Emerging Market Corporate Leverage and Global Financial Conditions

Corporate debt in emerging markets has risen significantly in recent years amid accommodative global financial conditions. This paper studies the relationship of leverage growth in emerging market (EM) firms to U.S. monetary conditions, and more broadly, to global financial conditions. We find that accommodative U.S. monetary conditions are reliably associated with faster EM leverage growth during the past decade. Specifically, a 1 percentage point decline in the U.S. policy rate corresponds to an appreciable increase in EM leverage growth of 9 basis points, on average (relative to the sample average leverage growth of 35 basis points per year). This impact is more pronounced for sectors dependent on external financing, for SMEs, and for firms in more financially open EMs with less flexible exchange rates. The findings suggest that global financial conditions affect EM firms’ leverage growth in part by influencing domestic interest rates and by relaxing corporate borrowing constraints.

Abstract

Corporate debt in emerging markets has risen significantly in recent years amid accommodative global financial conditions. This paper studies the relationship of leverage growth in emerging market (EM) firms to U.S. monetary conditions, and more broadly, to global financial conditions. We find that accommodative U.S. monetary conditions are reliably associated with faster EM leverage growth during the past decade. Specifically, a 1 percentage point decline in the U.S. policy rate corresponds to an appreciable increase in EM leverage growth of 9 basis points, on average (relative to the sample average leverage growth of 35 basis points per year). This impact is more pronounced for sectors dependent on external financing, for SMEs, and for firms in more financially open EMs with less flexible exchange rates. The findings suggest that global financial conditions affect EM firms’ leverage growth in part by influencing domestic interest rates and by relaxing corporate borrowing constraints.

I. Introduction

Corporate leverage in emerging markets has risen sharply in recent years amid exceptionally favorable global financial conditions. In fact, the corporate debt of nonfinancial EM firms quadrupled from about US$4 trillion in 2004 to well over US$18 trillion in 2014 (Figure 1). Likewise, the EM corporate debt-to-GDP ratio has also risen by 26 percentage points in the same period. In a comprehensive survey of the literature, which predominantly focuses on advanced economies, Frank and Goyal (2009) identify the most reliable determinants of corporate leverage. In particular, for the United States, these include firm-, sector-, and country-level determinants.2 We seek to complement this vast literature by investigating whether global financial conditions—such as global monetary conditions—influence leverage growth in emerging markets (EMs).

Figure 1.
Figure 1.
Figure 1.

Emerging Markets: Aggregate and Firm-Level Measures of Corporate Leverage

Citation: IMF Working Papers 2016, 243; 10.5089/9781475560480.001.A001

Sources: Bank for International Settlements; Dealogic; IMF; Orbis; and authors’ calculations.Note: The selected emerging markets are presented in Appendix Table 1.

Conceptually, accommodative global monetary conditions can encourage EM leverage growth via several related, and potentially mutually reinforcing channels. Since Calvo, Leiderman, and Reinhart (1993), studies have argued that monetary policy loosening in advanced economies is typically accompanied by greater EM capital inflows.3 For example, if central banks cut interest rates to mitigate appreciation pressures when confronted with large capital inflows, the ensuing lower domestic interest rates would then encourage corporate borrowing as it stimulates demand. At the same time, favorable global monetary conditions can foster leverage growth by relaxing financial (borrowing) constraints. In particular, firms that are most dependent on external finance for their business operations, which would likely include small- and medium-sized enterprises (SMEs) and/or companies with limited collateral to pledge, stand to benefit the most from accommodative global financial conditions, and would more likely disproportionately increase their leverage ratios relative to other types of firms.

Accordingly, the main question this paper addresses is the following: Are more accommodative global financial conditions associated with higher EM corporate leverage growth? In addition, our empirical framework sheds light on the following questions: What is the role of country-specific characteristics such as financial openness or the exchange rate regime? What can we say about the channels through which global financial conditions influence EM leverage growth?

Our empirical analysis begins by regressing leverage growth against a measure of global financial conditions, standard firm-level determinants of leverage, and other controls using data from more than 400,000 firms (including small- and medium-sized enterprises, SMEs) in 24 EMs. Initially, we proxy global financial conditions using a measure of the U.S. monetary policy stance, as is common in the literature, but we consider other indicators as well, including “shadow rates” and estimated monetary policy shocks. This setup helps sharpen identification because global monetary conditions can be seen as exogenous to any individual EM firm. Furthermore, to help distinguish the role of global financial conditions from other global factors, we differentiate firms based the degree of financial constraints they face (including, for example, firms’ dependence on external finance). This differentiation facilitates identification because it is more likely that global financial conditions would disproportionately affect more financially constrained firms as compared to, for instance, global growth or commodity prices fluctuations.

We find compelling evidence suggesting that accommodative U.S. monetary conditions are positively associated with faster EM corporate leverage growth:

  • A 1 percentage point decline in the U.S. policy rate corresponds to an increase in EM leverage growth of 9 basis points, on average, an appreciable increase given the sample average leverage growth of 35 basis points per year.

