Callable bonds, interest-rate risk, and the supply side of hedging

Transcription

1 Callable bonds, interest-rate risk, and the supply side of hedging Levent Güntay Indiana University Kelley School of Business Nagpurnanand Prabhala University of Maryland Smith School of Business Haluk Unal University of Maryland Smith School of Business July 2004 Corresponding author: Nagpurnanand Prabhala, tel: We thank Anup Agrawal, Yiorgos Allayannis, Alexandre Baptista, N. K. Chidambaran, Sanjiv Das, Jean Helwege, David Mauer, Abon Mozumdar, Tim Opler, Catherine Schrand, Lemma Senbet, Peter Tufano, participants at the American Finance Association meetings, and seminar participants at University of Alabama and George Washington University for helpful comments.

2 Callable bonds, interest-rate risk, and the supply side of hedging Abstract We show that firms attach call options to debt issues to manage interest-rate risk and characterize the empirical determinants of this hedging decision. Our results affirm that firms hedging choices are explained by theories of hedging demand, but more importantly, provide novel evidence that the supply side of hedging is equally important. In contrast to studies based on OTC derivatives, small firms are more likely to hedge in our setting, in which supply-side barriers are absent. We show that there is a secular, robust shift away from callable bonds in the 1990s, when supply-side barriers to hedging declined. This shift is more likely when firms disclose derivatives usage disclosed in their 10-K s. JEL classification: G30; G32 Keywords: Hedging; Risk Management; Callable Bonds 2

3 Every firm that issues fixed-rate debt must decide whether to attach a call option to the debt issue. The call option helps issuers hedge against declines in interest rates, by giving them the option to call back the bond at a preset strike price, any time before bond maturity after an initial protection period. While some firms attach call options to their debt issues, others do not. In this paper, we examine the empirical determinants of this choice between callable and noncallable debt. Our analysis leads to a detailed characterization of the interest rate hedging decisions of firms, and in particular, points to the central role played by the supply side in determining these hedging decisions. We start by extensively investigating the relation between callable bond usage and interest-rate risk. In contrast to the rather weak or mixed results reported in received studies, we provide strong evidence that callable bond usage is related to multiple proxies for interest-rate risk faced by issuers. We show that callable debt is more likely to be used when an issuer s operating income is positively related to interest rates. Greater correlation between stock returns and interest rates and higher operating income volatility increases the likelihood of call usage. To the extent that callable bonds are used to protect firms against interest rate declines, their issuance should be more likely when interest rates are high. Consistent with this hypothesis, we find a positive relation between the level of interest rates and callable bond usage. Callable bond usage is also more likely for larger size debt issues and for longer maturity issues, which confront firms with greater interest-rate risk. Our results provide strong support for the argument that calls are hedging tools used to hedge operating income fluctuations caused by interest-rate changes. 3

4 We augment our ex-ante findings by analyzing the ex-post ratings and interest rate changes after callable debt issues. If calls indeed hedge interest rate risk, we should observe that calls are exercised when interest rates decline. We test this hypothesis by analyzing firms that ex-post called their callable bonds. We find that interest rate changes are a significant motivation for calling callable bonds. There is an economically significant decline exceeding 200 (300) basis points in the long (short) Treasury rate between the callable bond issue date and call date. On the other hand, it can be argued that firms attach call provision to exploit ratings improvements after debt issues. However, we find no evidence that credit ratings are more likely to improve after callable bond issues. In sum, both the ex-post and ex-ante evidence show that interest-rate risk is a primary motivation for why firms attach call options to their nonconvertible debt issues. Our next tests examine why some firms use callable bonds to hedge interest-rate risk while others do not. While we begin by following existing studies in explaining hedging, based on theories of hedging demand and find affirmative results, our primary focus and the new results here concern the role of the supply side of hedging. Supply-side barriers such as scale diseconomies and transaction costs are identified as a key impediment to derivatives usage for hedging in surveys such as the Wharton derivatives usage survey (Bodnar, Hayt, and Marston, 1998). Additionally, Litzenberger (1992) and Mozumdar (2001) emphasize that supply-side barriers could also arise out of informational asymmetries that lead to limited access to derivatives for hedging. Such barriers could impact hedging by screening out lower quality firms from the derivatives market and limiting the risk management tools available to these firms. Callable bonds provide a unique avenue for empirically assessing the importance of the supply side. The 4

