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

Contents lists available at ScienceDirect

Journal of Multinational FinancialManagement

journal homepage: www.elsevier.com/locate/econbase

Review

Call feature and corporate bond yield spreads

Anis Sameta,∗, Lamia Obayb

a School of Business and Management, American University of Sharjah, P.O. Box 26666, Sharjah,United Arab Emiratesb 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 2013Accepted 7 June 2014Available online 14 June 2014

JEL classification:G30G32

Keywords:Credit spreadCallable bondCost of debtCross-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-levelvariables. Using an international sample of 13,936 bonds issuedbetween 1991 and 2007, we find that callable bonds have a pos-itive 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 spreadthan investment-grade callable bonds, which is consistent with thesignaling theory. The empirical results also show that highly lever-aged 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. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22. Literature review and hypotheses development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43. Data and methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3.1. Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53.2. Methodology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3.2.1. Bonds ratings and credit spreads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63.2.2. Control variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

4. Empirical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

∗ Corresponding author. Tel.: +971 6 515 2316; fax: +971 6 515 4065.E-mail addresses: asamet@aus.edu (A. Samet), Lamiaobay@uowdubai.ac.ae (L. Obay).

http://dx.doi.org/10.1016/j.mulfin.2014.06.0041042-444X/© 2014 Elsevier B.V. All rights reserved.

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

4.1. Univariate analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84.2. Multivariate analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

5. Sensitivity analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126. Callable yield premium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

1. Introduction

When issuing a bond, a firm has the choice between issuing a callable bond or a straight bond. A callprovision grants the issuer the right to buy back its already issued bonds prior to the maturity date. Inreturn for the opportunity to call back the bond, the issuer compensates the holder of a callable bondwith an option premium. In other words, a callable bondholder writes a call option and receives thepremium, but bears the risk to re-invest the proceeds at a lower rate should the issuer exercise itsright to call the bond.

The writing of the call option entitles the bondholder to a call premium. Hence, the price thebondholder pays for a callable bond is always lower than that of an equivalent straight bond. Morespecifically, the price of a callable bond is equal to the price of an equivalent straight bond minus theprice 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 reinvestmentrisk, 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 callablebonds and callable bond pricing. When issuing callable bonds, firms seek to: (1) hedge their interestrate risk (Güntay et al., 2004), (2) hedge investment risk (Chen et al., 2010), (3) benefit from theirfuture 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 anew reduced-form approach to value callable corporate bonds, which, according to them, fits callablebond 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 ofleverage. We define the call spread as the yield component that is due to the call provision aftercontrolling for bond-, firm-, and country-specific variables. To the best of our knowledge, no paperhas 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 inU.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 denomi-nation of the treasury security, used as a benchmark, to that of the bond for which the spread is beingcomputed.

The aim of our study is, therefore, to quantify the call spread in a global context and to compare thecall spread between high-rated and low-rated bonds and between bonds issued by high-leveragedfirms 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 spreadspecifications or include it in their robustness check analysis. Qiu and Yu (2010), using U.S. bonds issuedbetween 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 bondsissued by U.S. firms. Qi et al. (2010) look at only Eurobonds denominated in U.S. dollars. Ball et al. (2013) compute credit spreadusing 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 39countries between 1980 and 2006 and find that callable bonds have higher spread than straight bondswhen they include a dummy variable for callable bonds in three of their model specifications. Franciset al. (2010), using U.S. bonds issued between 1990 and 2000, find that the callable dummy variable isstatistically significant. Ball et al. (2013), using a global sample of public bonds but retaining only onebond per firm (the bond with the largest principal amount), find that the callable dummy is significantin only one of their specifications. It, however, becomes statistically not significant when they controlfor bond-specific variables. Ball et al. (2013) compute the credit spread as the yield to maturity of thebond minus the yield of an equivalent U.S. treasury security, independently of the currency in whichthe bond is denominated. We circumvent this shortcoming by matching the currency of denominationof 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 theirwillingness to hold the callable bond that would not otherwise be offered on equivalent non-callablebonds. We further hypothesize that firms with higher leverage have a higher call spread, all elseequal. According to the risk-shifting hypothesis, first introduced by Jensen and Meckling (1976) andlater developed by Barnea et al. (1980) and Chen et al. (2010), shareholders have an incentive toexpropriate bondholders’ wealth. Accordingly, we further assume that holders of callable bonds willrequire a higher call spread for highly leveraged firms to compensate them for the risk of wealthexpropriation.

Based on the signaling theory (Chen et al., 2010; Robbins and Schatzberg, 1986), we hypothesizethat junk bonds have a higher call spread than investment-grade bonds, all else being equal. In otherwords, we expect firms that issue junk bonds to have a higher incentive to exercise this option andcall back their bonds when they reveal their positive private information.

The univariate empirical results show that, on average, callable bonds have a higher credit spreadthan straight bonds. Moreover, callable bonds have lower ratings than straight bonds. The univariateresults further reveal that callable bonds are issued by smaller firms, firms with lower leverage, firmswith 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 havea positive call spread which is statistically and economically significant across all specifications. Wefind that the call feature adds between 52 and 58 basis points to the cost of debt, everything elsebeing equal. In other words, the call option premium paid by the callable bond issuer and received bythe callable bondholders is between 52 and 58 basis points. We also find that bonds issued by largefirms, firms with low leverage, more profitable firms, firms with higher residual ratings, and firmscross-listed on U.S. markets have lower credit spreads. Those results are consistent with the literatureon the determinants of the credit spread (e.g., Yu, 2005; Qi et al., 2010). In addition, we find that thecountry of incorporation has a significant effect on the credit spread. Indeed, issuing firms comingfrom countries with low sovereign ratings, firms coming from countries that issue a higher proportionof their debt on international markets, and firms that come from countries with no credit bureau, incura higher cost of debt.

