WPS8198 Policy Research Working Paper 8198 Which Emerging Markets and Developing Economies Face Corporate Balance Sheet Vulnerabilities? A Novel Monitoring Framework Erik Feyen Norbert Fiess Igor Zuccardi Huertas Lara Lambert Equitable Growth, Finance and Institutions Vice Presidency September 2017 Policy Research Working Paper 8198 Abstract This paper introduces a novel corporate financial vulnerabil- Middle East and North Africa, and Sub-Saharan Africa ity index that tracks financial conditions of the non-financial in recent years. The energy sector has exhibited the fastest corporate sector. Using the balance sheet information of deterioration, especially since 2014, in part driven by the 14,207 listed non-financial firms in 69 emerging markets decline in oil prices. However, if currently relatively benign and developing economies, the index shows that, at the global funding conditions and higher commodity prices global level, corporate vulnerability sharply increased since endure, companies may have an opportunity to strengthen 2013 and stabilized in 2016. Regional trends are more het- their balance sheets. The paper also finds that the index has erogeneous, pointing to significant corporate vulnerabilities leading indicator properties for socioeconomic outcomes, in Eastern Europe and Central Asia, as well as a deterio- such as a rise in unemployment and an economic recession, ration of firms’ financial conditions in Latin America, the and outperforms a commonly used “debt at risk” approach. This paper is a product of the Equitable Growth, Finance and Institutions Vice Presidency. 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 authors may be contacted at efeijen@worldbank.org, nfiess@worldbank.org, and izuccardi@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 Which Emerging Markets and Developing Economies Face Corporate Balance Sheet Vulnerabilities? A Novel Monitoring Framework Erik Feyen Norbert Fiess Igor Zuccardi Huertas Lara Lambert JEL Classification Numbers: F34, F65, G30. Keywords: Corporate Vulnerability, Non-Financial Sector, Debt Structure, Emerging and Developing Economies.  Erik Feyen is Lead Financial Sector Economist in the World Bank’s Finance & Markets Global Practice. Norbert Fiess is Lead Economist in the Macro and Fiscal Management Global Practice. Igor Zuccardi Huertas is Financial Sector Economist in the Finance & Markets Global Practice. Lara Lambert is Research Officer at the International Finance Corporation. The views expressed in this paper are those of the authors and do not necessarily reflect those of the World Bank or the International Financial Corporation. We are grateful to Carlos Arteta, Tatiana Didier, Carlos Felipe Jaramillo, Ayhan Kose, Bill Maloney, Yira Mascaro, Franziska Ohnsorge, Ceyla Pazarbasioglu, Claudio Raddatz, Sergio Schmukler, and the participants of the FSB-AGV Meeting, Basel April 2017 for comments. We also thank Diego Sourrouille and Yan Wang for research assistance. For questions, please contact the authors at efeijen@worldbank.org; nfiess@worldbank.org; izuccardi@worldbank.org; and LLambert@ifc.org. 1 Introduction Corporate debt in many emerging markets and developing economies (EMDEs) has risen significantly since the global financial crisis1 (IMF 2015), raising concerns about financial stability and spillover risks to the real sector. This paper introduces the Corporate Vulnerability Index (CVI), a novel country monitoring framework that tracks financial conditions of the non-financial corporate sector in EMDEs. Using readily available balance-sheet information of listed non- financial firms, the CVI is based on seven indicators2 which capture four key dimensions of firms’ financial vulnerabilities: debt service capacity, leverage, rollover risk, and economic performance. In recent years, a growing literature has attempted to quantify corporate financial vulnerabilities in EMDEs (Financial Stability Board, 2015; IMF, 2015, 2016a, 2017; IIF 2015 and 2017; World Bank, 2016; Beltran et al, 2016). While most of these studies assess vulnerabilities in terms of the interest coverage ratio (ICR) and leverage, some also consider corporate vulnerability along other dimensions, such as maturity mismatches (e.g. Gonzalez-Miranda, 2012; Rodrigues Bastos et al, 2015; Alfaro et al, 2017), and find that firms have changed their leverage and maturity structure between 2000 and 2013 to take advantage of benign global financial conditions. However, this literature focuses on a relatively small sample of firms in a select number of countries. Our paper is related and contributes to the literature in several respects. First, we propose a novel vulnerability index which extends the widely-used concept of “Debt at Risk” (see for example IMF, 2016a). Debt at risk is the total amount of debt in a country (or industry) associated with firms which are deemed financially vulnerable, typically for firms with an ICR below a threshold. Our contribution is to apply this concept across multiple financial vulnerability indicators since firms can be financially vulnerable across multiple dimensions at the same time. The CVI appears to have leading indicator qualities; an increase in the CVI tends to be associated with a future economic recession and an increase in unemployment. Our findings also suggest 1 See World Bank (2016), IMF (2015), IIF (2015), IS (2015) and Geneva Report (2015). 2 The seven indicators are: Interest Coverage Ratio (ICR), Leverage ratio, Net debt to EBIT ratio, Current liabilities to Long-term liabilities ratio, Quick ratio, Return on Assets (ROA), and Market to Book ratio. 2 the CVI is more informative for future socio-economic outcomes compared to the commonly used Debt at Risk approach based on ICR (IMF, 2017; FSB, 2015). The CVI is related to a large literature on corporate default modeling based on accounting data (e.g. Altman, 1968) and distance-to-default or contingent claim models based on market prices inspired by Merton (1974). Unlike the CVI, these models however require default and price data which are not readily available for a wide range of countries and/or firms. Another important advantage of our balance-sheet-based approach is that it can easily be extended to cover non- listed firms. We also add to the “early warning” literature which suggests corporate debt overhangs can be a leading indicator of crises and growth slowdowns as well as having the effect of amplifying shocks (for Europe: Goretti and Souto, 2013; Aiyar et al., 2015; Bergthaler et al. 2015; for emerging markets: IMF, 2015; Lindner and Jung, 2014). Finally, unlike many previous studies that are based on small samples or individual countries, our paper is based on a large sample of 14,207 listed firms, spanning 11 years (2006 to 2016) and 69 EMDEs; this should make our findings more representative. At the global level, the CVI suggests that vulnerabilities have risen sharply since 2013, but have stabilized in 2016. This global trend is caused by an increase in leverage and a deterioration of both profitability and debt service capacity. However, regional trends are heterogeneous as corporate vulnerabilities in Eastern Europe and Central Asia (ECA) have been elevated since 2007, while Latin America (LAC) shows a rapid increase in vulnerability since 2013. In 2016, most regions have stabilized except the Middle East and North Africa (MENA) and Sub-Saharan Africa (SSA), whose corporate sector’s vulnerabilities have risen. A key finding is that energy-linked sectors have shown an increase in financial vulnerabilities, especially since 2014, in conjunction with the decline in oil prices. This paper is organized as follows. Section 2 describes recent trends in non-financial corporate sector debt in EMDEs and discusses the methodology to construct the CVI. Section 3 describes the data used for the construction of the index. Section 4 presents the main results. Section 5 describes robustness tests and extensions. Section 6 concludes. 