  • Furthermore, this impact is more pronounced for firms with a relatively high intrinsic dependence on external financing. For instance, a decrease in the U.S. policy rate of one standard deviation is associated with leverage growth that is about 5 basis points greater for firms whose financial dependence is at the 75th percentile relative to firms whose financial dependence is at the 25th percentile. This difference is appreciable because incremental rises in leverage can build up over time, especially in the context of persistently loose global financial conditions. Likewise, relative to other types of firms, SMEs and/or firms with less collateral also disproportionately increase their leverage ratios amid accommodative U.S. monetary conditions.

  • We also find that the impact of U.S. monetary policy conditions on EM leverage growth is greater for sectors that are more heavily dependent on external funding in financially open EMs with relatively more rigid exchange rate regimes.

These findings suggest that global financial conditions affect EM firms’ leverage growth in part by influencing domestic interest rates and by relaxing corporate borrowing constraints. A battery of checks underscores the robustness of these results.

This paper contributes to the literature along several dimensions. First, in contrast to Feyen and others (2015), Gozzi and others (2015), among others, who primarily focus on bond issuance, this paper considers total debt (which encompasses both bond- and bank-based debt, among others), thus providing a much more comprehensive picture of how EM corporate leverage growth is influenced by global financial conditions. Second, as opposed to much of the corporate finance literature that focuses on listed firms (predominantly in the United States), we consider SMEs and other private firms in addition to listed firms, to get a more comprehensive picture of corporate leverage dynamics. The importance of considering SMEs and other non-listed firms is emphasized by Kalemli-Ozcan, Sorensen, and Yesiltas (2012) who spotlight leverage dynamics across advanced economies. Third, and in the spirit of Frank and Goyal (2009), we uncover a new, quantitatively important, and reliable determinant of capital structure that is likely to be of relevance for any small, financially integrated emerging or advanced economy.4 In addition, we highlight how the relationship between global financial conditions depends on sector- and country-specific features including dependence on external financing, financial openness, and exchange rate regime. Fourth, this paper provides novel empirical evidence that financial frictions play an important role in the transmission of monetary policy to the real economy, namely to non-financial corporate sector, across EMs.

We proceed as follows. The next section discusses the conceptual and empirical frameworks. Section III gives an overview of the data and variable definitions while relegating additional details to the Appendix. Section IV presents the main results of the paper, along with a very large array of robustness exercises, and Section V concludes.

II. Methodology

This section provides an overview of the conceptual and empirical frameworks that underpin the subsequent analysis.

A. Leverage Growth and Global Financial Conditions

In principle, global monetary conditions can influence EM leverage growth through several interrelated channels. Rather than providing a comprehensive survey, we provide an overview of the two broad channels that are most relevant for the empirical analysis. First, after Calvo, Leiderman, and Reinhart (1993, 1996), many other papers have documented that monetary policy loosening in advanced economies is characteristically accompanied by greater EM capital inflows. Likewise, more recent studies document a link between EM capital flows and global financial conditions—where U.S. monetary policy takes center stage—including Rey (2015), Miranda-Agrippino and Rey (2015), Fratzscher, Lo Duca, and Straub (2013) and Bruno and Shin (2015). During episodes of large capital inflows, if, for instance, EM central banks react by lowering policy rates more than they would otherwise to alleviate currency appreciation pressures, these lower rates would be transmitted to the real economy and foster corporate borrowing as it stimulates demand.

Second, accommodative global monetary conditions may promote leverage growth by relaxing borrowing constraints. Building on the work of Kiyotaki and Moore (1997), Bernanke, Gertler, Gilchrist (1999), and Iacoviello (2005), open-economy models developed by Gertler, Gilchrist, and Natalucci (2007), Elekdag and Tchakarov (2007) and Fernandez and Gulan (2015), among others, include financial frictions, which can take the form of borrowing constraints, thus prohibiting some firms from implementing their desired investment projects as they are not able to secure the needed funding. These frictions underpin a financial accelerator mechanism whereby the cost of debt, asset prices (including the exchange rate), and collateral valuation, jointly interact and determine the demand for capital and debt. If, for example, lower global interest rates push down domestic rates, this would raise the value of collateral, improve corporate financial positions, and therefore relax borrowing constraints. In turn, greater access to capital sets in motion a feedback loop where increased borrowing, leverage, investment, and output boost asset prices further, thereby further relaxing borrowing constraints. In sum, firms that are most dependent on external finance for their business operations, which would likely include SMEs and/or companies with limited collateral to pledge, stand to benefit the most from accommodative global financial conditions, and would therefore increase their leverage ratios disproportionately relative to other types of firms.

Accordingly, to find evidence that global financial conditions influence EM leverage growth by relaxing borrowing constraints, we use three proxies for these constraints. First, as our main proxy, we follow Rajan and Zingales (1998) and differentiate firms based on their intrinsic dependence on external financing. Second, in the spirit of Gertler and Gilchrist (1993), we argue that SMEs are more likely to face borrowing constraints. Third, as in Braun and Larrain (2005), we use asset tangibility to capture the binding nature of borrowing constraints. In other words, we differentiate firms based either on their dependence on external finance, their availability of collateral, or their status as an SME.