5 derivative security used for hedging a call option attached to the debt issue presents none of the supply-side barriers that researchers argue impedes hedging by firms. We exploit this unique facet of callable bonds to gain insights into the role played by the supply side of hedging. One of our main results on the supply side of hedging relates to firm size. Theories of hedging demand suggest that smaller firms should be more likely to hedge. For instance, small firms are faster growing and face greater bankruptcy costs, making them more likely to hedge to avoid bankruptcy (Smith and Stulz, 1985). Alternatively, smaller firms have more growth opportunities and thus have a greater need to hedge to avoid the underinvestment costs relating to lack of internal funds for investment (Froot, Scharfstein, and Stein, 1993). Yet, a robust finding in practically all hedging studies is that large firms are more likely to hedge. This result is obtained in multivariate settings with controls for other determinants of hedging. More puzzling is the fact that large firms are more likely to hedge even when firm size is the only explanatory variable used to explain hedging. The positive firm-size hedging relation is at odds with theories of hedging demand. We shed new light on the positive relation between firm size and hedging. In contrast to previous work, we find that in the case of callable bonds, there is a significant negative relation between size and hedging. This result holds up in both univariate and several multivariate specifications that include other controls. Thus, smaller firms are more likely to hedge in the callable bonds context. This finding suggests that in earlier studies, which are based on over the counter (OTC) derivatives usage for hedging, firm size acts as an instrument for the supply side of hedging. When informational or 5

6 transaction cost scale barriers exist and impede derivatives usage, as in the case of offthe-shelf derivatives, larger firms are more likely to hedge by using derivatives. However, when supply-side impediments to derivatives usage are absent, as in the callable bonds context here, small firms are indeed more likely to hedge, as predicted by theory. We provide other results related to the supply side of hedging. We show that first time issuers in the debt market are more likely to use callable bonds. In addition, we document a secular shift away from callable bonds in the later half of our sample period, the 1990s. Although over 70% of debt issues in the 1980s were callable, less than 30% of issues in the 1990s were. The shift away from call usage is robust even after we control for lower interest rates prevalent in the 1990s that could reduce demand for callable bonds as well as several other explanatory variables. Supply-side arguments plausibly explain why firms shifted away from calls. The 1990s represent a period characterized by a significant drop in supply-side barriers to the usage of interest rate options. The market for interest rate options showed significant growth over this period, as evidenced by Bank for International Settlements aggregate data on derivatives usage. With rapid growth and increased accessibility of these interest rate derivative products in the 1990s, firms had less need to manage interest-rate risk by bundling a call option with debt issues. Our findings are consistent with this hypothesis. We develop further evidence on the supply side by analyzing the switching behavior of issuers that moved away from callable bonds in the 1990s. If the shift away from callable bonds is explained by the increased accessibility and availability of off-theshelf interest-rate derivatives in the 1990s, firms with more access to these derivatives 6

7 should be more likely to switch away from callable to noncallable bonds in the 1990s. We hand compile a dataset to test this hypothesis. For firms that issued in both the pre and post-1990 subperiods, we identify firms that switched or didn t switch from callable bonds to noncallable bonds. We read proxies of the both sets of firms to identify firms that disclose derivatives usage in their 10-K statements. Consistent with the substitution of callable bonds by off-the-shelf alternatives in the 1990s, we find a reliable cross-sectional relation between switchers moving away from callable bonds and disclosed derivatives usage in annual 10-K s. Collectively, the cross-sectional evidence is consistent with the hypothesis that supply-side barriers to derivatives usage have a firstorder influence on the hedging strategies of the firm. The rest of the paper is organized as follows. Section 1 overviews the literature related to our study. Section 2 describes our dataset. Section 3 analyzes the relation between callable bond usage and interest-rate risk. Section 4 estimates of a multivariate probit specification to explain the decision to attach call options to bond issue. Section 5 looks at post-issue evidence. Section 6 analyzes firms switching from callable to noncallable bonds. Section 7 concludes. 1. Related literature Our analysis of callable bonds contributes to two streams of research: one on risk management by corporations and the other on choosing between callable and noncallable debt. To place our work in perspective, we briefly relate our findings to received literature in these two areas. 7

