Call feature and corporate bond yield spreads

J. of Multi. Fin. Manag. 25–26 (2014) 1–20

Contents lists available at ScienceDirect

Journal of Multinational Financial Management journal homepage: www.elsevier.com/locate/econbase

Review

Call feature and corporate bond yield spreads Anis Samet a,∗, Lamia Obay b a School of Business and Management, American University of Sharjah, P.O. Box 26666, Sharjah, United Arab Emirates b University of Wollongong in Dubai, Blocks 5, 14 & 15, Dubai Knowledge Village, P.O. Box 20183, Dubai, United Arab Emirates

a r t i c l e

i n f o

Article history: Received 9 February 2013 Accepted 7 June 2014 Available online 14 June 2014 JEL classification: G30 G32 Keywords: Credit spread Callable bond Cost of debt Cross-listing

a b s t r a c t Callable bonds offer higher yields compared to non-callable bonds. In this paper, we examine the call spread in a global framework, while controlling for firm-level, bond-level, and country-level variables. Using an international sample of 13,936 bonds issued between 1991 and 2007, we find that callable bonds have a positive call spread, which is statistically and economically significant. Our empirical results hold after a battery of robustness checks. We also find that junk callable bonds have a higher call spread than investment-grade callable bonds, which is consistent with the signaling theory. The empirical results also show that highly leveraged firms have a higher call spread than firms with low leverage, a finding that is consistent with the risk-shifting arguments. © 2014 Elsevier B.V. All rights reserved.

Contents 1. 2. 3.

4.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Literature review and hypotheses development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data and methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1. Bonds ratings and credit spreads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2. Control variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Empirical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

∗ Corresponding author. Tel.: +971 6 515 2316; fax: +971 6 515 4065. E-mail addresses: [email protected] (A. Samet), [email protected] (L. Obay). http://dx.doi.org/10.1016/j.mulfin.2014.06.004 1042-444X/© 2014 Elsevier B.V. All rights reserved.

2 4 5 5 6 6 6 8

2

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

5. 6. 7.

4.1. Univariate analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Multivariate analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sensitivity analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Callable yield premium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

8 8 12 19 19 19

1. Introduction When issuing a bond, a firm has the choice between issuing a callable bond or a straight bond. A call provision grants the issuer the right to buy back its already issued bonds prior to the maturity date. In return for the opportunity to call back the bond, the issuer compensates the holder of a callable bond with an option premium. In other words, a callable bondholder writes a call option and receives the premium, but bears the risk to re-invest the proceeds at a lower rate should the issuer exercise its right to call the bond. The writing of the call option entitles the bondholder to a call premium. Hence, the price the bondholder pays for a callable bond is always lower than that of an equivalent straight bond. More specifically, the price of a callable bond is equal to the price of an equivalent straight bond minus the price of the call option. Lower prices lead to higher yields offered by callable bonds over straight bonds. In return for the higher yields offered by callable bonds, investors stand ready to bear reinvestment risk, that is the risk of having to reinvest one’s money at a lower return should the bond be called back. The two strands of literature on callable bonds revolve around firms’ motivations for issuing callable bonds and callable bond pricing. When issuing callable bonds, firms seek to: (1) hedge their interest rate risk (Güntay et al., 2004), (2) hedge investment risk (Chen et al., 2010), (3) benefit from their future positive information, i.e. signaling theory (Chen et al., 2010; Robbins and Schatzberg, 1986), (4) decrease risk-shifting activities (Barnea et al., 1980), and (5) circumvent underinvestment problems (Barnea et al., 1980; Chen et al., 2010). Several theoretical and empirical papers have discussed the pricing of callable bonds. Berndt (2004) breaks down callable bond prices into three different components: a market interest rate component, a call option component, and a default and illiquidity risk component. Jarrow et al. (2010) develop a new reduced-form approach to value callable corporate bonds, which, according to them, fits callable bond prices well and outperforms the traditional structural approach (e.g., Acharya and Carpenter, 2002) and the reduced-form using American option pricing previously used by Duffie and Singleton (1999). In this paper, we analyze the call spread across different bond ratings and for different levels of leverage. We define the call spread as the yield component that is due to the call provision after controlling for bond-, firm-, and country-specific variables. To the best of our knowledge, no paper has empirically focused on the call spread that issuers offer to callable bondholders. Unlike previous research that looks mostly at bonds for the United States and/or denominated in U.S. dollars, we test our hypotheses in a global framework (an international sample of 13,963 bonds) and we use bonds denominated in different currencies.1 We further match the currency of denomination of the treasury security, used as a benchmark, to that of the bond for which the spread is being computed. The aim of our study is, therefore, to quantify the call spread in a global context and to compare the call spread between high-rated and low-rated bonds and between bonds issued by high-leveraged firms and by low-leveraged firms. This is, to our knowledge, the first study that attempts to do so. Previous empirical studies either use the call provision as a control variable in their credit spread specifications or include it in their robustness check analysis. Qiu and Yu (2010), using U.S. bonds issued between 1976 and 1991, find the callable dummy to be positive and statistically significant in their

1 Berndt (2004) considers only one firm when testing his model. Jarrow et al. (2010) and Qiu and Yu (2010) look at bonds issued by U.S. firms. Qi et al. (2010) look at only Eurobonds denominated in U.S. dollars. Ball et al. (2013) compute credit spread using U.S. treasury securities irrespective of the currency denomination of the bond.

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

3

robustness tests. Qi et al. (2010) look at Eurobonds denominated in U.S. dollars and originating from 39 countries between 1980 and 2006 and find that callable bonds have higher spread than straight bonds when they include a dummy variable for callable bonds in three of their model specifications. Francis et al. (2010), using U.S. bonds issued between 1990 and 2000, find that the callable dummy variable is statistically significant. Ball et al. (2013), using a global sample of public bonds but retaining only one bond per firm (the bond with the largest principal amount), find that the callable dummy is significant in only one of their specifications. It, however, becomes statistically not significant when they control for bond-specific variables. Ball et al. (2013) compute the credit spread as the yield to maturity of the bond minus the yield of an equivalent U.S. treasury security, independently of the currency in which the bond is denominated. We circumvent this shortcoming by matching the currency of denomination of the treasury security to that of the bond for which the spread is being computed. In this paper we hypothesize that callable bondholders receive a positive call spread for their willingness to hold the callable bond that would not otherwise be offered on equivalent non-callable bonds. We further hypothesize that firms with higher leverage have a higher call spread, all else equal. According to the risk-shifting hypothesis, first introduced by Jensen and Meckling (1976) and later developed by Barnea et al. (1980) and Chen et al. (2010), shareholders have an incentive to expropriate bondholders’ wealth. Accordingly, we further assume that holders of callable bonds will require a higher call spread for highly leveraged firms to compensate them for the risk of wealth expropriation. Based on the signaling theory (Chen et al., 2010; Robbins and Schatzberg, 1986), we hypothesize that junk bonds have a higher call spread than investment-grade bonds, all else being equal. In other words, we expect firms that issue junk bonds to have a higher incentive to exercise this option and call back their bonds when they reveal their positive private information. The univariate empirical results show that, on average, callable bonds have a higher credit spread than straight bonds. Moreover, callable bonds have lower ratings than straight bonds. The univariate results further reveal that callable bonds are issued by smaller firms, firms with lower leverage, firms with higher growth, and non-U.S. firms that are less likely to cross-list on U.S. markets. The multivariate analysis corroborates the univariate analysis and shows that callable bonds have a positive call spread which is statistically and economically significant across all specifications. We find that the call feature adds between 52 and 58 basis points to the cost of debt, everything else being equal. In other words, the call option premium paid by the callable bond issuer and received by the callable bondholders is between 52 and 58 basis points. We also find that bonds issued by large firms, firms with low leverage, more profitable firms, firms with higher residual ratings, and firms cross-listed on U.S. markets have lower credit spreads. Those results are consistent with the literature on the determinants of the credit spread (e.g., Yu, 2005; Qi et al., 2010). In addition, we find that the country of incorporation has a significant effect on the credit spread. Indeed, issuing firms coming from countries with low sovereign ratings, firms coming from countries that issue a higher proportion of their debt on international markets, and firms that come from countries with no credit bureau, incur a higher cost of debt. The empirical results further reveal that junk bonds have a higher call spread than investment grade bonds, which is consistent with the signaling theory (Chen et al., 2010; Robbins and Schatzberg, 1986) according to which firms issue low-rated bonds are more likely to benefit from signaling their private positive information. When testing for the effect of leverage we find that highly leveraged firms have a higher call spread than firms with low leverage, which is consistent with the risk-shifting hypothesis (Barnea et al., 1980) according to which highly leveraged firms are more likely to undertake riskier projects. The empirical results hold after a battery of robustness checks. Based on the above, our contribution to the callable bond literature is two-fold: first we quantify in basis points the extra cost bond issuers need to bear for issuing callable bonds. This has direct implications on the yield for holding callable bonds as well as the firm’s cost of debt, hence its cost of capital. Second, we provide further empirical support for the signaling and risk-shifting hypotheses underlying the issuance of callable bonds in a global context. The rest of the paper is as follows: in Section 2 we present the literature review and develop our hypotheses. Section 3 introduces our data and methodology. In Section 4, we report the univariate and multivariate results. In Section 5, we conduct a battery of robustness checks. We then compare