B. Regression Specifications

To investigate the relationship between EM corporate leverage growth and global financial conditions, we start by estimating the following equation:

ΔLeveragei.s.c.t=α*MonetaryConditionst+δ*Controlsi,s,c,t1+ɛi,s,c,t(1)

where i, s, c, and t, are indices of firms, sectors, countries, and time. Note that this is an annual panel regression, where firm-level leverage growth is regressed on, Monetary Conditionst, firm-specific controls, which are lagged first differences (profitability, size, and tangibility), and macroeconomic conditions (the ICRG index) in some specifications. Furthermore, firm-specific fixed effects are included to account for unobserved firm-level factors (as are combinations of time, country-time, and sector-time fixed effects). In the baseline specifications, we report standard errors that are corrected for clustering by sector, although we consider other possibilities as well, such as two-way clustering (for example, by sector and time). The slope coefficient, α, measures the extent to which the monetary conditions affects EM leverage growth; given the sharp rise in the latter amid favorable global financial conditions, we expect α > 0.

To identify the transmission of global financial conditions on corporate leverage more precisely, we differentiate firms based on their degree of financial constraints they face. Therefore, we introduce the interaction between the Monetary Conditionst and Financial Constraintss (which could, for example, include a measure of a sector’s dependence on external financing in the spirit of Rajan and Zingales, 1998):

ΔLeveragei.s.c.t=α*MonetaryConditionst+δ*Controlsi,s,c,t1(2)+β*MonetaryConditionst*FinancialConstraintss+ɛi,s,c,t

The slope coefficient on the interaction term, β, captures the extent to which the effect of monetary policy on leverage growth hinges on the nature of firms’ financial constraints. We anticipate that favorable global financial conditions will matter more for financially constrained firms, that is β> 0.

Lastly, we investigate if the impact of global financial conditions on EM corporate leverage varies across countries by adding interaction terms between various country characteristics (such as, financial openness and exchange rate regime) and the inverse shadow rate. In other words, the equation above is augmented as follows:

ΔLeveragei.s.c.t=α*MonetaryConditionst+δ*Controlsi,s,c,t1+β*MonetaryConditionst*FinancialConstraintss(3)+γ*MonetaryConditionst*CountryTraitc,t+ɛi,s,c,t

where the slope coefficient on the additional interaction term, γ, then captures the degree to which the effect of shadow rate fluctuations depends on a particular country trait. While the sign of the coefficient on the last interaction term, γ, varies according to the specific country under consideration, we would expect that global financial conditions matter more for EMs that are more financially integrated (i.e., more open capital accounts) and for EMs that have less flexible exchange rate regimes.

III. Data and Variable Definitions

This section summarizes the main variables and data sources used in the analysis, with details relegated to the Appendix.

A. ORBIS

The firm-level dataset used is this paper is ORBIS (Bureau van Dijk Electronic Publishing, BvD), an annual global panel dataset for over 130 million public and private companies. Relative to other firm-level cross-country databases, a key advantage of ORBIS is its wider coverage of both listed and non-listed firms—which includes SMEs. Although ORBIS has the advantage of being more comprehensive with millions of firms represented in the database, more detailed information on financial statements (such as debt) is harder to come by in the context of EMs.5 As explained in detail in the Appendix, our sample covers about 400,000 nonfinancial EM firms over 2004-2013, totaling more than 1.3 million firm-year observations.

B. Measures of leverage

We consider alternative definitions, initially using the total (non-equity) liabilities-to-total asset ratio, TLTA, as our baseline measure of EM corporate leverage (consistent with, for example, Rajan and Zingales, 1995). This is the broadest definition of leverage, and as discussed in detail in the Appendix, circumvents the issue of missing debt data for certain firms (especially SMEs).6 Furthermore, motivated by the clear upward trends in leverage documented in Figure 1, we focus on the growth (change) of EM corporate leverage, rather than its level. We appear to be in good company: De Angelo and Roll (2015) note that “capital structure stability is the exception, not the rule.” Graham, Leary, and Roberts (2015) also consider growth of leverage, in the context of the U.S., a mature economy, thus motivating our focus on leverage growth in the context of faster growing EMs.

C. Global Financial Conditions and Shadow Rates

We initially proxy global financial conditions with measures of the U.S. monetary policy stance, but also account for unconventional monetary policies. In particular, we follow the literature on “shadow rates” which are complementary indicators of the monetary policy stance and can be especially useful once the policy rate has reached the zero lower bound.

D. Controls

As measure of a sector’s intrinsic dependence on external finance, we use the financial dependence measure proposed by Rajan and Zingales (1998); at the firm level we control for size (log sales), profitability (return on assets), and asset tangibility (net property, plant, and equipment to total assets ratio). We also include a measure of overall macroeconomic conditions in certain regressions (see Appendix for details and discussions).

E. Descriptive Statistics

Table 1 reports summary statistics for selected variables. Financial dependence ranges from a low of −2.2 for the Tobacco and Cigarettes sectors, an industry that has been in decline over the last decades, to a high of 3.8 for the Electronic Repair and Related Services, an industry that has seen large growth.

Table 1.

Summary Statistics: Key Variables

article image
Sources: Orbis database; Reserve Bank of New Zealand; PRS Group; Authors’ calculations.Note: Leverage is the ratio of Total non-equity liabilities to Total assets. Sales is the logarithmic transformation of total sales. Profitability is Return-on-assets. Tangibility is defined as Net property, plant, and equipment to Total assets. Financial dependence is the updated version of the original Rajan-Zingales (1998) index based on Tong and Wei (2011). Macroeconomic conditions are proxied by the ICRG economic and financial index. The (inverse) shadow rate is estimated from a term-structure model based on Krippner (2014).