8 The theoretical literature on risk management puts forth alternative arguments for why firms have positive hedging demands. In other words, what benefit motivates firms to hedge? Smith and Stulz (1985) argue that firms hedge to reduce the probability of bankruptcy and attendant distress costs. The testable implication is that firms with greater bankruptcy risk are more likely to hedge. Froot, Scharfstein, and Stein (1993) hypothesize that firms hedge to minimize the need to raise external finance to fund investments. This argument suggests that firms with more growth opportunities are more likely to hedge, because these firms have greater need to avoid having to pass up growth options for the lack of internal finance. Smith and Stulz (1985) show that nonlinearities in the tax system can create benefits that lead to a demand for hedging instruments. Nance, Smith, and Smithson (1993) argue that firms other financial policies, such as the use of convertible debt or preferred stock in the firm s capital structure, should interact with the hedging decision. On the supply side of hedging, the literature argues that impediments to derivatives usage, such as scale diseconomies or information asymmetry-based rationing faced by smaller firms, could impact the ability to hedge. The existence of such supplyside barriers is reported as a key influence in firms hedging choices in the Wharton derivatives survey (Bodnar, Hayt, and Marston, 1998). Litzenberger (1992) highlights these barriers in the swaps market in his presidential address. Mozumdar (2001) formalizes Litzenberger s insights into a theoretical model that explains why firms could be unable to hedge due to asymmetric information based rationing in the derivatives market. 8

9 The empirical literature on hedging has developed primarily in the 1990s after hedging disclosures in annual financial statements became mandatory. We add to the empirical literature on three dimensions. First, we offer a novel approach to study hedging. Our study analyzes hedging at the level of the individual hedging transactions of the firm. This approach is complementary to the other approaches used in the literature, which includes the analysis of aggregate, firm-wide portfolio of derivatives (Géczy, Minton, and Schrand, 1997; Allayannis and Weston, 2000), analyses of responses to questionnaires sent to CEOs/CFOs (Nance, Smith, and Smithson, 1992), case studies involving specific firms (Chacko, Tufano, and Verter, 2001; Chidambaran, Fernando, and Spindt, 2001), or studies of particular industries (Mayers and Smith, 1990; Tufano, 1996; Schrand and Unal, 1998). Our second contribution is to bring in time series data on hedging. Our dataset covers a relatively long time span of 17 years from 1981 to 1997, and thereby adds to the primarily cross-sectional focus of earlier studies of hedging, such as Géczy, Minton, and Schrand (1997) or Allayannis and Weston (2000). Extending the evidence on the time-series dimension is important, because it gives some of the first insights into hedging behavior in the 1980s. There is little prior evidence from the 1980s because hedging-related disclosures in financial statements, the dominant source of data for recent studies, were not mandatory until the 1990s. We contribute towards filling in this gap. Finally, our study differs from previous work through its emphasis on the supply side. While we do include demand proxies proposed in previous work as controls and find supporting evidence, our central point is that regardless of any particular demand side reason for why firms hedge, the supply side matters and is robust in influencing firms hedging strategies. 9

10 Besides the risk management literature, our paper also adds to empirical studies on the determinants of the choice between callable and noncallable debt. These early studies test the signaling and security design theories of Barnea, Haugen, and Senbet (1980) and Robbins and Schatzberg (1986). The studies are quite fragmented in terms of samples, time periods, and explanatory variables. Thatcher (1985) analyzes 118 debt issues made in 1975, while Mitchell (1991) analyzes 265 issues made between 1982 and Kish and Livingston (1992) analyze 2061 issues made between 1977 and 1986, a period in which virtually all bonds were callable and that predates the structural change away from callable bonds in the late 1990s. More recently, Sarkar (2001) is based on a sample of 104 debt issues in 1996 and two months of In these studies even the basic proposition that callable debt is more likely when interest rates are high is very weakly supported or even rejected in existing studies. We consolidate and update the fragmented samples and evidence in the callable bond studies mentioned above. We offer a comprehensive set of explanatory variables, as well as a more extensive time series. In contrast to the narrower time periods examined by previous callable bond studies, our dataset spans about two decades over which there is substantial variation in call usage as well as interest rates, which allows for more powerful tests. Our study also differs from the previous studies in that we develop stronger evidence for the hedging explanation for issuing callable bonds through multiple approaches. We also add to the cross-sectional analyses of the previous studies by analyzing post-issue evidence on ratings and interest rate changes as well as evidence on switching behavior away from callable bonds in the 1990s, and develop their implications for the supply side of hedging. 10