The empirical results further reveal that junk bonds have a higher call spread than investment gradebonds, 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 privatepositive information. When testing for the effect of leverage we find that highly leveraged firms havea 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 riskierprojects. 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 quantifyin basis points the extra cost bond issuers need to bear for issuing callable bonds. This has directimplications on the yield for holding callable bonds as well as the firm’s cost of debt, hence its cost ofcapital. Second, we provide further empirical support for the signaling and risk-shifting hypothesesunderlying 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 ourhypotheses. Section 3 introduces our data and methodology. In Section 4, we report the univariateand 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 thelast section.

2. Literature review and hypotheses development

The callable bond literature has developed around two main strands: firm’s motivation for issuingcallable bonds and callable bond pricing. The most commonly cited explanation for issuing a callablebond is to hedge interest-rate risk. Indeed, firms issue callable bonds to call them back and refinancethem at lower costs when interest rates fall. As such, callable bonds are more likely to be issuedduring times of high interest rates. Firms are more likely to include a call provision in the case oflonger maturity bonds given their greater sensitivity to interest rate fluctuations. They are also morelikely to attach a call provision for larger issues because of “more material interest rate exposures thatare 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 callfeature alleviates the risk-shifting problem first introduced by Jensen and Meckling (1976) and givesthe firm the flexibility to liquidate a project that subsequently reveals to have a negative Net PresentValue. Chen et al. (2010) find strong evidence for the hedging of investment risk hypothesis. Thethird 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 callablebonds at lower prices and then capture the price appreciation and benefit from the option to refinancetheir 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 fourthexplanation is risk shifting according to which shareholders can expropriate bondholders’ wealthby shifting into riskier assets. According to Barnea et al. (1980), the value of the call option, ownedby shareholders, declines should there be an increase in firm’s risk, which reduces the incentives toexpropriate bondholders. The last explanation is the underinvestment problem where firms wouldnot invest in positive NPV projects as existing bondholders will take part in the benefits of investingin 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, anddefault 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 sim-ply control for the callable feature in some of their credit spread specifications or use it as robustnesscheck. Francis et al. (2010), using U.S. bonds issued between 1990 and 2000, find that the callabledummy variable is statistically significant. Qiu and Yu (2010), using U.S. bonds issued between 1976and 1991, find that callable dummy is positive and statistically significant in their robustness tests. Qiet al. (2010), using Eurobonds denominated in U.S. dollars and issued by borrowers incorporated in39 countries between 1980 and 2006, find that callable bonds have higher spread than straight bondsin three of their specifications. Ball et al. (2013), using a global sample of public bonds but retainingonly one bond per firm (the bond with the largest principal amount), find that the callable dummyis significant in only one of their specifications and becomes statistically not significant when theycontrol for bond-specific variables. In their paper, the credit spread is being computed as the yield tomaturity of the bond minus the yield of an equivalent U.S. treasury security, irrespective of the cur-rency of denomination of the bond. We circumvent this shortcoming by identifying treasury securitieswith similar currency denomination for the computation of the credit spread.

We believe our study makes an important contribution to the literature by quantifying the callspread for a global sample of 13,936 bonds issued between 1991 and 2007 originating from 36 differ-ent countries. To the best of our knowledge, no paper has empirically focused on the offering call spreadthat issuers offer to callable bondholders in a global framework and using bonds denominated in dif-ferent currencies. Our study is the first study to quantify the call spread in an international frameworkand compares the call spread between high-rated and low-rated bonds and between bonds issued byhigh-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 formulateour 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. Theprice of a callable bond is equal to the value of an equivalent straight bond less the value of the callprovision. Given the inverse relationship between bond prices and bond yields, the yield of a callablebond is, therefore, higher than that of an equivalent straight bond. Investors expect to be compensatedfor 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 economicallysignificant 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 firmsissue callable bonds in order to benefit from information asymmetry. Firms may exercise their calloption when their positive private information is revealed (i.e., signaling theory), when interest ratesgo 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 thatissue junk bonds, when they reveal their positive private information as there is no substantial gainfrom doing so. Recent studies (Banko and Zhou, 2010) show that below-investment-grade bonds aremainly issued to alleviate agency conflicts, while investment grade bonds are mainly issued to hedgeinterest rate risks. We expect that the callable dummy variable’s coefficient is strictly higher for junkbonds 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 toexpropriate bondholders’ wealth by shifting into riskier assets when the financial health of the firmdeteriorates. The call provision may mitigate such behavior. Indeed, when firms undertake riskierprojects, the price of the call option held by the shareholders falls, which in turn reduces the incentivesto shift into riskier asset and expropriate bondholders wealth. Highly leveraged firms are more likelyto undertake riskier projects. The issuance of callable bonds reduces the incentives to expropriatebondholders. To compensate for the risk of expropriation, we conjecture that bondholders require ahigher call spread for highly leveraged firms. We expect that the callable dummy variable’s coefficientis strictly higher for bonds issued by highly leveraged firms than those issued by firms with lowleverage.

The hypotheses above are tested in an international context using a sample of 13,936 bonds issuedbetween 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 worldbetween 1991 and 2007. We stop in 2007 to avoid the impact of the financial crisis. We excludegovernment bonds, sovereign agency issues, bonds issued by financial firms, floating rate bonds, andthose with missing values. From SDC, we extract bonds’ characteristics (call provision, coupon rate(fixed vs. floating), maturity, bond rating, convertibility provision, syndication, offering market (publicvs. 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 differentcountries, currencies, and maturities using Thomson Reuters DataStream, Bloomberg, and issuingfirms’ 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 legalrights index are from different sources. Table 1 defines the different variables and presents theirsources. 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 spreadsIn 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 variablein the credit spread regression in order to avoid any potential simultaneity between credit spreadsand ratings. We use the S&P bonds’ ratings reported in SDC. We convert the letter rating into a numberbetween 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 basedon the currency denomination of the bond and its maturity. For bonds issued in Euro by firms fromEuropean countries, those bonds are matched with the respective Euro government bond with thesame maturity. For instance a French bond issued in Euros is matched with the French governmentbond with the same maturity. Moreover, bonds issued before the adoption of the Euro are matchedwith their respective government bonds. If the maturity of the corporate bond and the governmentbond do not match, we use linear interpolation between the two closest government bond maturitiesto match the corporate bond yield to maturity with the government bond yield to maturity. Thedependent variable CS is then computed using the difference between the corporate bond yield tomaturity and the equivalent government bond yield to maturity.