Estimated shadow rates reasonably reflect monetary policy events in unconventional policy regimes. We initially use the U.S. shadow rate estimated by Krippner (2014), which entered negative territory in November 2008, when the Federal Reserve started the Large Scale Asset Purchases program (Figure 2). The shadow rate further declined as the Fed adopted additional unconventional policies. However, it bottomed out in May 2013, when the Fed raised the possibility of tapering its purchases of Treasury and agency bonds, and has continued to increase since then. Likewise, the global shadow rate has been virtually flat in recent years, reflecting that the tighter stances in the United States and the United Kingdom have been offset by accommodative stances in Japan and the euro area (Figure 2).

Figure 2.
Figure 2.
Figure 2.

The Shadow Rates

Citation: IMF Working Papers 2016, 243; 10.5089/9781475560480.001.A001

Sources: Reserve Bank of New Zealand home page; and authors’ calculations.Note: The global shadow rate is the first principal component of the shadow rates of the four central banks (Bank of England, Bank of Japan, European Central Bank, and U.S. Federal Reserve).

In what follows, to facilitate the interpretation of the results, we use the inverse shadow rate (which is just the shadow rate multiplied by -1; this simple transformation is applied to other measures of monetary policy for consistency).

IV. Empirical Results

After presenting the baseline results, this section discusses the implications of country-specific characteristics, and lastly, considers an array of sensitivity exercises to assess the robustness of the main findings.

A. Baseline Results

The baseline results are presented in Table 2. We include firm fixed effects throughout and, to start off with, cluster standard errors at the sector level. In Column 1, as a first pass, we examine the impact of changes in the inverse U.S. shadow rate on EM corporate leverage. We obtain a positive and statistically significant coefficient (0.088). This initial result suggests that expansionary global monetary conditions are associated with faster EM corporate leverage growth. In fact, an increase in the U.S. shadow rate (looser monetary conditions) of 1 percentage point corresponds to an increase in leverage growth of 9 basis points per year, which is not negligible relative to the sample average of 35 basis points (per year).7

Table 2.

Baseline: EM Corporate Leverage and Global Financial Conditions

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). The (inverse) shadow rate is based on Krippner (2014). Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.

Column 2 introduces an interaction term between the corporate sector’s varying dependence on external finance and the inverse U.S. shadow rate which is central to this paper.8 Indeed, in contrast to other firms, we expect that firms in sectors that are more reliant on external finance to increase their leverage ratios faster amid favorable global financing conditions because of less binding borrowing constraints. The regression does not include the financial dependence variable on its own, as it is fully absorbed by the firm fixed effects terms.

We find that the impact of U.S. shadow rate fluctuations is statistically significantly higher for sectors that depend more on external finance. Based on the estimated coefficient in Column 2 (0.039), an increase in the inverse U.S. shadow rate of one standard deviation—corresponding to more accommodative monetary conditions—is associated with leverage growth that is about 5 basis points greater for firms whose financial dependence is at the 75th percentile (Chemicals and Pharmaceuticals) relative to firms whose financial dependence is at the 25th percentile (the Construction sector). This is a notable effect compared to the sample average growth rate of 35 basis points per year. Considering the protracted nature of the exceptionally loose global financial conditions, it is clear how even seemingly incremental increases in leverage can build up over time.

In Column 3, we include dummies for each year to control of other contemporaneous time effects. The inverse shadow rate is now fully captured by these dummies (time fixed effects terms), and is therefore dropped from this specification. The interaction term of interest, Monetary Conditionst * Financial Constraintss, is again statistically significant at the 1 percent level, with an estimated coefficient value of 0.038, which is only marginally lower than in the regression without time dummies.

In Column 4, we include dummies for country-time pairs. These terms absorb the country-specific control, and therefore the control for country-specific macroeconomic conditions is omitted from the regression. Again, the interaction term is still highly statistically significant, and in line with the other coefficient estimates.

Finally, in Column 5, we also add dummies for sector-time pairs (in addition to the country-time fixed effects terms). These terms control for unobserved factors that vary over time for each sector. Not surprisingly, the coefficient associated with the interaction term declines, to 0.017, but is still statistically significant at the 5 percent level. The lower slope estimate most likely reflects the correlations between the interaction term, and the country-time and sector-time dummies.

In sum, these results support our first two hypotheses: (1) we find that accommodative U.S. monetary conditions are reliably associated with faster EM corporate leverage growth, and (2) this impact more pronounced for sectors that relatively more in need of external financing.

B. Country Traits

We now investigate whether and how the impact of the U.S. monetary conditions varies across countries. In Table 3, firm and time fixed effects terms are included in the regressions.

Table 3.

Leverage, Global Financial Conditions, and Country Traits

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Financial development index summarizes country-level information regarding financial institutions and markets based on Sahay and others (2015). Capital account openness is an index based on Chinn and Ito (2006). Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). The (inverse) shadow rate is based on Krippner (2014). Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.