11 Three papers in the callable bond literature analyze ex-post differences between these two sets of debt issuers. Post-issue rating changes are studied in Crabbe and Helwege (1994) for a sample of issuers in the 1980s. We extend and update their analysis to cover issues from 1981 to Two other post-debt-issue studies are Vu (1986) and King and Mauer (2000), the latter representing a considerably larger scale study with more recent data. Both papers analyze ex-post calls of callable bonds, focusing on call delays and events around the call date itself (e.g., announcement effects). We add to these studies by analyzing interest rate changes between the issue and call date. In particular, we report that there is a significant decline in the Treasury yield curve between issue and call, suggesting that interest rate changes are a significant part of the explanation for why firms call their callable bonds. 2. Data We identify nonconvertible fixed rate bonds issued between January 1981 and December 1997 from New Issues Database of Securities Data Company (SDC). We exclude issues made by financial firms and selected service firms. We filter out financials by omitting firms with 4-digit SIC codes between 6000 and We further screen out 146 firms in the filtered sample with names that contain the phrases "Acquisition," "Capital," "Credit," "Financial," "Finance," "Funding," "Leasing," and "Security." We also exclude leasing firms (SIC codes 7352, 7353, 7359, 7377, 7513, and 7515), and service firms in the educational services, social services sectors, membership organizations, and other non-classifiable establishments (SIC codes between 8200 and 8299, 8300 and 8399, 8600 and 8699, and ). We obtain 7943 bonds as a result 11

12 of these two filters. Additionally, we restrict our sample to firms for which crosssectional information is available in the Compustat database. The Compustat and SDC matched sample consists of 4,188 bond issues from 1981 to To classify a bond as callable or noncallable is not as straightforward as it might seem initially. The SDC database does not explicitly code a bond as being callable or not. Instead, the database gives a "call protection period" during which an issuer cannot call the bond. After the protection period the issuer can call the bonds if it wishes to do so, typically any time up to the final maturity of a bond. This reporting practice suggests that if the call protection period equals final maturity, a bond should be classified as noncallable, and protection periods less than the final maturity could qualify a bond as being callable. However, we modify this classification rule because call protection periods are sometimes close to but still less than the maturity to allow issuers some flexibility to time the reissuance of the bond close to its maturity. For example, a 30-year bond could be callable after 29 years, thus allowing the issuer some flexibility on timing the refunding in the last year of a bond's life. Such a bond is better classified as a noncallable bond. Therefore, we impose some separation between maturity and call protection to identify callable bonds. We use the SDC database data field "number of years until maturity" to identify the maturity of the bond. Where this field is missing, we calculate it by using the "issue date" and "final maturity date" fields. For bond maturity structures, it is a well-known fact that most corporate bond issues have standard at-issue maturity structures such as 12

13 five, ten, or thirty years to reflect the maturity structures of the most liquid on-the-run Treasuries off which the bonds are priced. We find a similar, though not identical, distribution for call periods. When we compare the call protection period with bond maturity, we find that for five-year bonds, the average call protection period is three years or lower, while the average call protection period is close to five years for all longer maturity callable bonds. This average is consistent across all maturity structures and the sample period. Hence, we define a bond as being callable if the call protection period is less than one year for bonds with three to seven year maturities, five years for bonds with seven to ten year maturities, seven years for bonds with ten to fifteen year maturities, and ten years for bonds with greater than fifteen year maturities. Table 1 gives descriptive statistics on callable bond usage for the full sample period as well as the two subperiods 1981 to 1988 and 1989 to We observe that callable bonds are the debt instruments of choice in the 1980s, but their popularity diminishes towards the late 1980s. Callable bonds constitute 74.3% of the sample bonds issued between 1981 and 1988, but the percentage of callable bonds drops off to 25.0% of the sample bonds issued in the second half of the period. The decline in call usage is more prominent in high and medium rated firms and less so in noninvestment grade firms, and is not particular to any one industry group. The dramatic decline in usage of callable debt is also illustrated by Figure 1, which reports the percentage of bonds in our sample of 4,188 Compustat and SDC matched issues that are callable for each year between 1981 and The transition from high to lower call usage occurs over the three-year period from 1988 to The figure also shows that a shift in callable bond 13

14 usage occurs in a period of rapidly expanding activity in bond issuance. For instance, the total dollar volume of the new corporate bond issues is $456 billion from 1989 to 1997; more than triple the volume in the period. Figure 2 shows that the decline in call usage coincides with a period of falling interest rates. Figures 3 and 4 indicate that the declining call usage is not confined to any particular maturity or rating category. However, Figure 4 shows that noninvestment grade bonds followed the general declining trend but with a lag and issue of callable bonds rebounded in this category after a sharp decline in Callable bonds and interest-rate risk This section develops evidence that callable bonds address the interest-rate risk hedging needs of firms. We consider four sources of interest-rate risk exposure of the firm the issuer s operating income sensitivity to changes in interest rates, the level of interest rates at the time of the issue, the maturity of the debt issue, and the issue size. The full set of variables used in the study and their definitions are given in the Appendix. 3.1 Interest-rate risk variables If a firm s operating cash flow is correlated with interest rates (or macroeconomic shocks related to interest rates), fluctuations in interest rates can induce volatility in firms operating income volatility. The direction of this correlation depends on the nature of the firm s business. For example, petroleum companies operating cash flows are positively correlated with interest rates to the extent that petroleum prices and interest rates have a positive relation. In contrast, the transportation firms cash flows can be 14