3 2 Methodology 2.1 Trends in Non-Financial Corporate Debt in EMDEs Non-financial corporate sector debt in many EMDEs has risen significantly since the global financial crisis3 (Figure 1.1). Moreover, given unprecedented accommodative monetary policies in the developed world, EMDE corporates have raised significant volumes of financing from global capital markets, mostly denominated in foreign currencies and typically not fully (naturally) hedged. This has had the effect of changing the composition of corporate debt in EMDEs away from bank credit and towards debt securities, with bond issuance from EMDE corporates more than doubling since 2010 (Figure 1.2).4 With the sharp decline in commodity prices since 2014 and the downward revisions to growth prospects across EMDEs, both firms’ profitability and debt service capacity has trended down in those countries. Consequently, financial risks from the EMDE non-financial corporate sector have emerged, which have been reflected by the widening of corporate Credit Default Swaps (CDS) spreads across regions, the rise of non-performing loans in the EMDE banking sector, and the significant declining of corporate bond issuance since 2015. 3 See World Bank (2016), IMF (2015), IIF (2015), BIS (2015) and Geneva Report (2015). 4 Becker and Ivanshina (2014), Cortina, Didier and Schmukler (2016). For details on the evolution and financial threats of corporate indebtedness, see Acharya et al (2015). 4 Figure 1: EMDE Corporate Debt Evolution 1. EMDE Corporate Debt Composition 2. EMDE Bond Composition Loans vs Bonds (USD billions) Local vs Foreign (USD billions) Source: IMF GFSR (October, 2015) Source: IMF GFSR (October, 2015) In the current global environment, questions have been raised about the financial stability risks of the corporate sector in EMDEs and the potential spillover effects on their financial sectors as corporations may face tightening in the global financial conditions and/or lower and more volatile commodity prices. 2.2 The Corporate Vulnerability Index The CVI is a composite indicator that assesses non-financial corporate vulnerability in emerging and developing economies (EMDEs). Based on corporates’ balance-sheet information, the CVI measures four key aspects5 of financial vulnerability that have been identified by the literature as leading indicators of corporate financial distress: Debt Service Capacity, Leverage, Rollover Risk, and Profitability/Market value. As shown in Figure 2, these four aspects of corporate vulnerability are measured using seven indicators for which data are readily and sufficiently available across a broad range of EMDEs: Interest Coverage Ratio (ICR), Leverage Ratio, Net Debt to EBIT Ratio, Current Liabilities to Long-term Liabilities Ratio, Quick Ratio, Return on Assets (ROA), and Market to Book Ratio. This set represents a diverse mix of indicators which are based 5 We recognize that other variables such as currency risks are important as well. However, as we explain in the Data section, data limitations prevent us from including them in our framework. 5 on both flow and stock data; it makes our approach more robust compared to, for example, the common “ICR-only” approach which could just flag a transient issue since the ICR is solely based on flow data. Indeed, although some of our indicators are conceptually correlated through basic accounting identities, empirically the pairwise correlations between the indicators are statistically significant, but relatively low across the four broad factors (Table 1) suggesting they capture different corporate vulnerability aspects and collectively can produce a more reliable result. Figure 2: Structure of Corporate Vulnerability Index The CVI is based on the concept of “Debt at Risk” (DaR), the total amount of outstanding debt in a country (or industry) associated with firms that are deemed financially vulnerable. DaR is an attractive concept to track corporate vulnerabilities since it exposes both the risk and magnitude present in the tail of the firm’s distribution, as opposed to other methodological approaches such as calculating averages or medians of (normalized) firm vulnerability indicators. Specifically, we define as the share of corporate debt in a country that is considered vulnerable according to indicator Y at time t and country c: 6 , ) = , (6) , 6 The CVI can be calculated at country-industry level. In that case, all calculations described in this section are conducted using information of corporate debt at country-industry level. Given that thresholds are calculated by industry, they are not modified when the VI at country-industry level is calculated. 6 where Y denotes one of our seven indicators. For each of the indicators, firms are classified as financially vulnerable if an indicator breaches an indicator-specific threshold at time t (Table 2). Table 1: Firm-Level Correlations Net Curr/Long ICR Leverage Quick ROA Debt Liab ICR 1 Leverage -0.1777* 1 Net Debt -0.0286* 0.0035 1 Curr/Long Liabilities 0.0115* -0.0545* -0.0199* 1 Quick 0.1307* -0.2482* -0.0394* -0.0175* 1 ROA 0.2078* -0.2897* 0.0306* -0.0226* 0.0951* 1 Market to Book 0.0566* 0.0035 -0.0285* 0.0374* -0.0006 0.0671* Note: Financial Indicators of 14,207 listed non-financial firms in 69 countries. Annual Information from 2006 to 2016. Variables winsorized at 1%. *Significance level 5% Table 2: Thresholds to classify a firm as financially vulnerable Indicator “At risk” Thresholds * Interest Coverage Ratio < 1 (profits less than interest expenses) * Leverage Ratio > 90th percentile value of the indicator for all * Net Debt to EBIT firms within the same industry, for the whole * Current liabilities to Long-term liabilities sample 2006-2016. One threshold per industry * Quick Ratio < 10th percentile value of the indicator for all * Return on Assets firms within the same industry, for the whole * Market to Book Ratio sample 2006-2016. One threshold per industry Note: Our sample includes financial information from 14.273 listed non-financial firms in 96 Emerging and Developing Economies, for years 2006 to 2016. A representativeness restriction is imposed in which countries with at least 5 firms in the sample are considered in the calculations. Therefore, the adjusted sample includes 14.207 firms from 69 countries. We use 1 as a threshold for ICR since firms with profits less than interest expenses are immediately highly vulnerable. This threshold is more conservative than those in other studies, but we find that a value of 1 provides consistent results. For Leverage Ratio, Net Debt to EBIT