In Column I, we consider the implications of financial development by adding a proxy for domestic financial development and its interaction with the inverse U.S. shadow rate. Many other studies, beginning with King and Levine (1993), have shown that financial development boosts economic growth by relaxing financial constraints. Following this literature, we initially measure domestic financial development with domestic credit to the private sector scaled by GDP.

The interaction term of domestic financial development and the inverse U.S. shadow rate is negative and statistically significant. This finding suggests that more financially developed countries are less sensitive to global financing conditions in part because they benefit from greater domestic funding opportunities.9 This finding is corroborated if we use the financial development index of Sahay and others (2015) as shown under Column II.

In Columns III and IV we consider the role of capital account openness. We find that in countries that have more open capital accounts (that is countries that are more financially open) firms’ leverage growth tends to be more responsive to U.S. monetary conditions. This result holds up when we control for exchange rate regimes and the degree of policy rate synchronization between the U.S. and individuals EMs.10

This last finding hints at an important channel that may be at work: U.S. monetary conditions may affect EM firms’ leverage growth through domestic interest rates. Given a completely liberalized capital account, theory suggests that when a country adopts a fixed exchange rate regime, it must forgo monetary autonomy; that is, its own interest rate must change in response to foreign monetary conditions. Moreover, even countries with flexible exchange rates in practice may choose to use monetary policy to dampen, though not fully prevent, currency fluctuations arising from changing external financial conditions.11 Therefore, we test whether U.S. monetary conditions have stronger effects in countries with more open capital accounts and with less flexible exchange rates, where domestic interest rates have to accommodate exchange rate policy.12

The results are shown in Table 4. Under Columns 1 and 2, using the sample median, countries are split into two groups: those with relatively more open and more closed capital accounts. Similarly, using the median, in Columns 3 and 4, we split the sample into two groups: those with relatively more rigid and more flexible exchange rate regimes. Lastly, under Columns 5 and 6, we compare EMs with less open capital accounts and more flexible exchange rate regimes with EMs that are more financially open and maintain more rigid exchange rate regimes.

Table 4.

Leverage, Global Financial Conditions, Financial Openness, and Exchange Rate Regimes

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Capital account openness is an index based on Chinn and Ito (2006). Exchange rate flexibility is a de facto exchange rate regime classification based on Ilzetzki, Reinhart, and Rogoff (2008). Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). The (inverse) shadow rate is based on Krippner (2014). Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.

The results in Columns 5 and 6 are of most interest. In particular, we find that the coefficient on interaction term, Monetary Conditionst * Financial Depdendences, under Column 6 is estimated to be 0.072, and is statistically significant at the 1 percent level, while the coefficient under Column 5 is statistically not different from zero. Thus, in financially open EMs with more rigid exchange rate regimes, the impact of U.S. monetary policy conditions is more pronounced for sectors that depend more on external finance. In sum, these results lend support to the relevance of the monetary policy transmission channel whereby U.S. monetary conditions influence domestic policy rates, especially in countries with open capital account and with less flexible exchange rates.

C. Robustness

This section summarizes an extensive set of sensitivity exercises. Alternative measures of monetary conditions, financial constraints, firm-specific characteristics, and leverage ratios are the variables considered in the empirical exercises discussed below. Overall, this section highlights the robustness of our baseline empirical setup and findings.

Monetary conditions

Thus far we have used a measure of the U.S. monetary policy stance as a proxy for global financial conditions. We now consider three complementary measures: First, we use the (inverse) global shadow rate in place of the U.S. shadow rate. Recall that the global shadow rate captures the common dynamics of the shadow rates across the major central banks (that is, the Bank of England, Bank of Japan, European Central Bank, and the Federal Reserve). Relative to the U.S. shadow rate, arguably, the global shadow rate is an even more exogenous measure of global financial conditions. In Table 5, the (inverse) U.S. shadow rate is replaced with its global counterpart. We find similar results: the global shadow rate is positively and statistically significantly correlated with EM leverage growth.13

Table 5.

Robustness: Global Shadow Rate

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). The (inverse) global shadow rate is the principal component of the shadow rates in euro area, Japan, and United States based on Krippner (2014). Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.

Second, we consider the Federal fund rate, as well as various Treasury rates. These are more common measures of the U.S. monetary policy stance. Interestingly, although constrained by the zero lower bound, the (inverse) Federal fund rate has the expected sign and is statistically significant at the one percent level (Table 6). Note however, that the coefficient estimate (0.028) is lower than when the shadow rate is used (0.038) most likely reflecting the Federal fund rate does not account for the unconventional policy measures (such as large-scale asset purchases). Treasury rates are various maturities are also presented, and further reinforce the baseline results.

Table 6.

Robustness: U.S. Policy Rates

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). The (inverse) shadow rate is estimated from a term-structure model based on Krippner (2014), while (inverse) federal funds rate is the interest rate at which US depository institutions lend reserve balances to other depository institutions overnight, on an uncollateralized basis. The 2-year rate, 5-year rate, and 10-year rate are the US treasury bond yields for those respective maturities. Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.