15 negatively correlated with changes in interest rates if petroleum prices fall parallel to a decline in interest rates, reducing the costs and increasing the firms operating cash flows. A call option attached to a bond issue offers a hedge for firms whose operating income is positively correlated to interest rates. Such firms are likely to have lower cash flows when interest rates are low and would face paying higher interest costs at this time if the debt issue were noncallable, giving rise to interest-rate risk. A call option attached to a debt issue would mitigate the risk by providing positive cash flows in the low interest-rate states of the world, thereby reducing the firm s cash flow volatility. Hence, we expect that firms having operating cash flows that are positively correlated with changes in interest rates are more likely to attach call features to their bond issues. Our proxy for the interest rate sensitivity of operating income is Operating Income Beta. Following Graham and Rogers (2002), we construct this variable by regressing quarterly changes in operating income of each firm in our sample on quarterly changes in the six-month LIBOR rate, using seven years of quarterly data. We have sufficient data to compute the variable for 75% of our sample (3,243 observations out of 4,188 sample bond issues). One issue with Operating Income Beta is that it does not account for the statistical significance of the beta estimate. Accordingly, we construct a second proxy on the lines suggested by Graham and Rogers. As in Graham and Rogers, we classify the Operating Income Exposure as being +1, 0, or -1 based on the sign and significance of the Operating Income Beta (OIB). Zero exposure occurs when OIB is not significant at 10%. We also consider related, though less direct proxies for operating income exposure to interest rates. One proxy is the correlation between interest rates and stock 15

16 market returns, which we estimate as the correlation between weekly stock returns of the firm and changes in the 10-year Treasury rates over two years prior to a debt issue. In addition, we also measure the operating income volatility as the standard deviation of the percentage changes in operating income over six years prior to the debt issue. We expect that those firms that have high volatility of operating income would prefer to use callable debt to avoid adding additional volatility caused by interest-rate risk. We consider three additional proxies for interest-rate risk exposure: the interestrate level (Risk-Free Rate), the logarithm of the bond maturity (Log Issue Maturity), and the logarithm of the issue amount (Log Issue Amount). We measure the interest-rate level as the Treasury bond rate that matches the bond maturity interpolated off a cubic spline fitted to the term structure. At high levels of interest rates, there is greater potential for interest rates to fall over the life of the bond, so the protection afforded by the call provision is more valuable. Conversely, at low levels of interest rates, calls are less useful as hedges. Hence, if calls are used to hedge against interest-rate risk, callable bonds should be more likely in periods with higher interest rates. Longer maturity bonds have greater sensitivity to interest-rate fluctuations and thus pose greater risk, so call provisions should be more likely for these bonds under the hedging hypothesis. Finally, we include Log Issue Amount as an explanatory variable because more material interest rate exposures are likely to be created by large-size debt issues Univariate results Table 2 reports univariate differences in interest-rate risk proxies between callable and noncallable bond issues. Firms with negative operating income sensitivity to interest 16

17 rates are more likely to issue noncallable bonds as predicted by the hedging explanation for issuing callable bonds. Of the 3,243 observations for which we can compute the exposure variable, 758 have negative exposure, 2,379 have insignificant (zero) exposure, and only 106 have positive exposure. Thus, most fixed rate issuers in our sample have zero or negative exposure. Callable debt usage varies according to the nature of the issuer s operating income exposure in a direction predicted by theory. In our sample, 27% of the negative exposure firms, 39% of the no-exposure firms, and 42.5% of the positive exposure firms use callable debt. The difference in call usage between the negative and zero exposure samples is significant (t-statistic for difference is 6.40), as is the difference between the negative and positive exposure samples (t-statistic for differences is 3.16). The difference between zero exposure and positive exposure sample goes in the right direction but is not significant (t-statistic is 0.71), perhaps because of less power due to the relative paucity of positive-exposure firms in our sample. Taken together, our results suggest that firms with negative operating income exposure to interest rates are measurably more reluctant to include a call option. Negative exposure firms are naturally hedged against falling interest rates, since their operating incomes rise with falling interest rates. Thus, these firms do not use or pay for the similar protection against falling interest rates that is provided by callable bonds. 1 1 We also examine exposures for floating rate issuers. Over our sample period, we have 788 floating rate issues. We have data to compute the Graham and Rogers (2002) operating exposure variable for 168 of them. Out of this sample, 18.4% of floating rate issuers have positive exposure, 67.3% have zero exposure, and 14.3% have negative exposures, versus 3.3%, 73.3%, and 23.4% for the fixed rate sample. The differences in proportions between positive and negative issuers in the two samples are significant (tstatistics = 3.87 and 2.26, respectively). Thus, while firms with negative exposure issue callable bonds, those with measurably positive exposure issue floating rate debt, consistent with predictions of the hedging hypothesis. 17