3.2.2. Control variablesWe 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 whichis equal to EBITA over total assets. GROWTH: Annual total assets’ growth rate. USLIST: A dummyvariable which is equal to one if the firm is cross-listed on U.S. markets, zero otherwise. Data areobtained 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 publicmarkets and zero otherwise. AGE: The natural logarithm of the bond’s age, which is also equal tonatural logarithm of the bond’s maturity. PROC: Is equal to the natural logarithm of the total proceedsin million U.S. dollars. CALLABLE: A dummy variable which is equal to one if the bond is a callablebond and zero otherwise. CONV: A dummy variable which is equal to one if the bond is a convertiblebond and zero otherwise. SYND: A dummy variable which is equal to one if the bond is syndicatedand zero otherwise. COUPON: The bond’s coupon rate.

• Institutional control variables: SOVRAT: Standard & Poor’s letter sovereign credit ratings convertedinto numbers ranging from 22 (AAA with positive outlook) to zero (C with negative outlook) follow-ing Gande and Parsley (2010). INTDB: International Debt Securities (outstanding amount) as a shareof GDP. CRBUR: The variable equals one if either a public registry or a private bureau operates in thecountry, 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 1Variables, definitions, and sources.

Variable Definition Source

Panel A: credit spreads and bond ratings variablesCS The credit spreads. Authors’

specificationRAT 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.Securities DataCompany

RATR The residuals from regressing bond ratings on control variables. Authors’specification

Panel B: control variablesFirm control variablesTA The natural logarithm of total assets in million U.S. dollars. CompustatLEV The leverage ratio which is equal to total liabilities over total assets. CompustatPROFIT The profitability ratio which is equal to EBITA over total assets. CompustatGROWTH The annual total assets’ growth rate. CompustatUSLIST A dummy variable which is equal to one if the firm is cross-listed on U.S. markets, zero

otherwise.NYSE, NASDAQ,JPMorgan, Bank ofNew York, andCitigroup websites

Bond control variablesPUB A dummy variable which is equal to one if the bond is issued on public markets, zero

otherwise.Securities DataCompany

AGE The natural logarithm of the bond’s maturity. Securities DataCompany

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

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

CONV A dummy variable which is equal to one if the bond is a convertible bond, zerootherwise.

Securities DataCompany

SYND A dummy variable which is equal to one if the bond is syndicated, zero otherwise. Securities DataCompany

COUPON The coupon rate. Securities DataCompany

Institutional control variablesSOVRAT Standard & Poors letter sovereign credit ratings converted into numbers ranging from

22 (AAA with positive outlook) to zero (C with negative outlook) following Gande andParsley (2010).

Standard andPoor’s website andGande and Parsley(2010)

INTDB International Debt Securities (outstanding amount) as a share of GDP Beck andDemirgüc -Kunt(2009), World Bankwebsite

CRBUR The variable equals one if either a public registry or a private bureau operates in thecountry, zero otherwise. A public registry is defined as a database owned by publicauthorities (usually the Central Bank or Banking Supervisory Authority) that collectsinformation on the standing of borrowers in the financial system and makes itavailable to financial institutions. A private bureau is defined as a private commercialfirm or non-profit organization that maintains a database on the standing ofborrowers in the financial system, and its primary role is to facilitate exchange ofinformation amongst banks and financial institutions.

Djankov et al.(2008)

CREDR The creditor rights index of the ultimate parent nation. The index ranges from zero(weak creditor rights) to four (strong creditor rights)

Djankov et al.(2008)

ENFOR The debt enforcement efficiency index of the ultimate parent nation. Djankov et al.(2008)

COMMON A dummy variable which is equal to one if the ultimate parent nation is a common lawcountry, zero otherwise.

Djankov et al.(2008)

LEGALR Measures the degree to which collateral and bankruptcy laws protect the rights ofborrowers and lenders and thus facilitate lending. The index ranges from zero to ten,with higher scores indicating that collateral and bankruptcy laws are better designedto expand access to credit. The index is taken for 2007.

Djankov et al.(2007) and doingbusiness 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 enforcementefficiency index of the ultimate parent nation. This index ranges from zero (weak enforcement) to100 (strong enforcement). COMMON: A dummy variable which is equal to one if the ultimate parentnation is a common law country, zero otherwise. LEGALR: Legal rights index. The index ranges fromzero to ten, with higher scores indicating that collateral and bankruptcy laws are better designed toexpand access to credit.

4. Empirical results

In this section, we present the empirical results. First, we start with the univariate analysis andthen 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 sampleis 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 statisticsof the main variables used in our regressions. Table 2 shows that in average 37.7% of the bonds issuedwere callable bonds. U.S. firms issued 10,830 bonds all over the world, of which 43.1% were callablebonds.

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 thatJapanese 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 dummyvariable (CALLABLE) is positively correlated (39%) with the credit spread (CS). Moreover, this tableshows 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 theaverage credit spread is 141.88 basis points and the standard deviation is 157.60 basis points. Thecredit 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 averagea higher credit spread (220.92 basis points) than straight bonds (94.04 basis points), giving furthersupport 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 growthratio, 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 between52 and 58 basis points. This result is significant across the different specifications at the 1% level. Thisresult 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 straightbonds issued all over the world and in different currencies. Indeed, when a firm’s manager chooses toissue a callable bond, the relative cost of debt increases between 52 and 58 basis points, everything elsebeing equal. In other words, the call option premium paid by the callable bond issuer and receivedby the callable bondholders is between 52 and 58 basis points. The positive and statistically andeconomically significant call premium is consistent with the literature on credit spread. Indeed, Qiuand Yu (2010) using U.S. bonds issued between 1985 and 1991, find that the callable dummy variableis positive and statistically significant. Qi et al. (2010), using Eurobonds denominated in U.S. dollarsand issued by borrowers incorporated in 39 countries between 1980 and 2006, find also that callablebonds have higher spread than straight bonds. Francis et al. (2010), using U.S. bonds issued between1990 and 2000, find that the callable dummy variable is statistically significant.