Ratio, and Current to Long-Term Liabilities, the vulnerability thresholds correspond to the 90th percentile value of the respective indicators for all firms within the same industry and across countries. By pooling by industry and across time (2006-2016) and countries, we focus on industry-specific effects and abstract from time and country effects. Similarly, for Quick Ratio, 7 Return on Assets, and Market to Book Ratio, the respective thresholds are equal to the 10th percentile value of the indicator by industry. We extend the notion of to multiple indicators which allows us to measure the “intensity” of debt at risk. We do so by focusing on debt of firms for which multiple indicators signal financial vulnerability at the same time. The underlying assumption is that debt that is associated with firms that are contemporaneously vulnerable according to multiple indicators is more risky. We operationalize this notion by defining . The captures the proportion of total corporate debt in a country that is held by firms that are vulnerable according to X or more indicators at the same time, where ∈ [0,7]: , ) = , . (7) is designed to exhibit a stronger signal-to-noise ratio compared to . The CVI is calculated as the average of for country c and time t: = ∑ ) , (8) where 0 ≤ ≤ 1. The definition has an intuitive graphical interpretation. As illustrated in Figure 3, the CVI is equivalent to the normalized area under the curve (the area under blue line). At the extremes, if no firm is financially vulnerable according to any indicator, then the value of the area under the curve (and the CVI) is equal to zero (i.e. equivalent to the area under the green line). In contrast, if all firms are financially vulnerable with respect to all seven indicators, then the area under the curve (and the value of CVI) would be equal to one (i.e. equivalent to the area under the red line). In practice and as expected, the CVI has values well below one; the sample mean is 0.11 with a maximum of 0.51. 8 Figure 3: Corporate Vulnerability Index and Area Under the Curve Our framework is underpinned by simplifying assumptions. The concept does not differentiate between indicators and treats them as interchangeable and of equal weight. While the credit scoring literature referenced earlier focuses on assessing indicator weights, these models are typically applied to a specific country or industry and require corporate default data. In the absence of such data for a large sample of countries, and to ensure wide applicability and transparency of the CVI, our approach is more pragmatic. Note also that the CVI de facto applies a greater weight on the debt of more vulnerable firms since the area under the curve is cumulative (i.e. is always 100 percent). In other words, the debt of a firm which is vulnerable according to seven indicators weighs seven times more than the debt of a firm that is vulnerable according to only one indicator. 3 Data We use balance-sheet information of 14,273 listed non-financial companies from 8 industries and 96 EMDE countries between 2006 and 2016 for the construction of the CVI. Bloomberg is the 9 source (see Figures 4.1 and 4.2).7 We exclude countries with fewer than five firms, which reduces our sample to 14,207 firms in 69 countries.8 It is important to highlight a few data limitations. First, our sample only covers listed non-financial firms. In most EMDEs listed firms are usually the biggest companies (by assets) and/or have better access to funding sources. This may bias our results, but we believe that information from listed companies, which generally also follow better accounting practices, nevertheless provides a good proxy of the general health of the overall non-financial corporate sector, in particular with regards to potential banking distress.9 Second, we do not have information about corporates’ financial information by currency and are therefore not able to accurately assess currency risks and external vulnerabilities. This is a general concern, as neither Bloomberg nor any other data source consistently collects this type of information on a cross-country basis. Finally, our data set does not contain information on derivatives, other risk management tools and counterparties which could be helpful to better understand exposures and transmission channels of corporate risk. 7 Coverage across regions and industries is heterogeneous: 38% of companies in the dataset are from EAP region, 26% from ECA, 17% from SA, 7.4% from MENA, 6.7% from LAC, 4.3% from SSA. In addition, 31.3% of the sample are firms from Industrials, 22.5% from Consumer Goods, 13.5% from Basic Materials, 11.6% from Consumer Services, 7.2% from Energy, 5.6% from Technology, 4.7% from Health Care, and 1.2% from Telecommunications. 8 The representativeness restriction of at least 5 firms in the database was not considered for calculating the VI at the country-industry level. 9 For instance, Gabaix (2011) highlights the importance of the firm’s size to explain shocks to the aggregate output. This author states that idiosyncratic shocks to large firms can lead to nontrivial aggregate shocks as modern economies are dominated by large corporations. 10 Figure 4: Data Coverage and Representativeness 1. Number of EMDEs’ companies per region 2. Number of EMDEs’ companies per industry No. Firms No. Firms 18000 18000 16000 16000 14000 14000 12000 12000 10000 10000 8000 8000 6000 6000 4000 4000 2000 2000 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 0 Basic Mat Cons Goods Cons Serv Energy 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Health Care Industrials Technology Telecom EAP ECA LAC MENA SA SSA Source: Bloomberg Source: Bloomberg 3. Reported total corporate debt 4. Common shares of firms covered (% GDP) (% Stock market capitalization) Total Shares/Total Debt/GDP St. Mkt (%) Capital (%) 35 90 80 30 70 25 60 20 50 15 40 10 30 20 5 10 0 0 EAP ECA LAC MENA SA SSA EAP ECA LAC MENA SA SSA Median per region, 75th and 25th percentiles. Median per region, 75th and 25th percentiles. GDP available as of 2015. Stock market capitalization available as of 2015. Source: Bloomberg, WB-World Development Indicators Source: Bloomberg, WB-Finstats Our sample is not homogenous across regions (see Figure 4.3). The regional median of total corporate debt (% of GDP) ranges from 2.9 percent of GDP for SSA to 10.8 percent of GDP for EAP. The companies covered in our data set are nevertheless an important part of EMDEs’ capital markets as the median value of their common shares (i.e. % of stock market capitalization) ranges from 17.6 percent in SSA to 61 percent in ECA (see Figure 4.4). 