18 Table 2 also shows that callable bond issuers have higher operating income volatility, consistent with the hypothesis that callable bonds are used to avoid adding additional volatility that can be produced by financing decisions. The more indirect proxy for operating exposure, correlation between stock returns and interest rate changes, is not significant. Callable issues appear to be more likely when interest rates are higher. The median (mean) Treasury interest rate level of 7.58 (8.37) for callable issuers exceeds the median (mean) for noncallable issuers of 6.63 (6.94). We also observe that the additional interest-rate risk variables, Issue Maturity, Log Issue Maturity, and Log Issue Amount, all prove to be significantly higher for callable bond issues, providing further support to the argument that call feature is an interest-rate risk hedging instrument Multivariate probit results Table 3 reports estimates of a multivariate probit model that incorporates all the interest rate exposure variables in one specification. The multivariate results are consistent with the univariate differences in Table 2. The operating income exposure variable is positively related to callable debt usage and its coefficient is significant for the full-period estimates. In subperiods, the first-period coefficient has a lower magnitude and is insignificant. This finding suggests that operating exposure has little incremental effect in period one when interest rates were high so most firms decided to opt for callable debt. However, the coefficient for the second subperiod and the full period are positive, greater in magnitude, and statistically significant. The signs of these coefficients are consistent with the hypothesis that the use of callable debt is positively related to the variation of operating income with changes in interest rates. 18

19 We also consider other exposure proxies, including the Operating Income Beta itself and other indirect exposure proxies. Using the estimated Operating Income Beta instead of the Graham and Rogers (2002) variable increases the errors-in-variable, because the error in estimating the coefficient probably exceeds the error in estimating whether it differs significantly from zero. Nevertheless, Table 3 reports that the coefficient for Beta is significant and positive, as predicted by the hedging hypothesis. Likewise, the correlation between stock market returns and interest rate changes has a positive and significant sign. Table 3 also shows that those firms, which use callable bonds have significantly higher operating income volatility. This finding supports the argument that callable bonds are used to avoid adding additional volatility that can be produced by financing decisions. Thus, all our measures of interest rate exposure produce results consistent with the hedging hypothesis for using callable bonds. However, the indirect proxies lose significance in multivariate specifications that incorporate firm size and credit ratings. We omit these variables in the probit estimates of Section 4, retaining the Graham and Rogers variable as the proxy for operating exposure to interest rates. Table 3 also shows that the coefficient for Risk-Free Rate is positive and significant, indicating that callable bond issues are more likely when interest rates are high. Coefficients for the two other proxies for interest-rate risk, i.e., Log Issue Maturity and Log Issue Amount, are significant in Table 3 in a direction consistent with the hedging hypothesis. Log Issue Maturity has a positive, significant coefficient during the full period as well as the subperiods. Log Issue Amount is positive and significant in both periods. Both maturity and issue amount have higher coefficients in period 1 when the risk posed by declining rates is higher. 19

20 In sum, the evidence presented thus far is consistent with the hedging argument for why firms attach a call feature to their bond issues. These findings support an early conjecture by Kraus (1983), who writes that the interest-rate hedging explanation for issuing callable debt has received little, if any, attention in the finance literature, [but] it offers another clue to the puzzle - one that gets closest to management's concern about the need to protect the company against exposure to changes in interest rates.'' 4. Callable bonds and the supply side of hedging The ex-ante relation between callable bond usage and interest-rate risk proxies suggests that callable bonds address interest-rate risk hedging needs of firms. Some firms choose to hedge using callable bonds while others do not. In this section, we examine the determinants of this hedging decision, focusing in particular on the supply side of hedging. While much of the received theoretical and empirical literature on hedging focus on the benefits that cause firms to have positive hedging demands, supply-side barriers could be equally important in determining firms risk management policies. Scale economies in derivatives usage represent one supply-side barrier, constraining all but large firms from using derivatives to hedge. Information asymmetry could represent yet another supply side obstacle to using derivatives by limiting their availability to only the low credit risk, transparent, and high quality firms. Litzenberger (1992) addresses such barriers in the context of the swap market. Mozumdar (2001) offers a formal model of information-asymmetry based rationing in the derivatives market. Survey data in Bodnar, 20