4

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

call spread on the basis of rating and leverage in Section 6. We present our concluding remarks in the last section.

2. Literature review and hypotheses development The callable bond literature has developed around two main strands: firm’s motivation for issuing callable bonds and callable bond pricing. The most commonly cited explanation for issuing a callable bond is to hedge interest-rate risk. Indeed, firms issue callable bonds to call them back and refinance them at lower costs when interest rates fall. As such, callable bonds are more likely to be issued during times of high interest rates. Firms are more likely to include a call provision in the case of longer maturity bonds given their greater sensitivity to interest rate fluctuations. They are also more likely to attach a call provision for larger issues because of “more material interest rate exposures that are likely to be created by large-size debt issues” (Güntay et al., 2004, p.16). Banko and Zhou (2010) find that interest rate hedging was the primary motive for issuing bonds during much of the ‘80s. Firms may also offer callable bonds when facing poor investment opportunities. Attaching a call feature alleviates the risk-shifting problem first introduced by Jensen and Meckling (1976) and gives the firm the flexibility to liquidate a project that subsequently reveals to have a negative Net Present Value. Chen et al. (2010) find strong evidence for the hedging of investment risk hypothesis. The third explanation is related to information asymmetry or “signaling theory” (e.g., Banko and Zhou, 2010; Chen et al., 2010; Choi et al., 2013). Firms with information asymmetry problems issue callable bonds at lower prices and then capture the price appreciation and benefit from the option to refinance their bonds at lower costs when their positive private information is revealed. Güntay et al. (2004), however, find no evidence of an improvement in rating subsequent to callable bond issue. The fourth explanation is risk shifting according to which shareholders can expropriate bondholders’ wealth by shifting into riskier assets. According to Barnea et al. (1980), the value of the call option, owned by shareholders, declines should there be an increase in firm’s risk, which reduces the incentives to expropriate bondholders. The last explanation is the underinvestment problem where firms would not invest in positive NPV projects as existing bondholders will take part in the benefits of investing in these new projects (e.g., Myers, 1977; Barnea et al., 1980; Chen et al., 2010). Several theoretical and empirical papers have discussed the pricing of callable bonds. Berndt (2004) disentangles the components of the callable bond prices due to market interest rates, call option, and default and illiquidity risk. Berndt (2004) tests the model on one single bond. Jarrow et al. (2010) develop a new reduced-form approach to value callable corporate bonds. Other empirical papers simply control for the callable feature in some of their credit spread specifications or use it as robustness check. Francis et al. (2010), using U.S. bonds issued between 1990 and 2000, find that the callable dummy variable is statistically significant. Qiu and Yu (2010), using U.S. bonds issued between 1976 and 1991, find that callable dummy is positive and statistically significant in their robustness tests. Qi et al. (2010), using Eurobonds denominated in U.S. dollars and issued by borrowers incorporated in 39 countries between 1980 and 2006, find that callable bonds have higher spread than straight bonds in three of their specifications. Ball et al. (2013), using a global sample of public bonds but retaining only one bond per firm (the bond with the largest principal amount), find that the callable dummy is significant in only one of their specifications and becomes statistically not significant when they control for bond-specific variables. In their paper, the credit spread is being computed as the yield to maturity of the bond minus the yield of an equivalent U.S. treasury security, irrespective of the currency of denomination of the bond. We circumvent this shortcoming by identifying treasury securities with similar currency denomination for the computation of the credit spread. We believe our study makes an important contribution to the literature by quantifying the call spread for a global sample of 13,936 bonds issued between 1991 and 2007 originating from 36 different countries. To the best of our knowledge, no paper has empirically focused on the offering call spread that issuers offer to callable bondholders in a global framework and using bonds denominated in different currencies. Our study is the first study to quantify the call spread in an international framework and compares the call spread between high-rated and low-rated bonds and between bonds issued by high-leveraged and low-leveraged firms.

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

5

In this paper we focus on the callable bonds’ offering call spread, while controlling for firm-level, bond-level, and country-level variables that have been shown to affect the credit spread. We formulate our research hypotheses as follows: Hypotheisis 1.

Call spread is positive.

This hypothesis is plausible since the callable bond price is lower than the straight bond price. The price of a callable bond is equal to the value of an equivalent straight bond less the value of the call provision. Given the inverse relationship between bond prices and bond yields, the yield of a callable bond is, therefore, higher than that of an equivalent straight bond. Investors expect to be compensated for their willingness to bear reinvestment risk when buying bonds that are most likely to be called. Therefore, we expect the callable dummy variable to have a positive and statistically and economically significant coefficient. Hypotheisis 2.

Junk bonds have a higher call spread than investment-grade bonds.

Junk callable bonds are more likely to be called than investment-grade ones. This assumes that firms issue callable bonds in order to benefit from information asymmetry. Firms may exercise their call option when their positive private information is revealed (i.e., signaling theory), when interest rates go down (i.e., to hedge their interest rate risk), or when seeking to hedge their investments. However, firms that issue investment-grade bonds are less likely to call back their bonds, compared to firms that issue junk bonds, when they reveal their positive private information as there is no substantial gain from doing so. Recent studies (Banko and Zhou, 2010) show that below-investment-grade bonds are mainly issued to alleviate agency conflicts, while investment grade bonds are mainly issued to hedge interest rate risks. We expect that the callable dummy variable’s coefficient is strictly higher for junk bonds compared to investment-grade bonds. Hypotheisis 3.

Firms with higher leverage have a higher call spread than firms with low leverage.

According to the risk-shifting explanation (Barnea et al., 1980), shareholders have an incentive to expropriate bondholders’ wealth by shifting into riskier assets when the financial health of the firm deteriorates. The call provision may mitigate such behavior. Indeed, when firms undertake riskier projects, the price of the call option held by the shareholders falls, which in turn reduces the incentives to shift into riskier asset and expropriate bondholders wealth. Highly leveraged firms are more likely to undertake riskier projects. The issuance of callable bonds reduces the incentives to expropriate bondholders. To compensate for the risk of expropriation, we conjecture that bondholders require a higher call spread for highly leveraged firms. We expect that the callable dummy variable’s coefficient is strictly higher for bonds issued by highly leveraged firms than those issued by firms with low leverage. The hypotheses above are tested in an international context using a sample of 13,936 bonds issued between 1991 and 2007. In what follows, we present our data and methodology. 3. Data and methodology 3.1. Data To examine the impact of call feature on credit spreads, we rely on the following databases. First, we use Securities Data Company (SDC) Platinum database to extract bonds issued all over the world between 1991 and 2007. We stop in 2007 to avoid the impact of the financial crisis. We exclude government bonds, sovereign agency issues, bonds issued by financial firms, floating rate bonds, and those with missing values. From SDC, we extract bonds’ characteristics (call provision, coupon rate (fixed vs. floating), maturity, bond rating, convertibility provision, syndication, offering market (public vs. private), total proceeds, market venue, ultimate parent nation, and currency of denomination). To calculate the credit spreads, we collect zero-coupon government-bond benchmarks for different countries, currencies, and maturities using Thomson Reuters DataStream, Bloomberg, and issuing firms’ respective central banks’ websites.