Third, we use a measure of U.S. monetary policy shocks in place of the shadow rate. The data is based on Gertler and Karadi (2015).14 This measure is advantageous because it abstracts from monetary policy actions that were already anticipated by the market, and like the shadow rate, it allows for the inclusion of the recent period when U.S. short-term rates are close to the zero lower bound (see also Debola, Rivolta, Stracca 2015). Using such a measure strengthens our case of treating U.S. monetary conditions as exogenous, since U.S. monetary policy is unlikely to be affected in a systematic way by idiosyncratic EM shocks.15 As shown in Table 7, the results once again reinforce the previous findings: there is a positive and statistically significant relationship between the U.S. monetary shocks and EM leverage growth. As we are now considering shocks, it is not surprising that the estimated coefficients are somewhat lower than those reported Table 2. It is also worthy to note that this last set of results takes an international perspective on the transmission channel of U.S. monetary policy, as we shed light on the role of U.S. monetary policy in influencing EM corporate leverage.

Table 7.

Robustness: U.S. Monetary Policy Shocks

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). The (inverse) monetary shocks are surprises in year-ahead futures on the 3-month Eurodollar deposits based on Gertler and Karadi (2015). Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.

We also consider the role of the VIX index, which has been used as an alternative and/or complementary measure of global financial conditions (see papers cited above, including, for example, Rey, 2015). Along with the baseline specifications, Table 8 presents that results when the inverse shadow rate is replaced with the inverse of the VIX (again, to facilitate interpretation). As is clear, the results are in line with our main findings and reaffirm the positive, statistically significant, and robust relationship between global financial conditions and EM leverage growth.

Table 8.

Robustness: VIX

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). The (inverse) shadow rate is based on Krippner (2014). The (inverse) VIX is the Chicago Board Options Exchange Market Volatility Index. Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.

Table 9 includes global growth and oil prices (which were previously accounted for by time fixed effects terms). The results are intuitive. For example, global oil prices and EM leverage appear to move in tandem. Indeed, IMF (2015a) notes that amid elevated commodity prices, energy firms have issued a significant share of nonfinancial EM corporate bonds. More importantly, the coefficients on the shadow rates are essentially unaltered when these alternative global factors are introduced.

Table 9.

Robustness: Other Global Controls

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). The (inverse) shadow rate is based on Krippner (2014). Oil prices and global growth are proxied by World Economic Outlook’ Commodity Price Index and Global real GDP growth, respectively. Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.

Small- and medium-sized enterprises

Our initial measure of financial frictions was based on firms’ dependence on external financing. A complementary measure of financial frictions is to categorize firms is by their size. As discussed in, for example, Gertler and Gilchrist (1993), firm size is a reasonable indicator of capital market access, there being a strong correlation between size and access to external finance. Specifically, SMEs on average rely heavily on intermediary credit, whereas large firms make far greater use of equity, longer-term debt, and commercial paper. In other words, SMEs have a greater tendency to face borrowing constraints.16 Therefore we construct a dummy variable that takes a value of unity if a firm is an SME. 17 The results shown in Table 10 indicate that (1) there is a positive relationship between SME leverage growth and the U.S. shadow rate, and (2) SME leverage growth increases disproportionately amid looser U.S. monetary conditions.

Table 10.

Robustness: SMEs

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). SME is a dummy variable for small and medium-sized enterprises. Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.

Asset tangibility

Asset tangibility is a complementary way to gauge the binding nature of borrowing constraints. As discussed in Braun and Larrain (2005), in an environment with incomplete financial contractibility, having assets that can be easily transferred to investors improves a firms’ access to external funding. As in the literature, we construct a tangible assets ratio by scaling “hard” assets such as (net) property, plant, and equipment by total assets. Because a firm-level measure of asset tangibility is already included, for the interaction terms we create a dummy variable that takes a value of one for firms in the bottom tertile of the distribution in terms of their tangible assets ratios. As in the previous exercises discussed thus far, this new ratio is interacted with the shadow rate. Evidence in Table 11 echoes the results discussed thus far. Briefly, firms with a lower share of tangible assets tend to disproportionately increase their leverage ratios when U.S. monetary conditions are loose. Table 12 considers a triple interaction term, and the main takeaway is that SMEs with less tangible assets to pledge as collateral (presumably these firms that face the most binding borrowing constraints) show an even greater tendency to increase their leverage ratios when global financial conditions are favorable.18

Table 11.

Robustness: Tangibility

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. TAN is a dummy variable indicating that a firm’s tangible assets are in the lower tertile of the distribution. Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.
Table 12.

Robustness: SMEs and Tangibility

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. SME is a dummy variable for small and medium-sized enterprises. TAN is a dummy variable indicating that a firm’s tangible assets are in the lower tertile of the distribution. Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.

Standard errors

To further assess robustness, we consider alternative ways to cluster the standard errors. In the baseline we clustered by sector. In Table 13, we cluster by sector and time, and the coefficient on the interaction term, for example, remains statistically significant. Table 14 summarizes several other ways to cluster standards errors. Again, the main coefficient of interest is statistically significant.19

Table 13.

Robustness: Clustering—Sector and Time

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). The (inverse) shadow rate is based on Krippner (2014). Standard errors are two-way clustered by sector (two-digit level) and time, respectively. Fixed-effects are not reported.
Table 14.

Robustness: Clustering—Other

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). The (inverse) shadow rate is based on Krippner (2014). Standard errors are clustered two-way in regressions 4 and 5. Fixed-effects are not reported.