11 The financial indicators for listed non-financial firms in our sample show downward trends in debt service capacity, profitability, and market valuation across all regions (Figures 5.1, 5.2, and 5.3). In addition, leverage has increased in most regions except South Asia (SA) and EAP, where leverage has been high, but decreasing. LAC and SSA have experienced a steep increase in leverage since 2011/2012 (Figure 5.4). Based on an increasing Quick Ratio, rollover risk has declined since the Global Financial Crisis, but ECA, SA, and SSA may be vulnerable to adverse liquidity shocks as the quick ratio has been persistently below one (Figure 5.5). Figure 5: Financial Indicators, 2006-2016 Country medians per region 1. Interest Coverage Ratio 2. Return on Assets ICR ROA % 7.0 10.0 6.0 8.0 5.0 4.0 6.0 3.0 4.0 2.0 1.0 2.0 0.0 0.0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 EAP ECA LAC MENA SA SSA EAP ECA LAC MENA SA SSA Note: Interest Coverage Ratio (ICR): Earnings Before Interest and Taxes/ Source: Bloomberg, own calculations Interest Expenses. Source: Bloomberg, own calculations 3. Market to Book Ratio 4. Leverage Ratio Market to Leverage % Book 35.0 2.6 30.0 2.1 25.0 20.0 1.6 15.0 1.1 10.0 0.6 5.0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 EAP ECA LAC EAP ECA LAC MENA SA SSA MENA SA SSA Source: Bloomberg, own calculations Note: Leverage Ratio: Total Debt/Total Assets. Source: Bloomberg, own calculations 12 5. Quick Ratio Quick Ratio 1.3 1.2 1.1 1.0 0.9 0.8 0.7 0.6 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 EAP ECA LAC MENA SA SSA Note: Quick Ratio: (Current Assets-Inventories)/Current liabilities. Source: Bloomberg, own calculations 4 Results 4.1 Trends in the Corporate Vulnerability Index Debt of non-financial listed firms in EMDEs has increased in both level and riskiness (Figure 6): EMDEs’ corporate debt increased by 46 percent, from $2.6 trillion to $3.8 trillion between 2010 and 2016. Over the same period, debt in the hands of firms considered vulnerable based on at least one indicator grew by 120 percent, from US $1 trillion to US $2.2 trillion. In 2016, about 58 percent of debt was considered ‘at risk’ based on at least one indicator; 35 percent according to two or more indicators; 12 percent based on three or more indicators; and 4 percent according to four or more indicators. 13 Figure 6: Corporate Debt at Risk (US$ Billion) US$ billion 4000 100% 3500 3000 2500 58% 2000 1500 35% 1000 500 12% 4% 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Total Debt DaR 1 indicator DaR 2 indicators DaR 3 indicators DaR 4 indicators DaR 5 indicators DaR 6 indicators DaR 7 indicators Note: DaR refers to ‘Debt at Risk’. Source: Bloomberg, own calculations At the global level, corporate vulnerability has increased sharply since 2013 according to our CVI10, although the speed of deterioration has moderated, especially in 2016 (see Figure 7). The global increase in corporate vulnerability has been driven by a rise in leverage ratios, and deteriorations in profitability and debt service capacity. Corporate vulnerability has also deepened: the country median value of DaR>=1 increased by 10.3 percentage points from 2013 to 2016, while DaR>=2, DaR>=3, and DaR>=4 rose by 8.6, 2.7, and 0.14 percentage points, respectively (Figure 8). 10 See Appendix for a CVI ranking at the country-level. 14 Figure 7: Global Corporate Vulnerability Index CVI 0.2 0.18 0.16 0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Global country median. 75th and 25th percentiles in dotted lines. Source: Own calculation Figure 8: Intensity of Corporate Vulnerability Debt at Risk for X or more indicators (DaR>=X) DaR>=X 60 50 40 30 20 10 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 DaR 1 or more DaR 2 or more DaR 3 or more DaR 4 or more Global country median. Source: Own calculations Trends across regions have not been uniform. In ECA, corporate vulnerability increased significantly in 2007 and has remained elevated since, while LAC experienced a steep rise in 15 vulnerability since 2013. Corporate vulnerability in EMDEs was stable in 2016, but MENA and SSA experienced some deterioration (Figure 9). The rise in corporate vulnerability in ECA has been associated with low profitability, deteriorating debt service capacity, high leverage, and increasing rollover risk. In LAC and MENA, increasing leverage ratios and deteriorating ICRs were the main drivers of vulnerabilities since 2013/2014. In SSA, low profitability drove the rise in corporate vulnerability between 2014 and 2016. Figure 10 shows the intensity of corporate vulnerabilities by region. The country median value of DaR>=1 increased for all regions between 2007 and 2016. ECA, SA, and SSA also experienced a rapid increase in the country medians of DaR>=2 and DaR>=3, particularly between 2013 and 2016. In SA, ECA, and EAP, the country median of DaR>=4 has increased, reaching values of 5.7%, 2.3%, and 1.7%, respectively. Figure 9: Regional Trends in Corporate Vulnerability VI 0.20 0.18 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 EAP ECA LAC MENA SA SSA Country median by region. Source: Own calculations At the country level, nine ECA countries are in the top 25 most vulnerable countries: companies in those countries are characterized by high levels of DaR by indicators Leverage Ratio, Return on Assets, and Quick Ratio. In addition, six MENA countries are in the top 25 most vulnerable countries: firms in those countries are highly vulnerable due to their important levels of DaR 16 associated with Leverage (i.e. high DaR by both Leverage and Net Debt to EBIT ratios) and debt service capacity (i.e. DaR based on the ICR).11 The CVI has increased for several countries between 2009 and 2016: the median rise of the CVI is 2.9 percentage points in that period. The non-financial corporate sectors of some regionally important EMDEs such as Brazil, India, the Russian Federation, Nigeria, China, and Indonesia have also shown weaker conditions, with a median CVI growth of 6 percentage points in the period 2009-2016. Figure 10: Intensity of corporate vulnerability by region Debt at Risk for X or more indicators (DaR>=X) DaR>=X 70 60 50 40 30 20 10 0 DaR>=1 DaR>=2 DaR>=3 DaR>=4 DaR>=5 DaR>=1 DaR>=2 DaR>=3 DaR>=4 DaR>=5 DaR>=1 DaR>=2 DaR>=3 DaR>=4 DaR>=5 DaR>=1 DaR>=2 DaR>=3 DaR>=4 DaR>=5 DaR>=1 DaR>=2 DaR>=3 DaR>=4 DaR>=5 DaR>=1 DaR>=2 DaR>=3 DaR>=4 DaR>=5 EAP ECA LAC MENA SA SSA 2007 2013 2016 Regional country medians. Source: Own calculations 11 An analysis by industry shows that corporate financial vulnerability in MENA countries is highly concentrated in the Energy sector. Debt at Risk by both Leverage and Net Debt to EBIT are above 95% of the total sectorial debt in 3 MENA countries, while Debt at Risk by ICR is above 60% of the total sectorial debt also in 3 MENA countries. 17 Figure 11: Corporate Vulnerability Index 2009 vs 2016 VI 2016 0.6 VEN 0.5 0.4 BIH TZA MNE 0.3 MKD TUN UKR KAZ LVA BRA ARE KEN MUS BGR HRV 0.2 EGY SAU IND SVK PAN MAR ZWE OMN CHN GHA HUN ROM RUS KWT ISR BOL GEOSRB BGD SVN ZMB COL ARG POL PAK NGA IDN VNM 0.1 CHL CIV THA ECU TUR PER MYS LTU MEX QAT LKA JOR CRI WBG PHL ZAF BHR MWI CZE JAM PRY 0 BWA TTO EST 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 VI 2009 Note: Scatter plot of the Vulnerability Index of 2009 versus 2016 showing that any country at the north of the 45-degree line has seen their corporate vulnerability position deteriorated during this period. Source: Own calculations At the industry level, corporate vulnerability has evolved unevenly over time: companies in basic materials, consumer goods, consumer services, and industrials increased their vulnerabilities during the financial crisis and this has continued since then. Energy, health care, and telecommunications firms have faced sharp deterioration of their financial conditions since 2011- 2012, a period characterized by accommodative financial conditions and increasing