6

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

Firms’ characteristics including total assets, leverage, and EBITA are extracted from Compustat. Institutional variables including sovereign ratings, international debt securities as a share of GDP, credit bureau, creditor rights index, the debt enforcement efficiency index, legal origin, and the legal rights index are from different sources. Table 1 defines the different variables and presents their sources. Our final sample includes 13,936 bonds issued by 1726 firms originating from 39 countries. In what follows, we present a discussion of the dependent variables and control variables in the study. 3.2. Methodology 3.2.1. Bonds ratings and credit spreads In our empirical analysis, we use a two-stage model to test our hypotheses. We first regress bond rating on control variables. We then include the residuals (RATR) of this regression as a control variable in the credit spread regression in order to avoid any potential simultaneity between credit spreads and ratings. We use the S&P bonds’ ratings reported in SDC. We convert the letter rating into a number between 21 and 1 to AAA and C, respectively. We estimate credit rating as a function of issuer-specific, bond-specific, and institutional variables. Our two dependent variables are, hence, bonds ratings (RAT) and credit spreads (CS). RAT = f (issuer, issue, and institutional characteristics)

(1)

CS = f (issuer, issue, institutional characteristics, and RATR)

(2)

To calculate CS, we first calculate the offering yield to maturity of each bond. For a fixed-rate bond, the offering yield to maturity is calculated using the coupon rate, the offering price, and the maturity. For each bond, we match the offering yield to maturity to a government bond yield to maturity based on the currency denomination of the bond and its maturity. For bonds issued in Euro by firms from European countries, those bonds are matched with the respective Euro government bond with the same maturity. For instance a French bond issued in Euros is matched with the French government bond with the same maturity. Moreover, bonds issued before the adoption of the Euro are matched with their respective government bonds. If the maturity of the corporate bond and the government bond do not match, we use linear interpolation between the two closest government bond maturities to match the corporate bond yield to maturity with the government bond yield to maturity. The dependent variable CS is then computed using the difference between the corporate bond yield to maturity and the equivalent government bond yield to maturity. 3.2.2. Control variables We control for firm variables, bond variables, and institutional variables. • Firm control variables: TA: Is equal to the natural logarithm of total assets in million U.S. dollars. LEV: Is the leverage ratio between total liabilities and total assets. PROFIT: Is a profitability ratio which is equal to EBITA over total assets. GROWTH: Annual total assets’ growth rate. USLIST: A dummy variable which is equal to one if the firm is cross-listed on U.S. markets, zero otherwise. Data are obtained at the end of the year prior to the bond issue. • Bond control variables: PUB: A dummy variable which is equal to one if the bond is issued on public markets and zero otherwise. AGE: The natural logarithm of the bond’s age, which is also equal to natural logarithm of the bond’s maturity. PROC: Is equal to the natural logarithm of the total proceeds in million U.S. dollars. CALLABLE: A dummy variable which is equal to one if the bond is a callable bond and zero otherwise. CONV: A dummy variable which is equal to one if the bond is a convertible bond and zero otherwise. SYND: A dummy variable which is equal to one if the bond is syndicated and zero otherwise. COUPON: The bond’s coupon rate. • Institutional control variables: SOVRAT: Standard & Poor’s letter sovereign credit ratings converted into numbers ranging from 22 (AAA with positive outlook) to zero (C with negative outlook) following Gande and Parsley (2010). INTDB: International Debt Securities (outstanding amount) as a share of GDP. CRBUR: The variable equals one if either a public registry or a private bureau operates in the country, zero otherwise. CREDR: The creditor rights index of the ultimate parent nation. The index

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

7

Table 1 Variables, definitions, and sources. Variable

Definition

Panel A: credit spreads and bond ratings variables The credit spreads. CS RAT RATR

The S&P bonds’ ratings reported in Securities Data Company. We convert the letter rating into a number between 21 and 1 to AAA and C, respectively. The residuals from regressing bond ratings on control variables.

Panel B: control variables Firm control variables The natural logarithm of total assets in million U.S. dollars. TA The leverage ratio which is equal to total liabilities over total assets. LEV The profitability ratio which is equal to EBITA over total assets. PROFIT The annual total assets’ growth rate. GROWTH A dummy variable which is equal to one if the firm is cross-listed on U.S. markets, zero USLIST otherwise.

Bond control variables A dummy variable which is equal to one if the bond is issued on public markets, zero PUB otherwise. The natural logarithm of the bond’s maturity. AGE PROC

The natural logarithm of the total proceeds in million U.S. dollars.

CALLABLE

A dummy variable which is equal to one if the bond is a callable bond, zero otherwise.

CONV SYND

A dummy variable which is equal to one if the bond is a convertible bond, zero otherwise. A dummy variable which is equal to one if the bond is syndicated, zero otherwise.

COUPON

The coupon rate.

Institutional control variables SOVRAT Standard & Poors letter sovereign credit ratings converted into numbers ranging from 22 (AAA with positive outlook) to zero (C with negative outlook) following Gande and Parsley (2010). INTDB

International Debt Securities (outstanding amount) as a share of GDP

CRBUR

The variable equals one if either a public registry or a private bureau operates in the country, zero otherwise. A public registry is defined as a database owned by public authorities (usually the Central Bank or Banking Supervisory Authority) that collects information on the standing of borrowers in the financial system and makes it available to financial institutions. A private bureau is defined as a private commercial firm or non-profit organization that maintains a database on the standing of borrowers in the financial system, and its primary role is to facilitate exchange of information amongst banks and financial institutions. The creditor rights index of the ultimate parent nation. The index ranges from zero (weak creditor rights) to four (strong creditor rights) The debt enforcement efficiency index of the ultimate parent nation.

CREDR ENFOR COMMON LEGALR

A dummy variable which is equal to one if the ultimate parent nation is a common law country, zero otherwise. Measures the degree to which collateral and bankruptcy laws protect the rights of borrowers and lenders and thus facilitate lending. The index ranges from zero to ten, with higher scores indicating that collateral and bankruptcy laws are better designed to expand access to credit. The index is taken for 2007.

Source Authors’ specification Securities Data Company Authors’ specification

Compustat Compustat Compustat Compustat NYSE, NASDAQ, JPMorgan, Bank of New York, and Citigroup websites Securities Data Company Securities Data Company Securities Data Company Securities Data Company Securities Data Company Securities Data Company Securities Data Company Standard and Poor’s website and Gande and Parsley (2010) Beck and Demirgüc¸-Kunt (2009), World Bank website Djankov et al. (2008)

Djankov et al. (2008) Djankov et al. (2008) Djankov et al. (2008) Djankov et al. (2007) and doing business website