A.

Samet,

L. O

bay /

J. of

Multi.

Fin. M

anag. 25–26

(2014) 1–20

9

Table 2Descriptive statistics by country.

Country CS TA LEV (%) PROFIT (%) GROWTH (%) USLIST (%) PUB RAT PROC AGE CALLABLE (%) # of obs.

Argentina 309.5 8.8 44.6 8.6 8.0 83.3 0.9 12.2 2.1 5.3 16.7 12Australia 128.2 8.1 56.9 7.6 14.1 79.3 0.6 14.2 2.2 4.8 15.2 92Austria 75.8 9.1 65.0 3.4 −2.0 100.0 1.0 13.0 2.1 6.6 0.0 3Belgium 97.2 7.7 55.8 6.8 4.3 83.3 0.3 14.7 2.1 4.4 0.0 6Brazil 521.1 8.8 70.1 1.7 49.9 70.4 0.9 9.7 1.6 5.1 14.8 27Canada 238.8 8.2 63.6 4.8 24.1 55.0 0.7 11.5 2.4 5.5 71.2 333Chile 166.8 8.1 50.5 7.4 14.9 26.7 0.8 14.1 2.2 5.3 0.0 15China 312.8 6.4 61.1 6.4 58.8 0.0 1.0 11.0 1.9 5.3 0.0 1Denmark 88.9 7.2 64.4 6.6 1.8 0.0 0.3 15.5 2.2 5.2 0.0 4Finland 118.7 7.6 49.6 7.8 −42.5 100.0 1.0 14.0 1.8 5.8 0.0 2France 130.4 10.3 75.7 4.2 −1.3 82.4 1.0 14.3 1.8 5.8 8.3 278Germany 111.9 11.5 77.8 1.9 10.6 97.3 1.0 15.4 1.3 3.9 8.1 669Greece 360.4 7.7 60.2 7.9 148.5 33.3 0.3 10.2 2.1 5.2 66.7 9Hong Kong 214.1 9.4 54.4 7.4 −17.2 71.4 0.8 13.9 2.2 5.9 19.0 42Hungary 128.4 9.0 55.1 16.1 6.7 100.0 1.0 12.0 1.6 6.8 0.0 1India 187.8 8.5 47.2 9.3 25.2 100.0 0.7 11.5 2.5 5.0 0.0 12Indonesia 401.4 4.7 55.2 14.5 4.3 33.3 0.8 9.2 1.7 5.1 50.0 6Italy 127.9 11.0 78.3 3.3 0.5 77.6 1.0 14.2 1.9 5.7 14.1 85Japan 38.4 11.0 76.6 2.5 1.9 95.8 1.0 17.4 1.8 4.9 6.1 575Luxembourg 160.0 8.6 76.1 1.4 3.5 54.5 0.9 12.8 1.9 5.6 9.1 11Malaysia 197.9 8.2 54.8 8.0 110.4 25.0 1.0 13.4 2.4 5.8 0.0 8Mexico 313.7 8.2 58.4 7.4 16.8 70.4 0.8 12.1 2.1 5.2 48.1 27Netherlands 253.7 9.2 73.4 2.2 37.8 75.3 0.9 13.1 2.1 6.0 29.9 77New Zealand 84.2 7.8 64.8 10.0 24.8 65.0 0.9 15.8 2.2 4.9 0.0 20Norway 226.1 9.0 63.6 11.4 57.8 16.7 0.9 13.6 2.2 5.1 44.4 36Philippines 390.8 8.1 59.2 5.1 16.3 14.3 1.0 9.7 2.1 5.1 7.1 14Portugal 61.1 9.1 68.6 7.2 −8.0 87.5 1.0 15.5 1.9 6.3 0.0 8Singapore 378.9 7.1 40.9 −4.2 32.4 100.0 0.6 7.4 2.1 5.3 80.0 5South Africa 433.9 7.2 54.5 6.2 8.6 60.0 0.8 9.2 2.4 5.7 100.0 5Spain 85.6 10.3 73.3 5.2 79.8 64.9 0.9 16.2 2.0 5.9 3.9 77Sweden 146.7 8.6 55.1 8.9 24.0 95.0 0.5 14.2 2.2 5.5 25.0 20Switzerland 111.2 10.0 65.6 7.5 9.6 90.3 1.0 16.9 1.8 5.4 9.7 62Taiwan 152.3 7.5 55.6 13.7 40.3 0.0 1.0 13.0 2.3 5.8 0.0 1Thailand 415.6 7.3 51.0 5.5 −11.5 0.0 1.0 8.0 2.1 5.5 50.0 2UK 123.4 9.6 68.1 8.3 43.3 65.1 0.9 15.2 1.9 5.6 20.0 561United States 144.5 9.4 72.4 6.0 16.4 0.0 0.8 14.4 2.0 4.7 43.1 10,830Total 141.9 9.5 72.1 5.7 17.0 17.6 0.8 14.5 2.0 4.8 37.7 13,936

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 providesthe definitions and data sources for these variables.

10

A.

Samet,

L. O

bay /

J. of

Multi.

Fin. M

anag. 25–26

(2014) 1–20

Table 3Correlation coefficients.

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

CS 1.00TA −0.39 1.00LEV 0.05 0.30 1.00PROFIT −0.20 −0.12 −0.34 1.00GROWTH 0.10 −0.07 −0.10 −0.03 1.00USLIST −0.10 0.25 0.01 −0.07 −0.01 1.00PUB −0.44 0.32 0.05 0.06 −0.09 0.10 1.00RAT −0.70 0.57 −0.01 0.22 −0.09 0.12 0.46 1.00AGE 0.11 −0.30 −0.17 0.05 0.02 −0.14 −0.09 −0.17 1.00PROC 0.16 −0.01 −0.07 0.01 0.04 0.08 −0.18 −0.17 0.25 1.00CALLABLE 0.39 −0.32 −0.12 0.00 0.06 −0.21 −0.29 −0.41 0.37 0.27 1.00CONV 0.10 −0.03 0.00 −0.03 0.00 0.03 −0.06 −0.06 −0.03 0.04 0.04 1.00SYND 0.01 −0.04 −0.08 0.02 −0.02 0.03 0.20 −0.09 0.29 0.43 0.19 −0.02 1.00SOVRAT −0.10 0.03 0.07 0.01 0.00 −0.24 −0.02 0.06 0.00 −0.07 0.06 0.01 −0.03 1.00INTDB 0.13 0.17 −0.02 −0.01 0.00 0.44 −0.06 −0.08 −0.03 0.21 0.12 0.06 0.21 0.01 1.00CRBUR −0.04 0.03 0.04 0.02 0.00 −0.03 0.01 0.04 0.00 −0.01 −0.01 0.00 0.01 0.05 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 11

Table 4Descriptive statistics of explanatory variables.