21 Hayt, and Marston (1998) suggest that such supply-side barriers constrain firms from using derivatives to hedge. Callable bonds offer a unique avenue for clarifying the role of the supply side of hedging. The hedging instrument involved the embedded call poses none of the supply-side barriers that have been argued to impede derivatives usage. The call requires no fixed investment that could represent a scale barrier. The derivative security involved a call option is familiar and is tied to a specific debt transaction, mitigating the necessity to screen for speculative intent that leads to rationing in Mozumdar s model. Thus, many of the supply-side barriers traditionally thought to impede derivatives usage for hedging are absent in the callable bond context. This feature leads to unique tests of hedging theories, as discussed below. The empirical results are presented as follows. Table 4 reports descriptive statistics for several characteristics of callable and noncallable issuers, with Wilcoxon z(p) or binomial test z-statistics for testing significance of differences between C and NC firms. We analyze the differences between C and NC firms more fully via a multivariate probit specification. We open the discussion by analyzing demand side probit results in Section 4.1. While not our focus, the demand side results do provide a useful point of departure to introduce control variables, and reaffirm that familiar hedging demand variables are significant and have the right signs in the callable bond context as well. Section 4.2 deals with the supply side results, while Section 4.3 examines the robustness of our probit estimates. 21

22 4.1 Demand side controls We control for the Smith and Stulz (1985) bankruptcy cost theory of hedging demand by including a proxy for bankruptcy risk in the probit specification. We transform the issuer s credit rating into a Rating Squared cardinal variable as suggested in Stohs and Mauer (1996): AAA=1, AA=2 2, A=3 2, BBB=4 2, BB=5 2, B=6 2, below B=7 2. The Stohs and Mauer transformation squares the numerical value for a rating. This allows the impact of rating deterioration to be greater as ratings move from being AAA toward below B. 2 The bankruptcy cost theory predicts a positive relation between squared rating and callable debt usage. To control for the Froot, Scharfstein, and Stein (1993) external finance avoidance rationale for hedging, we include a proxy for the growth prospects of the issuer, and specifically test whether more rapidly growing firms are more likely to issue callable debt. Our proxy for growth is an issuer s book-to-market (B/M) ratio: high growth firms have low B/M ratios. Nance, Smith, and Smithson (1993) emphasize that firms other financial policies could affect the hedging decision, by adding to hedging demand or by acting as substitute hedging mechanisms. To control for such effects, we follow previous work in including the convertible debt and preferred stock of the issuer as explanatory variables. The predicted signs for these variables are ambiguous. Both convertible debt and preferred stock involve periodic payments of fixed amounts of cash, creating constraints similar to fixed rate debt. Thus, their presence could be positively related to call usage. On the other hand, convertible debt can mitigate asset substitution problems or overinvestment, while 2 Other proxies for bankruptcy risk, such as leverage, interest coverage ratio, and a zero-one binary variable based on whether the bond issue is investment grade or not, give similar results. 22

23 preferred stock reduces bankruptcy risk, so both instruments reduce the need to hedge and could be negatively related to call usage. Table 5 shows that our proxies for the commonly accepted hedging demand arguments have their expected signs. The coefficients for Rating Squared is significant for both the full period and the two subperiods, providing strong support for the bankruptcy cost theory of why firms hedge. We also find that the B/M ratio has a negative and significant coefficient both in full period specification as well as the subperiod, suggesting that more rapidly growing firms are more likely to hedge using callable bonds, consistent with the Froot et al. predictions. 3 Convertible debt has a negative sign for the full period and the first subperiod, consistent with a substitution effect, while preferred stock has a positive sign in all estimates, consistent with its leverage effect driving firms to greater use of callable bonds Supply side results Firm size Theories of hedging suggest that smaller firms are more likely to hedge. For instance, bankruptcy cost theories of hedging suggest that small firms are more likely to hedge, since costs of financial distress do not increase proportionately with firm size (Warner, 1977). Alternatively, small firms are faster growing and therefore more likely to hedge to avoid underinvestment problems (Froot, Scharfstein, and Stein, 1993). Yet, the 3 Exploiting the hedging benefit of the call provision requires that firms refinance via a replacement debt issue after a call. The Froot, Schafstein, and Stein (1993) argument, however, emphasizes the advantage of hedging as avoiding external finance. Extending the Froot et al. argument to callable bonds requires that the cost of external financing via refinanced debt is cheaper than other forms of external finance. 4 Nance, Smith, and Smithson (1993) note that hedging creates tax benefits if there are tax-preference items, because it reduces the probability that these items will need to be carried forward into the future. In unreported regressions, we do not find the tax loss carryforward variable to be significant. 23