8

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

ranges from zero (weak creditor rights) to four (strong creditor rights). ENFOR: The debt enforcement efficiency index of the ultimate parent nation. This index ranges from zero (weak enforcement) to 100 (strong enforcement). COMMON: A dummy variable which is equal to one if the ultimate parent nation is a common law country, zero otherwise. LEGALR: Legal rights index. The index ranges from zero to ten, with higher scores indicating that collateral and bankruptcy laws are better designed to expand access to credit. 4. Empirical results In this section, we present the empirical results. First, we start with the univariate analysis and then we discuss the multivariate analysis. 4.1. Univariate analysis Our sample consists of 13,936 bonds issued by 1726 firms over the period 1991–2007. The sample is dominated by investment-grade bonds (76.8%) and is also dominated by bonds issued by U.S. firms (77.7%) followed by German firms (4.8%) and Japanese firms (4.1%). Table 2 presents the mean statistics of the main variables used in our regressions. Table 2 shows that in average 37.7% of the bonds issued were callable bonds. U.S. firms issued 10,830 bonds all over the world, of which 43.1% were callable bonds. The average credit spread differs across countries. Brazil has the highest average credit spread (521.1 basis points) and Japan has the lowest average (38.4 basis points). It is, thus, no surprise that Japanese bonds enjoy the highest rating (17.4), the lowest rating being recorded by Singaporean bonds. The sample includes both public and private bonds (84.9% and 15.1%, respectively). Table 3 presents the correlation between our main variables. As expected, the callable dummy variable (CALLABLE) is positively correlated (39%) with the credit spread (CS). Moreover, this table shows that large firms, firms with higher profitability, firms cross-listed on U.S. markets, public bonds, and bonds with high ratings have lower credit spreads. Table 4 presents the descriptive statistics of our explanatory variables. Table 4 shows that the average credit spread is 141.88 basis points and the standard deviation is 157.60 basis points. The credit spread median is 86.20 basis points, which means that the credit spread is highly skewed (Skewness = 2.99). Table 5 compares callable bonds and straight bonds. As expected callable bonds have on average a higher credit spread (220.92 basis points) than straight bonds (94.04 basis points), giving further support to the signaling hypothesis. Callable bonds have on average lower rating than straight bonds. Callable bonds are issued by smaller firms, firms with lower leverage, firms with higher asset growth ratio, and firms that are less likely to cross-list on U.S. markets. 4.2. Multivariate analysis The estimation results show that callable bonds, as expected, have a higher credit spread between 52 and 58 basis points. This result is significant across the different specifications at the 1% level. This result is consistent with our first hypothesis (H1) and it is statistically and economically significant. This result is important in quantifying the marginal cost of issuing callable bonds compared to straight bonds issued all over the world and in different currencies. Indeed, when a firm’s manager chooses to issue a callable bond, the relative cost of debt increases between 52 and 58 basis points, everything else being equal. In other words, the call option premium paid by the callable bond issuer and received by the callable bondholders is between 52 and 58 basis points. The positive and statistically and economically significant call premium is consistent with the literature on credit spread. Indeed, Qiu and Yu (2010) using U.S. bonds issued between 1985 and 1991, find that the callable dummy variable is positive and statistically significant. Qi et al. (2010), using Eurobonds denominated in U.S. dollars and issued by borrowers incorporated in 39 countries between 1980 and 2006, find also that callable bonds have higher spread than straight bonds. Francis et al. (2010), using U.S. bonds issued between 1990 and 2000, find that the callable dummy variable is statistically significant.

Table 2 Descriptive statistics by country. CS

TA

LEV (%)

PROFIT (%)

GROWTH (%)

USLIST (%)

PUB

RAT

PROC

AGE

CALLABLE (%)

Argentina Australia Austria Belgium Brazil Canada Chile China Denmark Finland France Germany Greece Hong Kong Hungary India Indonesia Italy Japan Luxembourg Malaysia Mexico Netherlands New Zealand Norway Philippines Portugal Singapore South Africa Spain Sweden Switzerland Taiwan Thailand UK United States Total

309.5 128.2 75.8 97.2 521.1 238.8 166.8 312.8 88.9 118.7 130.4 111.9 360.4 214.1 128.4 187.8 401.4 127.9 38.4 160.0 197.9 313.7 253.7 84.2 226.1 390.8 61.1 378.9 433.9 85.6 146.7 111.2 152.3 415.6 123.4 144.5 141.9

8.8 8.1 9.1 7.7 8.8 8.2 8.1 6.4 7.2 7.6 10.3 11.5 7.7 9.4 9.0 8.5 4.7 11.0 11.0 8.6 8.2 8.2 9.2 7.8 9.0 8.1 9.1 7.1 7.2 10.3 8.6 10.0 7.5 7.3 9.6 9.4 9.5

44.6 56.9 65.0 55.8 70.1 63.6 50.5 61.1 64.4 49.6 75.7 77.8 60.2 54.4 55.1 47.2 55.2 78.3 76.6 76.1 54.8 58.4 73.4 64.8 63.6 59.2 68.6 40.9 54.5 73.3 55.1 65.6 55.6 51.0 68.1 72.4 72.1

8.6 7.6 3.4 6.8 1.7 4.8 7.4 6.4 6.6 7.8 4.2 1.9 7.9 7.4 16.1 9.3 14.5 3.3 2.5 1.4 8.0 7.4 2.2 10.0 11.4 5.1 7.2 −4.2 6.2 5.2 8.9 7.5 13.7 5.5 8.3 6.0 5.7

8.0 14.1 −2.0 4.3 49.9 24.1 14.9 58.8 1.8 −42.5 −1.3 10.6 148.5 −17.2 6.7 25.2 4.3 0.5 1.9 3.5 110.4 16.8 37.8 24.8 57.8 16.3 −8.0 32.4 8.6 79.8 24.0 9.6 40.3 −11.5 43.3 16.4 17.0

83.3 79.3 100.0 83.3 70.4 55.0 26.7 0.0 0.0 100.0 82.4 97.3 33.3 71.4 100.0 100.0 33.3 77.6 95.8 54.5 25.0 70.4 75.3 65.0 16.7 14.3 87.5 100.0 60.0 64.9 95.0 90.3 0.0 0.0 65.1 0.0 17.6

0.9 0.6 1.0 0.3 0.9 0.7 0.8 1.0 0.3 1.0 1.0 1.0 0.3 0.8 1.0 0.7 0.8 1.0 1.0 0.9 1.0 0.8 0.9 0.9 0.9 1.0 1.0 0.6 0.8 0.9 0.5 1.0 1.0 1.0 0.9 0.8 0.8

12.2 14.2 13.0 14.7 9.7 11.5 14.1 11.0 15.5 14.0 14.3 15.4 10.2 13.9 12.0 11.5 9.2 14.2 17.4 12.8 13.4 12.1 13.1 15.8 13.6 9.7 15.5 7.4 9.2 16.2 14.2 16.9 13.0 8.0 15.2 14.4 14.5

2.1 2.2 2.1 2.1 1.6 2.4 2.2 1.9 2.2 1.8 1.8 1.3 2.1 2.2 1.6 2.5 1.7 1.9 1.8 1.9 2.4 2.1 2.1 2.2 2.2 2.1 1.9 2.1 2.4 2.0 2.2 1.8 2.3 2.1 1.9 2.0 2.0

5.3 4.8 6.6 4.4 5.1 5.5 5.3 5.3 5.2 5.8 5.8 3.9 5.2 5.9 6.8 5.0 5.1 5.7 4.9 5.6 5.8 5.2 6.0 4.9 5.1 5.1 6.3 5.3 5.7 5.9 5.5 5.4 5.8 5.5 5.6 4.7 4.8

16.7 15.2 0.0 0.0 14.8 71.2 0.0 0.0 0.0 0.0 8.3 8.1 66.7 19.0 0.0 0.0 50.0 14.1 6.1 9.1 0.0 48.1 29.9 0.0 44.4 7.1 0.0 80.0 100.0 3.9 25.0 9.7 0.0 50.0 20.0 43.1 37.7

# of obs. 12 92 3 6 27 333 15 1 4 2 278 669 9 42 1 12 6 85 575 11 8 27 77 20 36 14 8 5 5 77 20 62 1 2 561 10,830 13,936

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

Country

This table reports means by country for the main variables used in our regressions. Our sample consists of 13,936 bonds issued by 1726 firms over the period 1991–2007. Table 1 provides the definitions and data sources for these variables. 9

10

CS TA LEV PROFIT GROWTH USLIST PUB RAT AGE PROC CALLABLE CONV SYND SOVRAT INTDB CRBUR

CS

TA

LEV

PROFIT

GROWTH

USLIST

PUB

RAT

AGE

PROC

CALLABLE

CONV

SYND

SOVRAT

INTDB

CRBUR

1.00 −0.39 0.05 −0.20 0.10 −0.10 −0.44 −0.70 0.11 0.16 0.39 0.10 0.01 −0.10 0.13 −0.04