Mean Median Minimum Maximum Std. dev. # of obs.

Credit spreads and bond ratings variablesCS 141.88 86.20 0.02 2366.50 157.60 13,936RAT 14.51 15.00 1.00 21.00 3.76 13,936

Firm control variablesTA 9.55 9.53 −2.14 13.45 1.81 13,936LEV 0.72 0.72 0.00 2.96 0.17 13,936PROFIT 0.06 0.05 −1.55 1.79 0.07 13,936GROWTH 0.17 0.07 −0.89 41.09 0.97 13,936USLIST 0.18 0.00 0.00 1.00 0.38 13,936

Bond control variablesPUB 0.85 1.00 0.00 1.00 0.36 13,936AGE 1.99 2.08 0.00 3.66 0.75 13,936PROC 4.78 5.16 −2.30 8.42 1.47 13,936CALLABLE 0.38 0.00 0.00 1.00 0.48 13,936CONV 0.00 0.00 0.00 1.00 0.05 13,936SYND 0.57 1.00 0.00 1.00 0.50 13,936

Institutional control variablesSOVRAT 20.80 21.00 3.00 21.00 1.08 13,936INTDB 0.19 0.15 0.01 1.93 0.16 13,936CRBUR 1.00 1.00 0.00 1.00 0.02 13,913CREDR 1.29 1.00 0.00 4.00 0.80 13,913ENFOR 83.91 85.80 13.40 96.10 10.04 13,913COMMON 0.90 1.00 0.00 1.00 0.30 13,913LEGALR 7.80 8.00 3.00 10.00 0.79 13,930

The table reports summary statistics for our sample. Our sample consists of 13,936 bonds issued by 1726 firms over the period1991–2007. 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 revealsthat the higher the leverage of the issuer firm, the higher the credit spread. This result is statisticallysignificant at 1% level. Firms that cross-list on U.S. markets benefit from lower credit spreads (around50 basis points) compared to those that are not cross-listed on U.S. markets. This result is consistentwith Ball et al.’s (2013) evidence that foreign firms cross-listed on U.S. markets can lower their offeringyield 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 differencebetween cross-listed and non-cross-listed Eurobonds.

Table 6 shows that the higher the residual ratings (RATR) and sovereign rating (SOVRAT),2 the lowerthe credit spreads. Indeed, a unit increase in the sovereign rating decreases the credit spread between21 and 25 basis points. This result shows that bondholders price the sovereign risk or country risk ofthe country where the issuing firm is incorporated.

Table 6 also shows that firms originating from countries that issue higher proportion of their debt oninternational markets (INTDB) have a higher cost of debt (between 154 and 183 basis points). This resultis consistent with the fact that a higher level of international borrowing tends to correlate negativelywith the level of development of the domestic fixed income market. This result, undoubtedly leads toa 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 withstrong enforcement efficiency (ENFOR). This is consistent with the fact that the cost of debt is low incountries 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 5Comparison between callable and straight bonds.

Mean # of obs.

Callable Straight P-value All Callable Straight All

Bond ratings and credit spreads variablesCS 220.92 94.04 (0.000)*** 141.88 5255 8681 13,936RAT 12.53 15.71 (0.000)*** 14.51 5255 8681 13,936

Firm control variablesTA 8.81 9.99 (0.000)*** 9.55 5255 8681 13,936LEV 0.70 0.74 (0.000)*** 0.72 5255 8681 13,936PROFIT 0.06 0.06 (0.640) 0.06 5255 8681 13,936GROWTH 0.25 0.12 (0.000)*** 0.17 5255 8681 13,936USLIST 0.07 0.24 (0.000)*** 0.18 5255 8681 13,936

Bond control variablesPUB 0.72 0.93 (0.000)*** 0.85 5255 8681 13,936AGE 2.34 1.77 (0.000)*** 1.99 5255 8681 13,936PROC 5.29 4.48 (0.000)*** 4.78 5255 8681 13,936CONV 0.69 0.50 (0.000)*** 0.57 5255 8681 13,936SYND 0.01 0.00 (0.000)*** 0.00 5255 8681 13,936

Institutional control variablesSOVRAT 20.90 20.75 (0.000)*** 20.80 5255 8681 13,936INTDB 0.21 0.17 (0.000)*** 0.19 5255 8681 13,936CRBUR 1.00 1.00 (0.148) 1.00 5254 8659 13,913CREDR 1.11 1.40 (0.000)*** 1.29 5254 8659 13,913ENFOR 85.61 82.88 (0.000)*** 83.91 5254 8659 13,913COMMON 0.97 0.85 (0.000)*** 0.90 5254 8659 13,913LEGALR 7.86 7.76 (0.000)*** 7.80 5250 8680 13,930

This table presents the mean of the different variables for callable and straight bonds. Our sample consists of 13,936 bonds issuedby 1726 firms over the period 1991–2007. Table 1 provides the definitions and data sources for these variables. Differences inthe means of the variables between callable and straight bonds are tested using two-tailed t-test of means. P-values of this testare 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 resultsare reported in Table 7. Panel A of Table 7 shows that even after excluding U.S. firms, the callabledummy variable (CALLABLE) still has a positive coefficient (between 101 and 107 basis points), whichis 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 becomestatistically non-significant.

