* 0.286 0.438*** 0.496** (1.380) (0.170) (0.174) (0.149) (0.236) Pull Factor 0.753 -0.165** 0.00290 0.0571 -0.0110 (0.628) (0.0711) (0.00972) (0.0568) (0.0133) Interaction 0.0188 0.0423* 0.00396** -0.0188 -0.000521 (0.163) (0.0218) (0.00167) (0.0194) (0.00460) Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd 2,454 58 173 2,245 2,454 58 173 2,245 2,454 58 173 2,245 2,454 58 173 2,245 2,454 58 173 2,245 Robust standard errors clustered on the country-industry level in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Note: Obs. = Observations, #C = Number of Countries, #Id = Number of Industries, # Bd = Number of Bonds. R-squared for regressions with the VIX, RISK, and FED = 0.65; R-squared for regressions with LIBOR = 0.63. 41 Table 3. Bond Maturity OLS Regressions (Equation (2), excluding Chinese issuance) Dependent Variable: Maturity of Bond Tranche Push Factor GDPPC Interaction Push Factor GROWTH Interaction Push Factor EXT Interaction Push Factor CAD Interaction Push Factor PCRED Interaction VIX 2.399 -4.766** -4.108 -3.735* -2.003* (6.849) (2.198) (2.560) (1.869) (1.008) Pull Factor -0.0849 -0.416 -0.0407 -0.404 0.184 (2.825) (0.412) (0.0434) (0.404) (0.187) Interaction -0.747 0.175 0.00160 0.122 -0.0457 (0.780) (0.132) (0.0111) (0.122) (0.0541) Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd 3,247 60 184 2,887 3,247 60 184 2,887 3,247 60 184 2,887 3,247 60 184 2,887 3,247 60 184 2,887 RISK -0.122 -5.049*** -4.708** -4.486*** -3.325*** (3.943) (1.873) (1.928) (1.630) (1.089) Pull Factor -2.005 0.0198 -0.0366 -0.112 0.0761 (1.254) (0.117) (0.0229) (0.123) (0.0569) Interaction -0.531 0.120 0.00145 0.0990 -0.0352 (0.496) (0.115) (0.00741) (0.103) (0.0389) Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd 3,247 60 184 2,887 3,247 60 184 2,887 3,247 60 184 2,887 3,247 60 184 2,887 3,247 60 184 2,887 FED 5.666 3.432** 3.399** 3.649*** 4.318*** (3.871) (1.406) (1.471) (1.365) (1.530) Pull Factor 0.912 -0.656 -0.0704 -0.648 0.312 (5.961) (1.510) (0.118) (0.995) (0.228) Interaction -0.241 0.0552 0.00255 0.0431 -0.0190 (0.454) (0.109) (0.00797) (0.0693) (0.0151) Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd 3,247 60 184 2,887 3,247 60 184 2,887 3,247 60 184 2,887 3,247 60 184 2,887 3,247 60 184 2,887 LIBOR 0.154 -2.102* -1.496 -1.467 -0.675 (2.437) (1.107) (1.058) (0.883) (0.561) Pull Factor -2.772 -0.188 -0.0371 -0.220 0.120 (1.693) (0.183) (0.0300) (0.196) (0.0854) Interaction -0.206 0.105* -0.00277 0.0690 -0.0224 (0.293) (0.0616) (0.00381) (0.0535) (0.0196) Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd Obs. #C #Id #Bd 2,918 60 183 2,595 2,918 60 183 2,595 2,918 60 183 2,595 2,918 60 183 2,595 2,918 60 183 2,595 Robust standard errors clustered on the country-industry level in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Note: Obs. = Observations, #C = Number of Countries, #Id = Number of Industries, # Bd = Number of Bonds. R-squared for all regressions = 0.25; R-squared for regressions with LIBOR = 0.63 42