1.00 0.30 −0.12 −0.07 0.25 0.32 0.57 −0.30 −0.01 −0.32 −0.03 −0.04 0.03 0.17 0.03

1.00 −0.34 −0.10 0.01 0.05 −0.01 −0.17 −0.07 −0.12 0.00 −0.08 0.07 −0.02 0.04

1.00 −0.03 −0.07 0.06 0.22 0.05 0.01 0.00 −0.03 0.02 0.01 −0.01 0.02

1.00 −0.01 −0.09 −0.09 0.02 0.04 0.06 0.00 −0.02 0.00 0.00 0.00

1.00 0.10 0.12 −0.14 0.08 −0.21 0.03 0.03 −0.24 0.44 −0.03

1.00 0.46 −0.09 −0.18 −0.29 −0.06 0.20 −0.02 −0.06 0.01

1.00 −0.17 −0.17 −0.41 −0.06 −0.09 0.06 −0.08 0.04

1.00 0.25 0.37 −0.03 0.29 0.00 −0.03 0.00

1.00 0.27 0.04 0.43 −0.07 0.21 −0.01

1.00 0.04 0.19 0.06 0.12 −0.01

1.00 −0.02 0.01 0.06 0.00

1.00 −0.03 0.21 0.01

1.00 0.01 0.05

1.00 0.01

1.00

This table provides correlation coefficients between the variables used in our main regressions. Our sample consists of 13,936 bonds issued by 1726 firms over the period 1991–2007. Table 1 provides the definitions and data sources for these variables.

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

Table 3 Correlation coefficients.

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

11

Table 4 Descriptive statistics of explanatory variables. Mean

Median

Credit spreads and bond ratings variables 141.88 86.20 CS 14.51 15.00 RAT

Minimum

Maximum

Std. dev.

# of obs.

0.02 1.00

2366.50 21.00

157.60 3.76

13,936 13,936

Firm control variables TA LEV PROFIT GROWTH USLIST

9.55 0.72 0.06 0.17 0.18

9.53 0.72 0.05 0.07 0.00

−2.14 0.00 −1.55 −0.89 0.00

13.45 2.96 1.79 41.09 1.00

1.81 0.17 0.07 0.97 0.38

13,936 13,936 13,936 13,936 13,936

Bond control variables PUB AGE PROC CALLABLE CONV SYND

0.85 1.99 4.78 0.38 0.00 0.57

1.00 2.08 5.16 0.00 0.00 1.00

0.00 0.00 −2.30 0.00 0.00 0.00

1.00 3.66 8.42 1.00 1.00 1.00

0.36 0.75 1.47 0.48 0.05 0.50

13,936 13,936 13,936 13,936 13,936 13,936

21.00 0.15 1.00 1.00 85.80 1.00 8.00

3.00 0.01 0.00 0.00 13.40 0.00 3.00

21.00 1.93 1.00 4.00 96.10 1.00 10.00

1.08 0.16 0.02 0.80 10.04 0.30 0.79

13,936 13,936 13,913 13,913 13,913 13,913 13,930

Institutional control variables 20.80 SOVRAT 0.19 INTDB 1.00 CRBUR 1.29 CREDR 83.91 ENFOR 0.90 COMMON 7.80 LEGALR

The table reports summary statistics for our sample. Our sample consists of 13,936 bonds issued by 1726 firms over the period 1991–2007. Table 1 provides the definitions and data sources for these variables.

Bonds issued by large firms and more profitable firms have lower credit spreads. Table 6 also reveals that the higher the leverage of the issuer firm, the higher the credit spread. This result is statistically significant at 1% level. Firms that cross-list on U.S. markets benefit from lower credit spreads (around 50 basis points) compared to those that are not cross-listed on U.S. markets. This result is consistent with Ball et al.’s (2013) evidence that foreign firms cross-listed on U.S. markets can lower their offering yield spread by about 48 basis points. Qi et al. (2010) also find that Yankee bonds cross-listed on U.S. markets have a lower credit spreads than non-cross-listed ones. However, they find no difference between cross-listed and non-cross-listed Eurobonds. Table 6 shows that the higher the residual ratings (RATR) and sovereign rating (SOVRAT),2 the lower the credit spreads. Indeed, a unit increase in the sovereign rating decreases the credit spread between 21 and 25 basis points. This result shows that bondholders price the sovereign risk or country risk of the country where the issuing firm is incorporated. Table 6 also shows that firms originating from countries that issue higher proportion of their debt on international markets (INTDB) have a higher cost of debt (between 154 and 183 basis points). This result is consistent with the fact that a higher level of international borrowing tends to correlate negatively with the level of development of the domestic fixed income market. This result, undoubtedly leads to a higher cost of debt for firms operating in these countries. Finally, Table 6 shows that firms that come from countries that have a private or public credit bureau (CRBUR) have a lower cost of debt. The same relationship holds for firm originating from countries with strong enforcement efficiency (ENFOR). This is consistent with the fact that the cost of debt is low in countries where the protection of creditors is high.

2

According to Bissoondoyal-Bheenic (2005), sovereign ratings depend on the country’s economic and financial variables.

12

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

Table 5 Comparison between callable and straight bonds. # of obs.

Mean P-value

All

Callable

Straight

Bond ratings and credit spreads variables 220.92 94.04 CS RAT 12.53 15.71

(0.000)*** (0.000)***

141.88 14.51

5255 5255

8681 8681

13,936 13,936

Firm control variables 8.81 TA 0.70 LEV 0.06 PROFIT 0.25 GROWTH 0.07 USLIST

9.99 0.74 0.06 0.12 0.24

(0.000)*** (0.000)*** (0.640) (0.000)*** (0.000)***

9.55 0.72 0.06 0.17 0.18

5255 5255 5255 5255 5255

8681 8681 8681 8681 8681

13,936 13,936 13,936 13,936 13,936

Bond control variables 0.72 PUB 2.34 AGE 5.29 PROC 0.69 CONV 0.01 SYND

0.93 1.77 4.48 0.50 0.00

(0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)***

0.85 1.99 4.78 0.57 0.00

5255 5255 5255 5255 5255

8681 8681 8681 8681 8681

13,936 13,936 13,936 13,936 13,936

20.75 0.17 1.00 1.40 82.88 0.85 7.76

(0.000)*** (0.000)*** (0.148) (0.000)*** (0.000)*** (0.000)*** (0.000)***

20.80 0.19 1.00 1.29 83.91 0.90 7.80

5255 5255 5254 5254 5254 5254 5250

8681 8681 8659 8659 8659 8659 8680

13,936 13,936 13,913 13,913 13,913 13,913 13,930

Callable

Institutional control variables 20.90 SOVRAT 0.21 INTDB 1.00 CRBUR 1.11 CREDR 85.61 ENFOR 0.97 COMMON LEGALR 7.86

Straight

All

This table presents the mean of the different variables for callable and straight bonds. Our sample consists of 13,936 bonds issued by 1726 firms over the period 1991–2007. Table 1 provides the definitions and data sources for these variables. Differences in the means of the variables between callable and straight bonds are tested using two-tailed t-test of means. P-values of this test are reported in parentheses. * Significance at 10% level. ** Significance at 5% level. *** Significance at 1% level.