When we exclude U.S. firms, The COMMON dummy variable becomes statistically significant at the5% 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-denominatedbonds and exclude all others. As it is shown in Panel B of Table 7, the CALLABLE variable has a positivecoefficient (between 48.79 and 48.9 basis points), which is statistically significant at 1% level acrossthe different specifications.

Compared to Table 6, Panel B of Table 7 shows that, except for the dummy variable CONV, theregression coefficients are stable across different model specifications

A.

Samet,

L. O

bay /

J. of

Multi.

Fin. M

anag. 25–26

(2014) 1–20

13

Table 6Call feature and corporate bond yield spreads.

Variable Predicted sign (1) (2) (3) (4) (5) (6) (7) (8)

TA (−) −25.66*** −25.80*** −23.28*** −23.32*** −23.49*** −23.37*** −23.24*** −23.41***

(−11.64) (−11.71) (−11.80) (−11.84) (−11.97) (−11.79) (−11.72) (−12.01)LEV (+) 101.28*** 101.47*** 112.86*** 112.56*** 112.39*** 112.53*** 112.90*** 112.66***

(6.83) (6.89) (7.69) (7.70) (7.66) (7.69) (7.72) (7.70)PROFIT (−) −349.41*** −349.22*** −320.08*** −319.47*** −317.82*** −317.86*** −318.20*** −317.24***

(−5.39) (−5.38) (−5.29) (−5.27) (−5.22) (−5.29) (−5.30) (−5.26)GROWTH (+) 6.98 7.03 6.41 6.48* 6.51* 6.44* 6.43* 6.51*

(1.62) (1.62) (1.63) (1.66) (1.67) (1.66) (1.65) (1.67)USLIST (−) −10.17 −10.76 −50.45*** −47.85*** −50.91*** −53.92*** −51.68*** −49.62***

(−1.44) (−1.49) (−6.23) (−5.19) (−6.49) (−6.08) (−5.86) (−3.26)PUB (−) −91.62*** −89.10*** −90.17*** −90.18*** −90.27*** −90.32*** −90.48*** −90.44***

(−13.56) (−12.82) (−13.58) (−13.58) (−13.60) (−13.69) (−13.71) (−13.78)RATR (−) −9.03*** −8.87*** −9.84*** −9.84*** −9.79*** −9.80*** −9.85*** −9.84***

(−5.51) (−5.57) (−6.46) (−6.47) (−6.37) (−6.38) (−6.42) (−6.38)AGE (+) −2.51 −1.78 −0.21 −0.24 0.07 −0.09 −0.15 0.14

(−0.90) (−0.66) (−0.09) (−0.10) (0.03) (−0.04) (−0.06) (0.06)PROC (+) 9.73*** 10.18*** 9.73*** 9.72*** 9.99*** 9.82*** 9.74*** 10.00***

(5.72) (5.04) (5.94) (5.93) (6.39) (6.12) (6.00) (6.44)CALLABLE (+) 53.02*** 52.93*** 57.94*** 57.69*** 58.16*** 58.42*** 57.66*** 57.88***

(13.72) (13.91) (15.36) (15.14) (15.57) (15.36) (15.43) (14.81)CONV (−) 111.19 106.27 105.66 106.23 105.34 106.19 106.28

(1.22) (1.15) (1.14) (1.15) (1.14) (1.15) (1.15)SYND (−) −4.49 −7.35* −7.28* −7.84** −7.60** −7.04* −7.54**

(−0.89) (−1.90) (−1.87) (−2.09) (−2.00) (−1.82) (−2.00)SOVRAT (−) −24.07*** −24.14*** −21.99*** −23.20*** −22.71*** −21.58***

(−7.55) (−7.60) (−6.68) (−7.28) (−6.58) (−6.23)INTDB (+) 174.26*** 183.82*** 157.16*** 154.59*** 172.06*** 163.11***

(4.64) (5.05) (3.72) (4.65) (4.79) (5.61)CRBUR (−) −132.41*** −136.79*** −131.75** −132.63** −136.69*** −134.06**

(−2.61) (−2.71) (−2.54) (−2.59) (−2.64) (−2.57)CREDR (−) −3.19 −0.50

(−0.92) (−0.12)ENFOR (−) −0.55* −0.60

(−1.82) (−0.73)COMMON (−) −15.64 5.47

(−1.42) (0.17)LEGALR (−) −3.21 −1.21

(−0.96) (−0.33)

14

A.

Samet,

L. O

bay /

J. of

Multi.

Fin. M

anag. 25–26

(2014) 1–20

Table 6 (Continued)

Variable Predicted sign (1) (2) (3) (4) (5) (6) (7) (8)

Intercept (?) 362.85*** 359.89*** 957.71*** 966.87*** 961.97*** 956.45*** 958.75*** 963.76***

(15.12) (14.87) (11.58) (11.69) (11.66) (11.56) (11.40) (11.57)INDUSTRY & YEAR EFFECTS YES YES YES YES YES YES YES YESAdj. R2 (%) 43.93 44.07 46.57 46.58 46.65 46.61 46.61 46.67# of obs. 13,936 13,936 13,913 13,913 13,913 13,913 13,907 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 firmsover 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 associatedt-statistics are reported in parentheses.

* Significance at 10% level.** Significance at 5% level.

*** Significance at 1% level.

A.

Samet,

L. O

bay /

J. of

Multi.

Fin. M

anag. 25–26

(2014) 1–20

15

Table 7Sensitivity tests.