corporate leverage (Figure 12). Since 2014, with the end of the commodity super-cycle, financial conditions for energy companies worsened and their CVI has been high. 18 Figure 12: Corporate Vulnerability Index by Industry 1. CVI Industry trends I 2. CVI Industry trends II VI VI 0.16 0.16 0.14 0.14 0.12 0.12 0.1 0.1 0.08 0.08 0.06 0.06 0.04 0.04 0.02 0.02 0 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Basic Materials Consumer Goods Health Care Industrials Consumer Services Energy Technology Telecommunications Source: Authors’ calculations Source: Authors’ calculations Country median per industry Country median per industry 4.2 Association between the Corporate Vulnerability Index and Socio-Economic Outcomes Financial vulnerability in the non-financial corporate sector can have significant macroeconomic consequences as corporate distress may adversely impact the financial sector (e.g. through increases in non-performing loans, larger volatility of asset prices, rise of borrower risk); the public sector (e.g. through lower tax revenues, potential public support in case of public non- financial firms); other firms (e.g. through financial and/or trade interlinkages); and households (e.g. through the labor market).12 This section investigates the leading indicator properties of the CVI with regards to socio- economic outcomes. We use weighted logit regression models to establish whether an increase in unemployment; a GDP recession; or a reduction in welfare is associated with an increase in the Corporate Vulnerability Index CVIt, subject to controls (Xj): Pr = 1 | , , ) The respective dependent variables of the three logit regressions are three dummy variables that respectively take the value of 1 if (1) the unemployment rate increases from one year to the next; 12 See Gray et al (2006), and Ruscher and Wolff (2012). 19 (2) the annual GDP growth rate is negative (i.e., GDP recession); or (3) the annual per capita GDP is negative (i.e., Welfare reduction), and zero otherwise. The independent variables are the CVI and a set of control variables. We also include year dummies and weigh each observation by the ratio of the total sum of common shares of the firms in our sample to stock market capitalization in each country, which we use as a proxy for the representativeness of the CVI in a country. The CVI and control variables enter the regression lagged by one period to ameliorate endogeneity concerns. The macro-economic control variables are the current account balance, general government balance, government debt, real GDP per capita, and inflation rate. We further control for the commonly used DaR based on the ICR to test whether the CVI contains additional information. Table 2A: Descriptive Statistics Variable Obs Mean Std. p25 p75 Min Max Prob (Unemployment rise) (%) 300 21.7 18.0 0.0 36.4 0.0 54.5 Prob (Recession) (%) 367 7.5 9.6 0.0 9.1 0.0 36.4 Prob (Welfare reduction) (%) 367 14.4 13.6 0.0 18.2 0.0 54.5 Corp. Vulnerability Index (0 to 1 index) 367 0.11 0.07 0.06 0.16 0.00 0.36 DaR ICR (% Reported Total Corporate Debt) 367 19.0 16.9 6.4 26.0 0.0 89.3 Unemployment Rate (%) 300 9.0 6.1 5.5 10.1 0.7 34.9 GDP growth (yoy, %) 367 4.0 3.6 2.2 6.1 -7.8 18.0 GDP per capita growth (yoy, %) 367 2.5 3.7 0.5 4.9 -15.1 13.6 GDP per capita level (US$ constant 2010) 367 10051 11000 2981 12223 454 7468 Current Account Balance (% GDP) 353 -1.5 7.1 -5.4 2.5 -25.5 30.4 General Government Balance (% GDP) 362 -2.4 4.8 -4.8 -0.6 -16.5 22.2 General Government Debt (% GDP) 367 42.0 22.8 26.5 56.0 1.6 142.0 Inflation Rate (yoy, %) 350 5.7 4.3 2.7 7.9 -1.0 26.2 Note: Prob (X) is the unconditional probability that event X happens in a particular country between years 2006 and 2016. Unemployment rise: Dummy variable 1 if unemployment rate increases, 0 otherwise Recession: Dummy variable 1 if GDP growth is negative, 0 otherwise Welfare reduction: Dummy variable 1 if GDP per capita growth is negative, 0 otherwise Sources: WB World Development Indicators, IMF International Financial Statistics, Macro-Financial Initiative 20 Table 2B: Pairwise Correlations Corporate Current General General Unemploy Welfare Recession Vulnerabil DaR ICR Account Governme Governme ment Rise Reduction ity Index Balance nt Balance nt Debt Unemployment Rise 1 Recession 0.3487* 1 Welfare Reduction 0.3912* 0.6074* 1 Corporate Vulnerability Index 0.1244* 0.1059* 0.1315* 1 DaR based on ICR 0.0474 0.0246 0.0729 0.7121* 1 Current Account Balance 0.0381 0.0408 0.0411 -0.0258 -0.1204* 1 General Government Balance -0.2109* -0.1438* -0.0826 -0.1069* -0.2255* 0.4284* 1 General Government Debt 0.0738 0.0307 0.0068 -0.016 0.1016 -0.2174* -0.5510* 1 Inflation Rate -0.0098 -0.0647 0.0139 -0.0291 -0.1518* -0.2369* -0.0224 -0.0014 Unemployment rise: Dummy variable 1 if unemployment rate increases, 0 otherwise Recession: Dummy variable 1 if GDP growth is negative, 0 otherwise Welfare reduction: Dummy variable 1 if GDP per capita growth is negative, 0 otherwise *Significance level 5% 21 Table 3 presents our logit regressions. Panel A shows that the CVI is positively and statistically significantly associated with future socio-economic outcomes. Panel B indicates that, once the DaR based on ICR is included, the estimated coefficient of the CVI is still positive and statistically significant in most cases. Importantly, the DaR based on ICR is not statistically significant, suggesting that CVI has stronger predictive power for socio-economic outcomes. The CVI is correctly signed in all model specifications and, with the exception of the welfare specification, remains significant even when including macroeconomic controls.13 The first three columns of Panel A directly test the predictive power of the CVI. Panel A Column 1 shows that, based on the results of the marginal effects on the median, a one-unit increase of CVI from the median is associated with a rise in the probability of unemployment by 1.78 percentage points one year later.14 In other words, if CVI goes from its median value of 0.15 to 0.16, the probability of unemployment rise will increase 0.0178 percentage points. Column 1 also shows the area under the Receiver Operating Characteristic (ROC) curve (AUC) of 0.73, which means that the CVI has reasonable predictive accuracy for a future rise in unemployment.15 The results of Panel A Column 2 show that a one-unit increase of CVI from the median is associated with an increase in the probability of a recession next year by 0.91 percentage points. The AUC is 0.86 suggesting that the CVI has strong predictive accuracy power. Finally, Panel A 13 Information for Macedonia is not included in the logit estimations for unemployment rise as the dynamics of unemployment in that country in recent years are explained by policies targeting vocational training and on-the-job training under different government stimulus programs. As explained by the IMF’s Article IV (2016b), Macedonia has one of the highest unemployment rates in emerging Europe (25.4% in 2016), mostly reflecting skills shortage and mismatch resulting from emigration of skilled workers and low level of education. Since 2008, overall unemployment rate has declined by 10 percentage points, with lower unemployment rates in groups like workers with vocational training and tertiary education. 