5. Sensitivity analyses To check the robustness of our results, we conduct a battery of sensitivity tests. In the first test, we exclude bonds issued by U.S. firms given they account for 10,830 out of 13,936 issues. The results are reported in Table 7. Panel A of Table 7 shows that even after excluding U.S. firms, the callable dummy variable (CALLABLE) still has a positive coefficient (between 101 and 107 basis points), which is statistically significant at 1% level across the different specifications. Panel A of Table 7 also shows that our previous results hold, except for residual ratings (RATR), natural logarithm of total proceeds (PROC), and convertible dummy variable (CONV) that become statistically non-significant. When we exclude U.S. firms, The COMMON dummy variable becomes statistically significant at the 5% level. Indeed, coming from a common law country decreases the cost of debt by 26.23 basis points. In order to avoid the currency effect, we run a separate test by including dollar-denominated bonds and exclude all others. As it is shown in Panel B of Table 7, the CALLABLE variable has a positive coefficient (between 48.79 and 48.9 basis points), which is statistically significant at 1% level across the different specifications. Compared to Table 6, Panel B of Table 7 shows that, except for the dummy variable CONV, the regression coefficients are stable across different model specifications

Table 6 Call feature and corporate bond yield spreads. Predicted sign

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

TA

(−)

LEV

(+)

PROFIT

(−)

GROWTH

(+)

USLIST

(−)

PUB

(−)

RATR

(−)

AGE

(+)

PROC

(+)

CALLABLE

(+)

−25.66*** (−11.64) 101.28*** (6.83) −349.41*** (−5.39) 6.98 (1.62) −10.17 (−1.44) −91.62*** (−13.56) −9.03*** (−5.51) −2.51 (−0.90) 9.73*** (5.72) 53.02*** (13.72)

CONV

(−)

SYND

(−)

−25.80*** (−11.71) 101.47*** (6.89) −349.22*** (−5.38) 7.03 (1.62) −10.76 (−1.49) −89.10*** (−12.82) −8.87*** (−5.57) −1.78 (−0.66) 10.18*** (5.04) 52.93*** (13.91) 111.19 (1.22) −4.49 (−0.89)

SOVRAT

(−)

INTDB

(+)

CRBUR

(−)

−23.28*** (−11.80) 112.86*** (7.69) −320.08*** (−5.29) 6.41 (1.63) −50.45*** (−6.23) −90.17*** (−13.58) −9.84*** (−6.46) −0.21 (−0.09) 9.73*** (5.94) 57.94*** (15.36) 106.27 (1.15) −7.35* (−1.90) −24.07*** (−7.55) 174.26*** (4.64) −132.41*** (−2.61)

−23.49*** (−11.97) 112.39*** (7.66) −317.82*** (−5.22) 6.51* (1.67) −50.91*** (−6.49) −90.27*** (−13.60) −9.79*** (−6.37) 0.07 (0.03) 9.99*** (6.39) 58.16*** (15.57) 106.23 (1.15) −7.84** (−2.09) −21.99*** (−6.68) 157.16*** (3.72) −131.75** (−2.54)

−23.37*** (−11.79) 112.53*** (7.69) −317.86*** (−5.29) 6.44* (1.66) −53.92*** (−6.08) −90.32*** (−13.69) −9.80*** (−6.38) −0.09 (−0.04) 9.82*** (6.12) 58.42*** (15.36) 105.34 (1.14) −7.60** (−2.00) −23.20*** (−7.28) 154.59*** (4.65) −132.63** (−2.59)

−23.24*** (−11.72) 112.90*** (7.72) −318.20*** (−5.30) 6.43* (1.65) −51.68*** (−5.86) −90.48*** (−13.71) −9.85*** (−6.42) −0.15 (−0.06) 9.74*** (6.00) 57.66*** (15.43) 106.19 (1.15) −7.04* (−1.82) −22.71*** (−6.58) 172.06*** (4.79) −136.69*** (−2.64)

CREDR

(−)

−23.32*** (−11.84) 112.56*** (7.70) −319.47*** (−5.27) 6.48* (1.66) −47.85*** (−5.19) −90.18*** (−13.58) −9.84*** (−6.47) −0.24 (−0.10) 9.72*** (5.93) 57.69*** (15.14) 105.66 (1.14) −7.28* (−1.87) −24.14*** (−7.60) 183.82*** (5.05) −136.79*** (−2.71) −3.19 (−0.92)

ENFOR

(−)

COMMON

(−)

LEGALR

(−)

−23.41*** (−12.01) 112.66*** (7.70) −317.24*** (−5.26) 6.51* (1.67) −49.62*** (−3.26) −90.44*** (−13.78) −9.84*** (−6.38) 0.14 (0.06) 10.00*** (6.44) 57.88*** (14.81) 106.28 (1.15) −7.54** (−2.00) −21.58*** (−6.23) 163.11*** (5.61) −134.06** (−2.57) −0.50 (−0.12) −0.60 (−0.73) 5.47 (0.17) −1.21 (−0.33)

−0.55* (−1.82) −15.64 (−1.42) −3.21 (−0.96)

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

Variable

13

14

Variable

Predicted sign

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Intercept

(?) YES

359.89*** (14.87) YES 44.07 13,936

957.71*** (11.58) YES 46.57 13,913

966.87*** (11.69) YES 46.58 13,913

961.97*** (11.66) YES 46.65 13,913

956.45*** (11.56) YES 46.61 13,913

958.75*** (11.40) YES 46.61 13,907

963.76*** (11.57)

INDUSTRY & YEAR EFFECTS Adj. R2 (%) # of obs.

362.85*** (15.12) YES 43.93 13,936

46.67 13,907

This table presents regression estimates of credit spreads on callable variable, residual rating, and firm, bond, and institutional control variables. Industry group dummies (not reported) are based on the two-digit SIC codes following Campbell (1996). Year dummies (not reported) are also included in the estimation. Our sample consists of 13,936 bonds issued by 1726 firms over the period 1991–2007. Table 1 provides the definitions and data sources for these variables. These models are estimated using OLS, correcting for clustering by firm. The associated t-statistics are reported in parentheses. * Significance at 10% level. ** Significance at 5% level. ***

Significance at 1% level.

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

Table 6 (Continued)

Table 7 Sensitivity tests. Variable

Predicted sign

Panel A (1)

(−)

LEV

(+)

PROFIT

(−)

GROWTH

(+)

USLIST

(−)

PUB

(−)

RATR

(−)

AGE

(+)

PROC

(+)

CALLABLE

(+)

CONV

(−)

SYND

(−)

SOVRAT

(−)

INTDB

(+)

CRBUR

(−)

CREDR

(−)

ENFOR

(−)

COMMON

(−)

LEGALR

(−)

−34.11 (−5.97) 106.42*** (4.18) −527.09*** (−2.98) 3.95 (1.41) −41.78*** (−2.96) −43.86*** (−2.85) 0.02 (0.01) −4.16 (−0.57) −3.27 (−1.27) 101.64*** (7.63) −22.34 (−0.26) 3.29 (0.39) −20.65*** (−6.13) 180.22*** (5.83) −122.06** (−2.30) −6.07 (−1.59) ***

(3)

−35.60 (−6.37) 102.15*** (3.92) −519.61*** (−2.87) 3.82 (1.37) −38.71*** (−2.72) −46.66*** (−3.08) −0.95 (−0.26) −1.64 (−0.23) −2.20 (−0.85) 105.16*** (8.21) −24.66 (−0.29) 1.87 (0.22) −17.04*** (−4.86) 137.92*** (3.76) −104.53** (−2.08) ***

(4)

−34.97 (−6.10) 103.24*** (4.17) −522.67*** (−3.05) 3.74 (1.40) −39.70*** (−2.85) −44.60*** (−2.93) −0.79 (−0.22) −2.59 (−0.36) −2.49 (−0.94) 105.45*** (8.01) −27.31 (−0.31) 2.37 (0.28) −19.65*** (−5.71) 139.16*** (4.36) −110.39** (−2.20) ***

(5)

−33.89 (−5.87) 108.92*** (4.27) −537.68*** (−3.08) 3.74 (1.34) −38.80*** (−2.78) −46.11*** (−3.00) −0.61 (−0.17) −3.56 (−0.49) −2.48 (−0.94) 102.05*** (7.76) −22.01 (−0.25) 4.51 (0.54) −20.41*** (−5.56) 167.81*** (5.32) −116.77** (−2.26) ***

−0.92*** (−3.00) −26.23** (−2.39)