Firm fundamentals

While not the main focus of the paper, we now consider other firm-specific fundamentals. Although we find sales, profitability, and tangibility to be quite robust across an array of specifications—in line with Rajan and Zingales (1995), for example—other studies, such as Frank and Goyal (2009), use total assets as a proxy for size and also include median industry leverage as firm-specific controls. To this end, in Table 15 we consider combinations that replace sale with assets and/or include median firm leverage. In these specifications, the interaction of the inverse shadow rate and financial dependence again remains statistically significant.

Table 15.

Robustness: Firm Fundamentals

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is the total liabilities-to-total assets ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Total assets are an alternative measure for size. Median sector leverage is computed for each sector (two-digit level) and each year. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). The (inverse) shadow rate is based on Krippner (2014). Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.

Leverage ratios

We also consider alternative leverage ratios (Table 16). These ratios were described above, and the interaction term, Monetary Conditionst * Financial Depdendences, remains statistically significant when it is considered in turn in place of the total liabilities-to-total assets ratio used in the baseline specification.

Table 16.

Robustness: Alternative Leverage Ratios

article image
Source: Authors’ calculations.Note: ***, **, and * denote statistical significance at the 1, 5, and 10 percent levels, respectively. The dependent variable is a different measure of leverage in each regression. TLTE stands for the total liabilities-to-total equity ratio (first differenced). TATE stands for the total assets-to-total equity ratio (first differenced). NTLTA stands for the total liabilities (net of cash)-to-total assets ratio (first differenced). NTLTE stands for the total liabilities (net of cash)-to-total equity ratio (first differenced). NTATE stands for the total assets (net of cash)-to-total equity ratio (first differenced). Sales is the logarithmic transformation of total sales. Profitability is measured by the return-on-assets, while tangibility is the ratio of net property, plant, and equipment to total assets. Firm-specific regressors are first differenced and lagged. Macroeconomic conditions are measured by the ICRG economic and financial index. Financial dependence is an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011). The (inverse) shadow rate is based on Krippner (2014). Standard errors are clustered by sector (two-digit level). Fixed-effects are not reported.

Sectors and countries

To gauge whether a particular sector or country might be driving the results, we conduct two related exercises. We re-estimate the baseline regression, but exclude each sector one by one, and do the same for each country in our sample. Although not reported (results available upon request), the baseline specification is extremely robust to the exclusion of individual countries, including large ones such as China. Indeed, the interaction between the shadow rate and financial dependence, for example, remain statistically significant at the 1 percent level. Exclusion of individual sectors presents a similar picture. The general contractors (construction) sector is the only sector for which exclusion from the regression lowers statistical significance notably.

V. Conclusions

This paper is motivated by sharp rise in emerging market corporate debt in recent years when global financial conditions were exceptionally accommodative. Accordingly, it investigates if there is a reliable relationship between firms’ leverage growth in EMs and global financial conditions—initially proxied using a measure of U.S. monetary conditions. The results suggest an economically meaningful and statistically robust relationship whereby accommodative U.S. monetary conditions are associated with faster EM corporate leverage growth. Moreover, this effect is more pronounced for sectors that depend more on external financing, for SMEs, and for more financially open EMs with less flexible exchange rate regimes. These findings suggest that U.S. monetary conditions affect EM firms’ leverage growth in part by influencing domestic interest rates and by relaxing corporate borrowing constraints.

To our knowledge, this is the first paper to demonstrate that global financial conditions are a reliable determinant of firm-level leverage dynamics. The focus is on EMs; however, this result is likely to be relevant for advanced small open economies as well. The greater role of global factors during a period when they have been exceptionally favorable indicates that EMs must prepare for the implications of a potential tightening of global financial conditions.20

A potential area for future research is to explore the role of institutional environments, particularly corporate governance, in explaining firm capital structure and leverage dynamics. This topic could be relevant especially in the context of EMs where comprehensive crosscountry empirical evidence is relatively scarce. Indeed, country- and firm-level heterogeneity in corporate governance structures could shed light on differential sensitivities to global financial conditions.

Appendix I. ORBIS

This appendix provides further details on the data and variables used in the analysis.

ORBIS

The firm-level dataset used is this paper is ORBIS (Bureau van Dijk Electronic Publishing, BvD), an annual global panel dataset for over 130 million public and private (non-listed) companies. A notable advantage of ORBIS is that it includes non-listed firms, such as SMEs. Data on firms’ financial positions and productive activities is sourced from their balance sheets and income statements. Because ORBIS includes non-listed firms, by construction, all available data is based on book values. Although ORBIS has the advantage of being more comprehensive with millions of firms represented in the database, more detailed information on financial statements is harder to come by in the context of EMs. For example, debt is not reported by many EM firms.21

As with other large micro data sets, the data need to be managed carefully before they can be used for formal econometric analysis. Kalemli-Ozcan and others (2015) discuss challenges of the ORBIS data base and methods to overcome them. Accordingly, when cleaning ORBIS for our purposes, we are guided by the methods laid out in Kalemli-Ozcan and others (2015), Kalemli-Ozcan, Laeven, and Moreno (2015), Kalemli-Ozcan, Sorensen, and Yesiltas (2012), Fons-Rosen and others (2013), and for instance, Gopinath and others (2015). For instance, to avoid double counting and to improve comparability across countries consolidated accounts are considered. We focus on private EM non-financial corporations with total assets in excess of $1 million. As a result, about 60 percent our sample covers SMEs. Finally, all variables are winsorized at 2.5 percent to account for outliers, especially owing to input errors.22 The ORBIS-based firm-level dataset is then merged with a country-specific measure of macroeconomic conditions (ICRG index) and global factors (for example, a measure of the U.S. monetary policy stance, both which are discussed below. In sum, the dataset comprises over 400,000 firms for 24 EMs during 2004-2013, resulting in an unbalanced panel comprising nearly 1.3 million firm-year observations (Appendix Table 1).