14 In logit models, as the sign and significance of the estimated coefficients are important to determine the statistical relationship between the probability of the outcome (in this case unemployment rise) and the independent variables, the values of the coefficients have a difficult interpretation in economic terms. The coefficients of the marginal effects have an economic interpretation as they measure the change in predicted probability of outcome when the independent variables change in one unit. 15 The area under the Receiver Operating Characteristic (ROC) curve (AUC) is a measure that that reflects how well a model is able to correctly classify a binary outcome variable. The ROC compares the proportion of cases in which the model specification correctly predicts the result of the outcome variable (i.e. true positive rate) versus the proportion of cases in which the model specification incorrectly predicts the result of the outcome variable (i.e. false positive rate). The AUC goes from 0 to 1, where the value 0 represents that the true positive rate is 0 percent and the false positive rate is 100 percent, and a value of 1 represents that the true positive rate is 100 percent and the false positive rate is 0 percent. The closer the AUC is to 1, the better the model. The AUC of a random classifier is 0.5. 22 Column 3 establishes that the probability of welfare reduction (i.e., fall in GDP per capita) goes up by 0.78 percentage points when the lagged CVI rises one unit from the median. The AUC is 0.81. The last three columns of Panel A control for macroeconomic variables. The results indicate that the coefficients of CVI are still positive and statistically significant. In other words, our CVI captures relevant features related to the corporate sector which are distinct from general macroeconomic conditions. As shown in Panel A column 4, the probability of a future rise in unemployment increases by 1.36 percentage points when the CVI rises one unit from its median value. Similarly, Panel A column 5 shows that the probability of a recession rises 1.02 percentage points when CVI goes up one unit from its median. The AUC increase to a very high 0.9 when macroeconomic variables are included. Finally, Panel A column 6 shows that the probability of a welfare reduction rises 0.66 percentage points with a change in CVI in one unit from the median. Interestingly, the AUC is only marginally affected by the inclusion of macroeconomic controls, suggesting that the model’s predictive classification power is mostly derived from the CVI. Panel B shows that the coefficients of CVI are still positive and statistical significant after including the commonly used DaR based on ICR which is insignificant in all specifications. This finding suggests that the CVI contains empirically relevant additional information compared to the DaR based on ICR. Note that the correlation between DaR based on ICR and the CVI is relatively high (0.71). However, the size of the CVI coefficient and statistical significance do not change dramatically (with the exception of Column 6) due to the inclusion of the DaR based on ICR, suggesting collinearity is not an overriding concern. 5 Sensitivity Analysis to Different ICR Thresholds Our results are consistent under different thresholds for ICR. Using ICR thresholds 1.5, 2, and 3, we recalculated the CVI to evaluate how sensitive our results are to thresholds commonly used in the literature (FSB, 2015; IMF, 2017). 23 As shown in the Appendix, Table B1, the country ranking does not have much variability under different ICR thresholds, particularly for both the top 10 and the bottom 15 countries. In addition, Tables B2, B3, and B4 show the logit regressions for unemployment rise, recession, and welfare reduction, respectively, using the CVI modified by different levels of ICR thresholds. The results show that the estimated coefficient for the CVI is positive and statistically significant in all cases. 24 Table 3: Corporate Vulnerability Index and Socio-economic Outcomes Logit Regressions, 2006-2016 Panel A: Corporate Vulnerability Index Variables Unemployment risea Recession Welfare reduction Unemployment risea Recession Welfare reduction L. Vulnerability Index (CVI) 8.880*** 9.825*** 6.509*** 6.835** 11.06*** 5.832** (2.220) (3.229) (1.915) (2.673) (3.816) (2.307) Constant -3.661*** -3.111*** -3.738*** -3.036*** -4.770*** -4.940*** (1.004) (0.899) (0.639) (0.979) (1.320) (1.059) Macro Controls^ No No No Yes Yes Yes Observations 294 281 367 287 232 340 ROC (Area under the curve) 0.734 0.857 0.810 0.717 0.896 0.841 Marginal Effects at Median 1.776*** 0.912*** 0.780*** 1.359** 1.016*** 0.657*** Panel B: Including DaR based on ICR Variables Unemployment risea Recession Welfare reduction Unemployment risea Recession Welfare reduction L. Vulnerability Index (CVI) 12.97*** 11.68** 9.496*** 9.045*** 10.84** 5.214 (4.124) (4.922) (3.673) (3.448) (5.066) (4.147) L.DaR ICR -2.430 -0.922 -1.483 -1.352 0.107 0.308 (1.914) (1.581) (1.398) (1.529) (2.035) (1.731) Constant -3.442*** -3.133*** -3.720*** -2.962*** -4.772*** -4.946*** (0.897) (0.896) (0.607) (0.978) (1.325) (1.056) Macro Controls^ No No No Yes Yes Yes Observations 294 281 367 287 232 340 ROC (Area under the curve) 0.745 0.859 0.815 0.720 0.895 0.841 Marginal Effects at Median 2.541*** 1.072*** 1.125*** 1.790*** 0.998** 0.590 Unemployment rise: Dummy variable 1 if unemployment rate increases, 0 otherwise. Recession: Dummy variable 1 if GDP growth is negative, 0 otherwise. Welfare reduction: Dummy variable 1 if GDP per capita growth is negative, 0 otherwise. ROC denotes Receiver Operating Characteristic, a common indicator that captures the ability of the specified model to replicate the results of the outcome variable. Unit of observation: country-year. ^ Macro controls: Current Account Balance (%GDP), General Government Balance (% GDP), Government Debt (%GDP), real GDP per capita, Inflation rate. a. Information for Macedonia not included Weights: Sum of value common shares for all firms in sample/Stock market capitalization. Information of Stock market capitalization available for 33 countries. Calculations per country-year. Year dummies included, errors clustered at country level. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 25 6 Conclusion Using firms’ balance-sheet information for 69 Emerging Markets and Developing Economies (EMDEs), this paper introduces the Corporate Vulnerability Index (CVI) which provides a framework to monitor financial conditions of the non-financial corporate sector. The CVI employs seven indicators for which data are readily available that capture four aspects of firms’ financial vulnerabilities: debt service capacity, leverage, rollover risk, and return-to-market value. The Corporate Vulnerability Index suggests that vulnerabilities have increased sharply since 2013, but have stabilized in 2016. At the global level, increased leverage ratio and deteriorations in profitability and debt service capacity were the main drivers behind corporate vulnerability. But global trends mask regional diversity. Corporate vulnerability in Eastern Europe has been elevated since 2007, while Latin America experienced a steep increase in vulnerability since 2013. Corporate vulnerability in EMDEs has been stable in 2016, but the Middle East and Sub-Saharan Africa have shown deterioration. Relative to 2009, financial conditions in the non-financial corporate sector have deteriorated in several EMDEs. However, if currently relatively benign global funding conditions and higher commodity prices endure, companies may have an opportunity to strengthen their balance sheets. At the industry level, we find that energy-linked sectors in particular have experienced rising financial vulnerabilities; especially since the 2014 peak in global oil prices. We also find that debt of non-financial listed firms in EMDEs has increased both in level and in riskiness: EMDEs’ corporate debt rose by 46 percent, from US $2.6 trillion in 2010 to US $3.8 trillion in 2016. Over the same period, debt in the hands of firms that are considered financially vulnerable in at least one indicator grew by 120 percent, from $1 trillion to $2.2 trillion. In 2016, approximately 58 percent is considered at risk according to at least one indicator. Logit regressions suggest that the CVI has some leading indicator qualities: an increase in the CVI is positively associated with a future rise in unemployment and an economic recession. Results are robust to controlling for macroeconomic conditions. The CVI also outperforms the commonly used Debt at Risk measure based on the Interest Coverage Ratio (ICR). 26 As financial vulnerabilities in non-financial corporate sectors of many EMDEs appear to be growing, it seems important to consider the efficiency of the EMDEs’ institutional and policy frameworks to monitor vulnerabilities and to deal with distressed firms in case adverse shocks materialize. For instance, it is unknown whether recently established macroprudential frameworks to monitor corporate debt are effective. In addition, inadequate regulatory frameworks like deficient insolvency regimes, poor financial institutions to deal with non- performing loans in the banking sector, or macro policies that discourage hedging of firms’ foreign currency positions might work as amplifiers of adverse shocks. Consequently, policies aimed at minimizing corporate vulnerabilities and controlling their spillover effects as well as the legal framework providing a diverse “menu” of options for ailing firms to obtain efficient financial restructuring, are vital elements of a strategy to improve the resilience of EMDEs’ corporate sector to adverse shocks. 27 7 Bibliography Acharya, V., S. Cecchetti, J. De Gregorio, S. Kalemli-Ozcan, P. Lane, and U. Panizza, 2015. “Corporate debt in emerging economies: A threat to financial stability?”. Committee for International Policy Reform. Brookings Institution, Washington DC. Aiyar, S., W. Bergthaler, J.M. Garrido, A. Ilyina, A. Jobts, K. Kang, D. Kovtun, Y. Liu, D.Monaghan, and M. Moretti, 2015. “A Strategy for Resolving Europe’s Problem Loans”. IMF Staff Discussion Note 15/19. International Monetary Fund, Washington DC, September. Alfaro, L., G. Asis, A. Chari, and U. Panizza, 2017. “Lessons Unlearned? Corporate Debt in Emerging Markets”. NBER Working Paper No. 23407. National Bureau of Economic Research, Cambridge, MA, May. Alloway, T. and J. Weisenthal, 2015. “These Charts Show the Astounding Growth in Emerging Market Corporate Bonds,” Bloomberg, June 8. Altman, E., 1968. Financial Ratios, Discriminant Analysis, and the Prediction of Corporate Bankruptcy. Journal of Finance 23:589–609. Bank of International Settlements, 2015. 85th Annual Report, April 2014-March 2015. Basel, Switzerland, June Becker, B. and V. Ivashina, 2014. “Cyclicality of credit supply: Firm level evidence,” Journal of Monetary Economics 62(C), pages 76-93. Beltran, D., K. Garud, and A. Rosenblum, 2016. “Emerging Market Nonfinancial Corporate Debt: Recent Developments and Risks”. IFDP Note. Bergthaler, W., K. Kang, Y. Liu, and D. Monaghan, 2015, “Tackling Small and Medium Sized Enterprise Problem Loans in Europe.” IMF Staff Discussion Note 15/04. International Monetary Fund, Washington DC, March. Brauer, J., 2017. “GEMs Flow Talk, Bye Bye Debt outflows”, Bank of America Merrill Lynch, 05 January. Bruno, V. and H.S. Shin, 2015a. “Capital Flows and the RiskTaking Channel of Monetary Policy”. Journal of Monetary Economics 71, pages 119-132. Bruno, V. and H.S. Shin, 2015b. “Global Dollar Credit and Carry Trades: A Firm-Level Analysis.” Working Paper, BIS. Caruana, J., 2016. “Credit, Commodities, and Currencies”. Lecture at the London School of Economics and Political Science, London, 5 February 2016. 28 Chan-Lau, J., and T. Gravelle, 2005. “The END: A New Indicator of Financial and Nonfinancial Corporate Sector Vulnerability”. IMF Working Paper 05/231. International Monetary Fund, Washington DC, December. Committee on the Global Financial System, 2014. “EME Banking Systems and Regional Financial Integration,” Report submitted by a Study Group Established by the Committee on the Global Financial System, Basel, March. Cortina, J. J., T. Didier, and S. Schmukler, 2016. “How Long Do Corporates Borrow? Evidence from Capital Raising Activity.” Working Paper, World Bank, Washington, DC. Dehn, J. and J. Arthur, 2015. “The Emerging Markets Investment Universe,” The Emerging View, Ashmore, August. Didier, T., R. Levine, and S. Schmukler, 2015. “Capital Market Financing, Firm Growth, and Firm Size Distribution”. Policy Research Working Paper 7353. World Bank, Washington DC, July. Feyen, E., S. Ghosh, K. Kibuuka, and S. Farazi, 2015. “Global Liquidity and External Bond Issuance in Emerging Markets and Developing Economies.” World Bank Policy Research Working Paper 7363, World Bank, Washington. Financial Stability Board, 2015. Corporate Funding Structures and Incentives – Final Report. Basel, Switzerland, 28 August. Gabaix, X., 2011. “The Granular Origins of Aggregate Fluctuations”. Econometrica 79, pp. 733-72. Geneva Report, 2014. Deleveraging? What Deleveraging? 16th CEPR-ICMB Geneva Report on the World Economy. CERP-International Center for Monetary and Banking Studies. Geneva, Switzerland, September Gonzalez-Miranda, M., 2012. “Nonfinancial Firms in Latin America: A Source of Vulnerability?”. IMF Working Paper 12/279. International Monetary Fund, Washington DC, November. Goretti, M., and M. Souto, 2013. “Macro-Financial Implications of Corporate (De)Leveraging in the Euro Area Periphery”. IMF Working Paper 13/154. International Monetary Fund, Washington DC, June. Gray, D., 1999. “Assessment of Corporate Sector Value and Vulnerability”. WB Working Paper Series WTP 455. World Bank Group. Washington DC, November. Gray, D., R. Merton, and Z. Bodie, 2006. “A New Framework for Analyzing and Managing Macrofinancial Risks of an Economy”. Working Paper #07-026. Harvard Business School. Boston, MA. October. International Monetary Fund, 2015. “Chapter 3: Corporate Leverage in Emerging Markets – A Concern?” Global Financial Stability Report—Vulnerabilities, Legacies, and Policy Challenges: Risks Rotating to Emerging Markets. Washington DC. October. 