−34.96 (−6.19) 106.16*** (4.16) −521.77*** (−2.90) 3.83 (1.34) −37.97*** (−2.65) −48.21*** (−3.14) −1.26 (−0.34) −1.12 (−0.15) −2.05 (−0.82) 106.35*** (7.65) −27.65 (−0.32) 3.06 (0.36) −19.03*** (−5.09) 148.28*** (4.78) −103.51** (−2.08) −4.31 (−0.85) −0.89 (−1.21) −2.38 (−0.08) 5.53 (1.37) ***

(7)

−22.01 (−10.40) 128.12*** (7.65) −263.02*** (−4.48) 6.69 (1.63) −33.95** (−2.29) −78.26*** (−11.45) −14.50*** (−9.07) −0.82 (−0.40) 13.03*** (7.64) 48.80*** (12.93) 89.85 (0.74) −15.13*** (−3.80) −27.65*** (−8.31) 171.26** (2.46) −109.25** (−2.02) −1.74 (−0.38) ***

(8)

−21.95 (−10.36) 128.27*** (7.65) −263.44*** (−4.48) 6.64 (1.61) −34.44*** (−2.65) −78.15*** (−11.45) −14.55*** (−9.21) −0.83 (−0.40) 12.94*** (7.49) 48.88*** (13.03) 89.31 (0.74) −15.03*** (−3.79) −28.15*** (−8.18) 165.68** (2.29) −105.82** (−1.98) ***

(9)

−22.16 (−10.22) 127.98*** (7.65) −261.98*** (−4.49) 6.70* (1.65) −41.20** (−2.28) −78.23*** (−11.51) −14.42*** (−8.99) −0.71 (−0.35) 13.27*** (7.30) 49.07*** (12.94) 90.62 (0.75) −15.37*** (−3.83) −26.84*** (−8.04) 151.44** (2.38) −108.70** (−1.97) ***

(10)

−22.02 (−10.40) 128.19*** (7.67) −262.42*** (−4.49) 6.71* (1.65) −37.31** (−2.31) −78.26*** (−11.57) −14.45*** (−9.00) −0.74 (−0.36) 13.07*** (7.60) 48.79*** (12.97) 90.64 (0.75) −15.12*** (−3.80) −26.53*** (−6.99) 160.45** (2.26) −110.38* (−1.92) ***

0.11 (0.34) −16.64 (−0.88) −3.93 (−0.79)

−21.90*** (−10.42) 128.28*** (7.67) −262.02*** (−4.51) 6.65* (1.66) −43.21* (−1.83) −77.81*** (−11.26) −14.51*** (−9.10) −0.81 (−0.39) 12.98*** (7.51) 48.90*** (13.04) 88.88 (0.73) −14.88*** (−3.74) −29.08*** (−7.21) 133.75*** (2.60) −106.27* (−1.93) 0.17 (0.03) 0.91 (0.88) −39.38 (−0.95) −1.74 (−0.31)

15

−1.58 (−0.54)

(6)

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

TA

Panel B (2)

16

Variable

Predicted sign

Intercept

(?)

INDUSTRY & YEAR EFFECTS Adj. R2 (%) # of obs.

YES

Panel A

Panel B

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

979.92*** (10.09) YES 39.42 3083

976.71*** (10.23) YES 39.95 3083

973.15*** (10.07) YES 39.63 3083

967.91*** (9.89) YES 39.25 3077

977.30*** (10.32)

984.68*** (11.43) YES 51.80 11,572

980.08*** (11.41) YES 51.80 11,572

982.91*** (11.32) YES 51.83 11,572

991.77*** (11.25) YES 51.84 11,568

985.02*** (11.22) YES 51.89 11,568

39.93 3077

This table presents sensitivity tests. In Panel A, we exclude U.S. firms and in Panel B we include only bond issues denominated in $US. Our sample consists of 13,936 bonds issued by 1726 firms over the period 1991–2007. Table 1 provides the definitions and data sources for these variables. These models are estimated using OLS, correcting for clustering by firm. Industry group dummies (not reported) are based on the two-digit SIC codes following Campbell (1996). Year dummies (not reported) are also included in the estimation. The associated t-statistics are reported in parentheses. * Significance at 10% level. ** ***

Significance at 5% level. Significance at 1% level.

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

Table 7 (Continued)

Table 8 Callable yield premium. Variable

Predicted sign

(−)

LEV

(+)

PROFIT

(−)

GROWTH

(+)

USLIST

(−)

PUB

(−)

RATR

(−)

AGE

(+)

PROC

(+)

CALLABLE

(+)

CONV

(−)

SYND

(−)

SOVRAT

(−)

INTDB

(+)

CRBUR

(−)

CREDR

(−)

ENFOR

(−)

COMMON

(−)

LEGALR

(−)

Panel B

Inv. grade

Junk

−8.19 (−5.41) 29.79*** (3.62) −166.90*** (−6.50) 0.54 (0.55) −22.86*** (−4.13) −37.77*** (−8.09) −1.16 (−1.03) 7.70*** (3.84) 2.36 (1.63) 20.75*** (7.38) 784.07*** (3.20) 5.76* (1.67) −4.65** (−2.22) 86.04*** (4.64) −75.46 (−0.10)

−33.52 (−8.02) 76.68*** (3.15) −208.41** (−2.55) 12.19*** (3.80) −21.49 (−0.88) −54.79*** (−5.39) 5.49 (1.63) −53.25*** (−5.77) 12.54*** (2.77) 62.90*** (7.15) −143.13* (−1.82) −32.41*** (−3.50) −16.56*** (−4.82) 385.27*** (2.75) −65.7 (−1.04)

***

***

Junk

−8.49 (−5.63) 30.60*** (3.81) −165.64*** (−6.55) 0.65 (0.63) −17.35*** (−3.28) −38.48*** (−8.28) −1.15 (−1.02) 8.01*** (4.05) 2.71** (2.00) 19.82*** (7.07) 780.42*** (3.18) 5.23 (1.59) −3.60 (−1.55) 91.52*** (5.47) −57.05 (−0.14) 0.61 (0.29) −1.22*** (−5.11) 37.25*** (4.07) −1.14 (−0.50)

−33.14 (−8.11) 76.72*** (3.17) −200.95** (−2.59) 10.83*** (3.21) −49.26 (−1.60) −60.35*** (−6.11) 5.10 (1.49) −51.43*** (−5.76) 11.99*** (2.67) 65.83*** (7.33) −156.34** (−1.97) −29.99*** (−3.35) −12.69*** (−2.60) 236.24*** (2.66) −50.73 (−0.77) 14.29 (1.20) 0.89 (0.81) −130.71** (−2.24) 0.24 (0.03)

***

***

Low leverage

High leverage

Low leverage

High leverage

−22.66 (−10.43) 8.50 (0.37) −226.73*** (−3.95) 2.80 (0.81) −47.57*** (−5.32) −81.06*** (−11.47) −14.27*** (−12.00) −2.12 (−0.76) 10.28*** (5.11) 48.36*** (11.57) 148.28 (0.99) −9.53** (−2.14) −22.18*** (−11.75) 137.26*** (3.82) −113.09** (−2.46)

−23.92 (−8.03) 134.11*** (4.76) −365.24*** (−3.32) 15.91*** (4.01) −46.50*** (−3.84) −94.58*** (−8.44) −5.98*** (−2.88) 3.07 (0.84) 6.48*** (3.01) 65.93*** (10.97) 46.93 (0.53) −4.95 (−0.91) −27.64*** (−2.94) 194.88*** (3.66) −93.13 (−0.13)

−22.41 (−10.38) 6.47 (0.27) −224.85*** (−3.94) 2.78 (0.81) −47.13*** (−3.92) −81.80*** (−11.54) −14.37*** (−12.08) −1.85 (−0.65) 10.09*** (4.99) 48.38*** (11.47) 148.96 (1.00) −8.75** (−1.97) −19.66*** (−7.60) 133.52*** (3.81) −115.65** (−2.47) 0.95 (0.22) −0.36 (−0.72) 5.03 (0.24) −3.08 (−0.78)

−24.34*** (−8.17) 142.95*** (4.95) −354.82*** (−3.23) 16.06*** (4.05) −42.64* (−1.82) −93.73*** (−8.68) −5.95*** (−2.84) 3.59 (0.99) 6.64*** (3.17) 66.10*** (10.27) 41.46 (0.48) −5.33 (−0.98) −25.76*** (−2.67) 186.79*** (4.08) −87.02 (−0.18) −6.37 (−0.97) −0.79 (−0.50) 5.61 (0.09) −0.69 (−0.11)

***

***

***

17

Inv. grade

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

TA

Panel A

18

Variable

Predicted sign

Intercept

(?)