Measures of Leverage

Leverage, or financial leverage, is the degree to which a company uses fixed-income securities such as debt. A high degree of financial leverage entails larger interest payments, which negatively affect firm’s profitability. Leverage is usually presented as a ratio, such as debt to assets. The broadest definitions of leverage consider total non-equity liabilities. An advantage of using total liabilities is that it implicitly recognizes that some firms can use trade credit as a means of financing, rather than purely for transactions (Rajan and Zingales, 1995). Another benefit of using total liabilities is its availability. In contrast, for some countries, debt may not be reported in larger datasets that include non-listed firms, which is the reality we face when using ORBIS.

For these reasons, we initially consider the total (non-equity) liabilities-to-total asset ratio, TLTA, as our measure of EM corporate leverage (consistent with, for example, Rajan and Zingales, 1995). Later, we also consider alternative definitions of leverage including the total liabilities-to-total equity and total assets-to-total equity ratios. Furthermore, to account for the fact that leverage may have risen owing to the accumulation of precautionary cash buffers, we consider variations of these ratios where cash is netted out.23

Financial dependence index

As measure of a sector’s intrinsic dependence on external finance, we use the financial dependence measure proposed by Rajan and Zingales (1998). Conceptually, the Rajan and Zingales index aims to identify sectors that are naturally more dependent on external financing for their business operation. They compute a sector’s dependence on external finance as:

FinancialDependence=(CapitalExpendituresCashFlow)/CapitalExpenditures

where cash flow = cash flow from operations + decreases in inventories + decreases in receivables + increases in payables. The index is computed using data on publicly listed US firms, which are judged to be least likely to suffer from financing constraints relative to generally smaller firms in other countries, including EMs. We use an updated version of the original Rajan and Zingales (1998) index based on Tong and Wei (2011) over 1990-2006, which allows us to consider over 50 sectors.24

Firm-level controls

Building on the literature (for example, Rajan and Zingales, 1995) and based on data availability, size (log sales), profitability (return on assets), and asset tangibility (net property, plant, and equipment to total assets ratio) are firm-level controls used in the baseline specification. As noted by Frank and Goyal (2009), the expected signs of these controls are ambiguous based on opposing theoretical predictions.25

Leverageandprofitability: Profitable firms face lower expected costs of financial distress (and find interest tax shields more valuable), and therefore the tax and bankruptcy costs perspective predicts that profitable firms taken on more debt.26 Moreover, the agency costs perspective predicts that the discipline provided by debt is more valuable for profitable firms with more acute free cash flow problems (Jensen, 1986). In contrast, the pecking order theory argues that firms prefer internal finance over external funds, implying that profitability and leverage are negatively correlated.

Leverageandsize: Large, and potentially more diversified, firms face lower default risk. Therefore, the trade-off theory predicts larger firms to have relatively more debt. Conversely, the pecking order theory is usually interpreted as implying an inverse relationship between leverage and firm size (Frank and Goyal, 2009).

Leverageandassettangibility: Tangible assets, such as property, plant, and equipment, are easier for outsiders to value than intangibles, such as goodwill. Therefore, a greater share of tangible assets relative to total assets lowers expected distress costs, and therefore suggests a positive relationship between tangibility and leverage.27 The pecking order theory makes the opposite prediction. Low information asymmetry associated with tangible assets makes equity issuance less costly, and therefore leverage ratios should be lower for firms with a greater share of tangible assets.28

Country-specific controls

In some specifications, we explicitly attempt to account for country-specific macroeconomic conditions. In particular, we follow Bekaert and others (2014), and take the average of the International Country Risk Guide (ICRG) Economic and Financial Risk Ratings. The ICRG economic risk indicator is designed to capture a country's current economic strengths and weaknesses. It combines information on five economic statistics: GDP levels, GDP growth, inflation, government budgets, and the current account. The ICRG financial risk indicator is designed to assess a country's ability to finance its official, commercial, and trade debt obligations. It combines data from five statistics: foreign debt as a percentage of either GDP or exports, the current account as a percentage of exports, official reserves, and exchange rate stability. In both cases, a higher value indicates stronger fundamentals.29 Recall that various theoretical studies have differing predictions regarding the cyclicality of leverage, further motivating our empirical analysis. Although we use the ICRG to control for country-specific macroeconomic conditions, we also consider regressions that include country-time fixed effects, thereby controlling for a wider array of factors that may be affecting firm-level leverage depending on their location and time period in question. 30

Appendix Table 1

Country and Firm Coverage

article image
Sources: Orbis Database; Authors’ calculations.Note: BvD = Bureau van Dijk Electronic Publishing. Cross-sectional statistics presented for 2007. The criteria for firm size categories follow BvD’s definitions.