29 International Monetary Fund, 2016a. “Chapter 1: Financial Stability Challenges in a low-growth, low-rate era”. Global Financial Stability Report – Fostering Stability in a Low-Growth, Low-Rate Era. Washington DC, October. International Monetary Fund, 2016b. Staff Report for the 2016 Article IV Consultation, Former Yugoslav Republic of Macedonia. Washington DC, October. International Monetary Fund, 2017. “Chapter 1: Getting the Policy Mix Right”. Global Financial Stability Report—Getting the Policy Mix Right. Washington DC. April. Institute of International Finance, 2015. “Corporate Debt in Emerging Markets. Where are the risks?” CMM Research Note. Washington DC, June. Institute of International Finance, 2017. “Corporate Capex: Financing Not a Constraint…Yet” CCM Research Note. Washington DC, March. Kamil, H., 2012. “How Do Exchange Rate Regimes Affect Firms’ Incentives to Hedge Currency Risk? Micro Evidence for Latin America”. IMF Working Paper 12/69. International Monetary Fund, Washington DC, March. Lindner, P., and S. Jung, 2014. “Corporate Vulnerabilities in India and Banks’ Loan Performance”. IMF Working Paper 14/232. International Monetary Fund, Washington DC, December. Merton, R., 1974. On the Pricing of Corporate Debt: The Risk Structure of Interest Rates. Journal of Finance, 29(2), pp. 449-470. Rodrigues, F., H. Kamil, and B. Sutton, 2015. “Corporate Financing Trends and Balance Sheet Risks in Latin America: Taking Stock of ‘The Bon(d)anza’”. IMF Working Paper 15/10. International Monetary Fund, Washington DC, January. Ruscher, E., and G. Wolff, 2012. “Corporate balance sheet adjustment: stylized facts, causes and consequences”. Economic Papers 449. European Commission, Economic and Financial Affairs. Brussels, Belgium, February. World Bank, 2016. “Special Focus 1: Recent Credit Surge in Historical Context”. Global Economic Prospects - Divergences and Risks. World Bank. Washington DC, June. World Bank, 2017. Global Economic Prospects - Weak Investment in Uncertainty Times. World Bank. Washington DC, January. 30 8 Appendix Table A1: Country Ranking 2016 based on Corporate Vulnerability Index (CVI) * Data as of 2015. GDP for Venezuela, BR is not available. The Corporate Vulnerability Index (CVI) ranges from 0 (i.e., firms are not financially vulnerable in any of the 7 indicators) to 1 (i.e., all firms are financially vulnerable in all 7 indicators). The CVI calculated for 69 countries, covering 14.207 firms. For CVI calculation, countries require to have 5 or more firms in the database. YoY Change of CVI: percentage change of CVI between 2015 and 2016. Red: change of CVI in the highest tercile across countries. Green: change of CVI in the lowest tercile across countries. DaR >=X denotes the Debt at Risk in "X or more indicators": debt of financially vulnerable firms in X or more indicators as a share of reported total corporate debt, where X=1,2,…,7. DaR denotes the Debt at Risk regarding a particular indicator. Source: Bloomberg; own calculations. 31 Table A1: Country Ranking 2016 based on Corporate Vulnerability Index (CVI) (continued) 32 Table B1: Sensitivity Analysis of CVI Country Ranking 2016 using different ICR thresholds CVI Country Ranking St CVI Country Ranking St Country Average Country Average Benchmark ICR<=1.5 ICR<=2 ICR<=3 Deviation Benchmark ICR<=1.5 ICR<=2 ICR<=3 Deviation Venezuela, RB 1 1 1 1 1.0 0.00 Bangladesh 36 35 32 34 34.3 1.71 Bosnia and Herzegovina 2 2 2 2 2.0 0.00 Israel 37 39 39 42 39.3 2.06 Tanzania 3 3 3 4 3.3 0.50 Kuwait 38 37 40 38 38.3 1.26 Russian Montenegro 4 4 5 5 4.5 0.58 39 40 41 43 40.8 1.71 Federation Macedonia 5 5 4 3 4.3 0.96 Colombia 40 42 24 18 31.0 11.83 Ukraine 6 6 6 7 6.3 0.50 Poland 41 41 43 39 41.0 1.63 Tunisia 7 8 9 9 8.3 0.96 Uruguay 42 24 28 30 31.0 7.75 Kazakhstan 8 10 8 6 8.0 1.63 Mongolia 43 43 44 47 44.3 1.89 Latvia 9 7 7 8 7.8 0.96 Argentina 44 44 30 29 36.8 8.38 United Arab Emirates 10 12 16 12 12.5 2.52 Nigeria 45 45 46 46 45.5 0.58 Kenya 11 13 19 22 16.3 5.12 Côte d’Ivoire 46 46 45 48 46.3 1.26 Mauritius 12 17 14 16 14.8 2.22 Thailand 47 47 48 51 48.3 1.89 Bulgaria 13 16 21 25 18.8 5.32 Turkey 48 50 51 45 48.5 2.65 Brazil 14 15 15 15 14.8 0.50 Ecuador 49 52 53 57 52.8 3.30 Croatia 15 14 17 17 15.8 1.50 Chile 50 49 49 50 49.5 0.58 Egypt, Arab Rep 16 19 12 13 15.0 3.16 Peru 51 53 52 53 52.3 0.96 Saudi Arabia 17 20 22 26 21.3 3.77 Lithuania 52 54 55 59 55.0 2.94 Slovak Republic 18 11 13 19 15.3 3.86 Malaysia 53 51 47 49 50.0 2.58 India 19 18 18 20 18.8 0.96 Qatar 54 48 50 36 47.0 7.75 Morocco 20 23 25 27 23.8 2.99 Mexico 55 56 56 56 55.8 0.50 Oman 21 22 10 10 15.8 6.65 Sri Lanka 56 57 57 55 56.3 0.96 Zimbabwe 22 25 29 23 24.8 3.10 Jordan 57 55 54 58 56.0 1.83 Panama 23 9 11 11 13.5 6.40 South Africa 58 58 58 60 58.5 1.00 Ghana 24 21 27 28 25.0 3.16 Philippines 59 59 59 52 57.3 3.50 China 25 26 26 24 25.3 0.96 Palestine 60 60 60 62 60.5 1.00 Hungary 26 29 34 37 31.5 4.93 Costa Rica 61 61 61 63 61.5 1.00 Romania 27 34 37 32 32.5 4.20 Bahrain 62 62 62 64 62.5 1.00 Indonesia 28 30 33 33 31.0 2.45 Jamaica 63 63 63 65 63.5 1.00 Vietnam 29 27 23 21 25.0 3.65 Estonia 64 64 64 54 61.5 5.00 Trinidad & Serbia 30 32 38 40 35.0 4.76 65 65 65 68 65.8 1.50 Tobago Czech Bolivia 31 31 36 41 34.8 4.79 66 67 67 69 67.3 1.26 Republic Slovenia 32 33 20 14 24.8 9.29 Paraguay 67 68 68 66 67.3 0.96 Zambia 33 36 35 35 34.8 1.26 Botswana 68 66 66 61 65.3 2.99 Pakistan 34 28 31 31 31.0 2.45 Malawi 69 69 69 67 68.5 1.00 Georgia 35 38 42 44 39.8 4.03 Source: Own calculations. 33 Sensitivity Analysis of Association between CVI and social outcomes, using different ICR thresholds Logit Regressions, 2006-2016 Table B2: Unemployment Rise Variables Benchmark ICR<=1.5 ICR<=2 ICR<=3 L. Vulnerability Index (CVI) 6.835** 5.606** 5.898*** 5.737*** (2.673) (2.232) (2.146) (1.932) Constant -3.036*** -2.981*** -3.044*** -3.154*** (0.979) (0.982) (0.993) (1.013) Macro Controls^ Yes Yes Yes Yes Observations 287 287 287 287 ROC 0.717 0.718 0.722 0.720 Marginal Effects at Median 1.359** 1.114** 1.174*** 1.145*** Unemployment rise: Dummy variable 1 if unemployment rate increases, 0 otherwise. Information for Macedonia not included Table B3: Recession Variables Benchmark ICR<=1.5 ICR<=2 ICR<=3 L. Vulnerability Index (CVI) 11.06*** 9.182*** 8.588** 8.837** (3.816) (3.491) (3.369) (3.635) Constant -4.770*** -4.509*** -4.492*** -4.680*** (1.320) (1.276) (1.236) (1.171) Macro Controls^ Yes Yes Yes Yes Observations 232 232 232 232 ROC 0.896 0.887 0.886 0.887 Marginal Effects at Median 1.016*** 0.889*** 0.833** 0.855** Recession: Dummy variable 1 if GDP growth is negative, 0 otherwise. Table B4: Welfare Reduction Variables Benchmark ICR<=1.5 ICR<=2 ICR<=3 L. Vulnerability Index (CVI) 5.832** 4.803** 4.712** 4.419* (2.307) (2.235) (2.121) (2.360) Constant -4.940*** -4.825*** -4.850*** -4.862*** (1.059) (1.035) (1.042) (1.053) Macro Controls^ Yes Yes Yes Yes Observations 340 340 340 340 ROC 0.841 0.840 0.839 0.838 Marginal Effects at Median 0.657*** 0.549** 0.538** 0.508* Welfare reduction: Dummy variable 1 if GDP per capita growth is negative, 0 otherwise Unit of observation: country-year. ^ Macro controls: Current Account Balance (%GDP), General Government Balance (% GDP), Government Debt (%GDP), real GDP per capita, Inflation rate. Weights: Sum of value common shares for all firms in sample/Stock market capitalization. Information of Stock market capitalization available for 33 countries. Calculations per country-year. Year dummies included, errors clustered at country level. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Source: Own calculations. 34