INDUSTRY & YEAR EFFECTS Adj. R2 (%) # of obs.

YES

Panel A

Panel B

Inv. grade

Junk

Inv. grade

Junk

Low leverage

High leverage

Low leverage

High leverage

236.96*** (5.37) YES 33.83 10,689

978.63*** (8.94) YES 29.86 3224

291.63*** (6.46) YES 34.30 10,686

920.82*** (8.26)

945.02*** (15.33) YES 48.68 6950

924.02*** (4.51) YES 47.62 6963

944.10*** (15.28) YES 48.75 6950

952.08*** (4.83)

30.76 3221

47.80 6963

This table presents regression estimates of credit spreads on callable variable, residual rating, and firm, bond, and institutional control variables. Panel A consists of the breakdown of our sample into investment grade (bonds rated BBB and above) and junk bonds (bonds rated below BBB). Panel B breaks down our sample into firms with high leverage and firms with low leverage ratio (compared to the firms’ median leverage). Industry group dummies (not reported) are based on the two-digit SIC codes following Campbell (1996). Year dummies (not reported) are also included in the estimation. Our sample consists of 13,936 bonds issued by 1726 firms over the period 1991–2007. Table 1 provides the definitions and data sources for these variables. These models are estimated using OLS, correcting for clustering by firm. The associated t-statistics are reported in parentheses. * Significance at 10% level. ** ***

Significance at 5% level. Significance at 1% level.

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

Table 8 (Continued)

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

19

6. Callable yield premium To test our two other hypotheses (H2 and H3) discussed above, we break down our sample into two subsamples. The first breakdown consists of investment-grade bonds (bonds that are rated BBB and above) and junk bonds (bonds that are rated below BBB). The second breakdown consists of bonds that are issued by firms that have a leverage ratio lower than the sample’s median leverage and bonds that are issued by firms with a leverage ratio higher than the sample’s median leverage. We run the regression on each of the subsamples defined above and report the results in Table 8. Table 8 shows that our results are consistent with our conjecture in H2. Indeed, we find that junk bonds have a higher callable yield premium than investment grade bonds. We may then infer that junk bonds are more likely to be called back than investment-grade bonds. The call spread is equal to 20.75 (19.82) basis points for investment grade bonds and 62.90 (65.83) basis points for junk bonds in the first (second) specification. The difference is statistically significant at 1% level. This result is consistent with the signaling theory according to which firms can benefit from their bond price appreciation after revealing their positive private information. When we break our sample down into highly leveraged firms and firm with low leverage ratio (compared to the firms’ median leverage), we find that our results are consistent with our third conjecture (H3). The call spread is equal to 48.36 (48.38) basis points for firms with low leverage ratio, compared 65.93 (66.10) for firms with high leverage in the first (second) specification. This difference is statistically significant at 1% level. This result is consistent with the risk-shifting explanation according to which firms with high leverage are more likely to undertake riskier projects and then expropriate bondholders’ wealth. 7. Conclusion In this paper we examine the call spread in a global framework using an international sample and controlling for institutional characteristics. We conjecture that callable bonds have a positive call spread compared to their equivalent non-callable (straight bonds). We also conjecture that the call spread of junk bond is higher than the call spread of investment grade bonds, which is consistent with the signaling hypothesis. We finally conjecture that the call spread of highly leveraged firms is higher than the call spread of firms with low leverage, which is consistent with risk-shifting arguments. Our empirical evidence shows that callable bonds have a positive call spread, which is statistically and economically significant. Our empirical results hold after a battery of robustness checks. We also find that junk callable bonds have a higher call spread than investment-grade callable bonds, which is consistent with the signaling theory. Our empirical results also show that highly leveraged firms have a higher call spread than firms with low leverage, which is consistent with risk-shifting arguments. Our results have implications on the cost of debt and hence on the cost of capital of firms choosing to issue callable bonds. Indeed, these firms pay a positive call premium to bondholders who accept to invest in their callable bonds, thus increasing their cost of borrowing. The extra-yield received by callable bondholders is to compensate these bondholders for the risk that the firm may call back its callable bonds. References Acharya, V., Carpenter, J., 2002. Corporate bond valuation and hedging with stochastic interest rates and endogenous bankruptcy. Rev. Financ. Stud. 15, 1355–1383. Ball, R.T., Hail, L., Vasvari, F.P., 2013. Equity cross-listings in the U.S. and the price of debt. Working paper. University of Chicago. Banko, J., Zhou, L., 2010. Callable bonds revisited. Financ. Manage. 39, 613–641. Barnea, A., Haugen, R., Senbet, L., 1980. A rationale for debt maturity structure and call provisions in the agency theoretic framework. J. Finance 35, 1223–1234. Beck, T., Demirgüc¸-Kunt, A., 2009. Financial institutions and markets across countries and over time: data and analysis. World Bank policy research. Working paper. Berndt, A., 2004. Estimating the term structure of yield spreads from callable corporate bond price data. Working paper. Carnegie Mellon University. Bissoondoyal-Bheenic, E., 2005. An analysis of the determinants of sovereign ratings. Global Finance J. 15, 251–280. Campbell, J., 1996. Understanding risk and return. J. Polit. Econ. 104, 298–345. Chen, Z., Mao, C.X., Wang, Y., 2010. Why firms issue callable bonds: hedging investment uncertainty. J. Corp. Finance 16, 588–607.

20

A. Samet, L. Obay / J. of Multi. Fin. Manag. 25–26 (2014) 1–20

Choi, S., Jameson, M., Jung, M., 2013. The issuance of callable bonds under information asymmetry. J. Empir. Finance 21, 1–14. Djankov, S., Hart, O., McLiesh, C., Shleifer, A., 2008. Debt enforcement around the world. J. Polit. Econ. 116, 1105–1149. Djankov, S., McLiesh, C., Shleifer, A., 2007. Private credit in 129 countries. J. Financ. Econ. 84, 299–329. Duffie, J.D., Singleton, K.J., 1999. Modeling term structure of defaultable bonds. Rev. Financ. Stud. 12, 687–720. Francis, B.B., Hasan, I., John, K., Waisman, M., 2010. The effect of state antitakeover laws on the firm’s bondholders. J. Financ. Econ. 96, 127–154. Gande, A., Parsley, D., 2010. Sovereign credit ratings, transparency and international portfolio flows. Working paper. Southern Methodist University. Güntay, L., Prabhala, N.R., Unal, H., 2004. Callable bonds and hedging. Working paper. University of Maryland. Jarrow, R., Li, H., Liu, S., Wu, C., 2010. Reduced-form valuation of callable corporate bonds: theory and evidence. J. Financ. Econ. 95, 227–248. Jensen, M.C., Meckling, W.H., 1976. Theory of the firm: managerial behavior, agency costs, and ownership structure. J. Financ. Econ. 3, 305–360. Myers, S.C., 1977. Determinants of corporate borrowing. J. Financ. Econ. 5, 147–175. Qi, Y., Roth, L., Wald, J., 2010. Political rights and the cost of debt. J. Financ. Econ. 95, 202–226. Qiu, J., Yu, F., 2010. The market for corporate control and the cost of debt. J. Financ. Econ. 93, 505–524. Robbins, E.H., Schatzberg, J.D., 1986. Callable bonds: a risk-reducing signaling mechanism. J. Finance 41, 935–950. Yu, F., 2005. Accounting transparency and the term structure of credit spreads. J. Financ. Econ. 75, 53–84.