Variable Predicted sign Panel A Panel B

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

TA (−) −34.11*** −35.60*** −34.97*** −33.89*** −34.96*** −22.01*** −21.95*** −22.16*** −22.02*** −21.90***

(−5.97) (−6.37) (−6.10) (−5.87) (−6.19) (−10.40) (−10.36) (−10.22) (−10.40) (−10.42)LEV (+) 106.42*** 102.15*** 103.24*** 108.92*** 106.16*** 128.12*** 128.27*** 127.98*** 128.19*** 128.28***

(4.18) (3.92) (4.17) (4.27) (4.16) (7.65) (7.65) (7.65) (7.67) (7.67)PROFIT (−) −527.09*** −519.61*** −522.67*** −537.68*** −521.77*** −263.02*** −263.44*** −261.98*** −262.42*** −262.02***

(−2.98) (−2.87) (−3.05) (−3.08) (−2.90) (−4.48) (−4.48) (−4.49) (−4.49) (−4.51)GROWTH (+) 3.95 3.82 3.74 3.74 3.83 6.69 6.64 6.70* 6.71* 6.65*

(1.41) (1.37) (1.40) (1.34) (1.34) (1.63) (1.61) (1.65) (1.65) (1.66)USLIST (−) −41.78*** −38.71*** −39.70*** −38.80*** −37.97*** −33.95** −34.44*** −41.20** −37.31** −43.21*

(−2.96) (−2.72) (−2.85) (−2.78) (−2.65) (−2.29) (−2.65) (−2.28) (−2.31) (−1.83)PUB (−) −43.86*** −46.66*** −44.60*** −46.11*** −48.21*** −78.26*** −78.15*** −78.23*** −78.26*** −77.81***

(−2.85) (−3.08) (−2.93) (−3.00) (−3.14) (−11.45) (−11.45) (−11.51) (−11.57) (−11.26)RATR (−) 0.02 −0.95 −0.79 −0.61 −1.26 −14.50*** −14.55*** −14.42*** −14.45*** −14.51***

(0.01) (−0.26) (−0.22) (−0.17) (−0.34) (−9.07) (−9.21) (−8.99) (−9.00) (−9.10)AGE (+) −4.16 −1.64 −2.59 −3.56 −1.12 −0.82 −0.83 −0.71 −0.74 −0.81

(−0.57) (−0.23) (−0.36) (−0.49) (−0.15) (−0.40) (−0.40) (−0.35) (−0.36) (−0.39)PROC (+) −3.27 −2.20 −2.49 −2.48 −2.05 13.03*** 12.94*** 13.27*** 13.07*** 12.98***

(−1.27) (−0.85) (−0.94) (−0.94) (−0.82) (7.64) (7.49) (7.30) (7.60) (7.51)CALLABLE (+) 101.64*** 105.16*** 105.45*** 102.05*** 106.35*** 48.80*** 48.88*** 49.07*** 48.79*** 48.90***

(7.63) (8.21) (8.01) (7.76) (7.65) (12.93) (13.03) (12.94) (12.97) (13.04)CONV (−) −22.34 −24.66 −27.31 −22.01 −27.65 89.85 89.31 90.62 90.64 88.88

(−0.26) (−0.29) (−0.31) (−0.25) (−0.32) (0.74) (0.74) (0.75) (0.75) (0.73)SYND (−) 3.29 1.87 2.37 4.51 3.06 −15.13*** −15.03*** −15.37*** −15.12*** −14.88***

(0.39) (0.22) (0.28) (0.54) (0.36) (−3.80) (−3.79) (−3.83) (−3.80) (−3.74)SOVRAT (−) −20.65*** −17.04*** −19.65*** −20.41*** −19.03*** −27.65*** −28.15*** −26.84*** −26.53*** −29.08***

(−6.13) (−4.86) (−5.71) (−5.56) (−5.09) (−8.31) (−8.18) (−8.04) (−6.99) (−7.21)INTDB (+) 180.22*** 137.92*** 139.16*** 167.81*** 148.28*** 171.26** 165.68** 151.44** 160.45** 133.75***

(5.83) (3.76) (4.36) (5.32) (4.78) (2.46) (2.29) (2.38) (2.26) (2.60)CRBUR (−) −122.06** −104.53** −110.39** −116.77** −103.51** −109.25** −105.82** −108.70** −110.38* −106.27*

(−2.30) (−2.08) (−2.20) (−2.26) (−2.08) (−2.02) (−1.98) (−1.97) (−1.92) (−1.93)CREDR (−) −6.07 −4.31 −1.74 0.17

(−1.59) (−0.85) (−0.38) (0.03)ENFOR (−) −0.92*** −0.89 0.11 0.91

(−3.00) (−1.21) (0.34) (0.88)COMMON (−) −26.23** −2.38 −16.64 −39.38

(−2.39) (−0.08) (−0.88) (−0.95)LEGALR (−) −1.58 5.53 −3.93 −1.74

(−0.54) (1.37) (−0.79) (−0.31)

16

A.

Samet,

L. O

bay /

J. of

Multi.

Fin. M

anag. 25–26

(2014) 1–20

Table 7 (Continued)

Variable Predicted sign Panel A Panel B

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Intercept (?) 979.92*** 976.71*** 973.15*** 967.91*** 977.30*** 984.68*** 980.08*** 982.91*** 991.77*** 985.02***

(10.09) (10.23) (10.07) (9.89) (10.32) (11.43) (11.41) (11.32) (11.25) (11.22)INDUSTRY & YEAR EFFECTS YES YES YES YES YES YES YES YES YES YESAdj. R2 (%) 39.42 39.95 39.63 39.25 39.93 51.80 51.80 51.83 51.84 51.89# of obs. 3083 3083 3083 3077 3077 11,572 11,572 11,572 11,568 11,568

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 1726firms 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. Industrygroup 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-statisticsare reported in parentheses.

* Significance at 10% level.** Significance at 5% level.

*** Significance at 1% level.

A.

Samet,

L. O

bay /

J. of

Multi.

Fin. M

anag. 25–26

(2014) 1–20

17

Table 8Callable yield premium.

Variable Predicted sign Panel A Panel B

Inv. grade Junk Inv. grade Junk Low leverage High leverage Low leverage High leverage

TA (−) −8.19*** −33.52*** −8.49*** −33.14*** −22.66*** −23.92*** −22.41*** −24.34***

(−5.41) (−8.02) (−5.63) (−8.11) (−10.43) (−8.03) (−10.38) (−8.17)LEV (+) 29.79*** 76.68*** 30.60*** 76.72*** 8.50 134.11*** 6.47 142.95***

(3.62) (3.15) (3.81) (3.17) (0.37) (4.76) (0.27) (4.95)PROFIT (−) −166.90*** −208.41** −165.64*** −200.95** −226.73*** −365.24*** −224.85*** −354.82***

(−6.50) (−2.55) (−6.55) (−2.59) (−3.95) (−3.32) (−3.94) (−3.23)GROWTH (+) 0.54 12.19*** 0.65 10.83*** 2.80 15.91*** 2.78 16.06***

(0.55) (3.80) (0.63) (3.21) (0.81) (4.01) (0.81) (4.05)USLIST (−) −22.86*** −21.49 −17.35*** −49.26 −47.57*** −46.50*** −47.13*** −42.64*

(−4.13) (−0.88) (−3.28) (−1.60) (−5.32) (−3.84) (−3.92) (−1.82)PUB (−) −37.77*** −54.79*** −38.48*** −60.35*** −81.06*** −94.58*** −81.80*** −93.73***

(−8.09) (−5.39) (−8.28) (−6.11) (−11.47) (−8.44) (−11.54) (−8.68)RATR (−) −1.16 5.49 −1.15 5.10 −14.27*** −5.98*** −14.37*** −5.95***

(−1.03) (1.63) (−1.02) (1.49) (−12.00) (−2.88) (−12.08) (−2.84)AGE (+) 7.70*** −53.25*** 8.01*** −51.43*** −2.12 3.07 −1.85 3.59

(3.84) (−5.77) (4.05) (−5.76) (−0.76) (0.84) (−0.65) (0.99)PROC (+) 2.36 12.54*** 2.71** 11.99*** 10.28*** 6.48*** 10.09*** 6.64***

(1.63) (2.77) (2.00) (2.67) (5.11) (3.01) (4.99) (3.17)CALLABLE (+) 20.75*** 62.90*** 19.82*** 65.83*** 48.36*** 65.93*** 48.38*** 66.10***

(7.38) (7.15) (7.07) (7.33) (11.57) (10.97) (11.47) (10.27)CONV (−) 784.07*** −143.13* 780.42*** −156.34** 148.28 46.93 148.96 41.46

(3.20) (−1.82) (3.18) (−1.97) (0.99) (0.53) (1.00) (0.48)SYND (−) 5.76* −32.41*** 5.23 −29.99*** −9.53** −4.95 −8.75** −5.33

(1.67) (−3.50) (1.59) (−3.35) (−2.14) (−0.91) (−1.97) (−0.98)SOVRAT (−) −4.65** −16.56*** −3.60 −12.69*** −22.18*** −27.64*** −19.66*** −25.76***

(−2.22) (−4.82) (−1.55) (−2.60) (−11.75) (−2.94) (−7.60) (−2.67)INTDB (+) 86.04*** 385.27*** 91.52*** 236.24*** 137.26*** 194.88*** 133.52*** 186.79***

(4.64) (2.75) (5.47) (2.66) (3.82) (3.66) (3.81) (4.08)CRBUR (−) −75.46 −65.7 −57.05 −50.73 −113.09** −93.13 −115.65** −87.02

(−0.10) (−1.04) (−0.14) (−0.77) (−2.46) (−0.13) (−2.47) (−0.18)CREDR (−) 0.61 14.29 0.95 −6.37

(0.29) (1.20) (0.22) (−0.97)ENFOR (−) −1.22*** 0.89 −0.36 −0.79

(−5.11) (0.81) (−0.72) (−0.50)COMMON (−) 37.25*** −130.71** 5.03 5.61

(4.07) (−2.24) (0.24) (0.09)LEGALR (−) −1.14 0.24 −3.08 −0.69

(−0.50) (0.03) (−0.78) (−0.11)

18

A.

Samet,

L. O

bay /

J. of

Multi.

Fin. M

anag. 25–26

(2014) 1–20

Table 8 (Continued)

Variable Predicted sign Panel A Panel B

Inv. grade Junk Inv. grade Junk Low leverage High leverage Low leverage High leverage

Intercept (?) 236.96*** 978.63*** 291.63*** 920.82*** 945.02*** 924.02*** 944.10*** 952.08***

(5.37) (8.94) (6.46) (8.26) (15.33) (4.51) (15.28) (4.83)INDUSTRY & YEAR EFFECTS YES YES YES YES YES YES YESAdj. R2 (%) 33.83 29.86 34.30 30.76 48.68 47.62 48.75 47.80# of obs. 10,689 3224 10,686 3221 6950 6963 6950 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 ofour 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 withlow 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 (notreported) 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 forthese 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 19

6. Callable yield premium

To test our two other hypotheses (H2 and H3) discussed above, we break down our sample intotwo subsamples. The first breakdown consists of investment-grade bonds (bonds that are rated BBBand above) and junk bonds (bonds that are rated below BBB). The second breakdown consists of bondsthat are issued by firms that have a leverage ratio lower than the sample’s median leverage and bondsthat are issued by firms with a leverage ratio higher than the sample’s median leverage. We run theregression 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 junkbonds have a higher callable yield premium than investment grade bonds. We may then infer that junkbonds 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 thefirst (second) specification. The difference is statistically significant at 1% level. This result is consistentwith the signaling theory according to which firms can benefit from their bond price appreciation afterrevealing 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 con-jecture (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 differ-ence is statistically significant at 1% level. This result is consistent with the risk-shifting explanationaccording to which firms with high leverage are more likely to undertake riskier projects and thenexpropriate bondholders’ wealth.

7. Conclusion

In this paper we examine the call spread in a global framework using an international sampleand controlling for institutional characteristics. We conjecture that callable bonds have a positive callspread compared to their equivalent non-callable (straight bonds). We also conjecture that the callspread of junk bond is higher than the call spread of investment grade bonds, which is consistent withthe signaling hypothesis. We finally conjecture that the call spread of highly leveraged firms is higherthan 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 statisticallyand economically significant. Our empirical results hold after a battery of robustness checks. We alsofind that junk callable bonds have a higher call spread than investment-grade callable bonds, which isconsistent with the signaling theory. Our empirical results also show that highly leveraged firms havea 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 choosingto issue callable bonds. Indeed, these firms pay a positive call premium to bondholders who acceptto invest in their callable bonds, thus increasing their cost of borrowing. The extra-yield received bycallable bondholders is to compensate these bondholders for the risk that the firm may call back itscallable 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.