24 available evidence suggests a robust relation in the opposite direction: large firms are more likely to hedge using OTC derivatives. The relation is documented in multivariate tests that control for other determinants of hedging, but more surprisingly, shows up even in univariate tests in which size is the only explanatory variable. Supply-side arguments plausibly explain the positive size and hedging relation in OTC-based studies. Small firms are less likely to have the scale economies that allow them to exploit OTC derivatives. Alternatively, greater information gathering costs for smaller firms make it harder for dealers to gauge the speculative or hedging intent of small firms, so small firms are effectively screened out of OTC derivatives markets. Thus, supply-side barriers could make OTC hedging instruments accessible only for large firms, and thereby induce a positive correlation between size and hedging. However, in the callable bonds context, supply-side barriers are minimal, given the fact the absence of fixed investments to set up and manage a portfolio of derivatives and the absence of asymmetric information issues that could lead to screening out small firms. Absent supply effects, demand side effects should dominate, so the positive size-hedging relation should be reversed in callable bonds. The supply-side hypothesis predicts that small firms are more likely to use callable bonds, reversing the relation documented so extensively in the OTC derivatives context. We measure the size of a firm by its annual sales in the year preceding the debt issue deflated by an annual GDP index that is defined to be 1.0 in 1980 and grows at the GDP growth rate each year thereafter. We use the natural logarithm of deflated sales as a proxy for firm size in the empirical tests, although we obtain similar results when we use the logarithm of the GDP-deflated book value of assets rather than sales. From Table 4, 24

25 we see that callable bond issuers are smaller. In the probit results reported in Table 5, firm size has a negative and significant coefficient both in the full sample period and the two subperiods. Small firms are more likely to hedge using callable bonds, as predicted by the supply-side hypothesis. Our size result is consistent with the negative relation between hedging and firm size in the particular case of the insurance industry (Mayers and Smith, 1990). Mayers and Smith find that small property-casualty-insurance companies are more likely to hedge their insurance obligations through reinsurance. They argue familiarity with the hedging instrument and its role in hedging can act as a significant supply-side barrier to derivatives usage. This barrier is absent in the insurance industry, where insurers can readily understand and use the hedging instrument reinsurance so small firms hedge more in this market. Likewise, in the interest rate hedging context, when supply-side barriers are minimal as in callable bonds, small firms are more likely to hedge. When such barriers are present, our results suggest that size should be interpreted as an instrument for the supply side First-time issuers To further investigate the role of the supply side, the probit specification includes a dummy variable for whether a debt issue is the first made by a new entrant into the bond market. Given the absence of an extensive prior bond issuance program, a first-time issuer is less likely to have the personnel, expertise, or economies of scale needed to manage an OTC derivatives portfolio to hedge interest-rate risk. If such expertise and scale economies are in fact significant barriers to OTC derivatives usage, new issuers 25

26 should be most likely to issue callable debt rather than noncallable debt. Consistent with this prediction, Table 4 shows that callable issuers are more likely to be first time issuers. Table 5 reports a significant, positive coefficient for the First Issue dummy variable in the probit specification. Newcomers to the bond market seem to find it convenient to meet their interest-rate hedging needs through a straightforward instrument that does not entail the fixed costs discussed above The time period dummy The final variable for supply side effects is a zero-one dummy variable (Time Dummy) for whether a debt issue occurred between 1981 and 1988, in the first half of our sample period. We include the time dummy variable to test whether the decline in call usage goes beyond that explained by interest-rate changes. If, for instance, lower supply-side barriers to OTC derivatives usage have a first-order effect on firms' hedging decisions, we expect the use of callable bonds to be negatively related to the time-shift variable even after controlling for interest rate shifts and other influences on the hedging decision. We obtain similar results when the inter-period cutoff is defined to start in 1989 or The estimates in Table 5 show that the time dummy variable is negative and significant at better than 1%. This result is particularly interesting because the bond market has grown in the 1990s mainly from firms issuing at the lower end of the ratings spectrum. Lower rated firms are more likely to use callable bonds, as discussed in Section 4.1, so their increasing presence in the 1990s should lead to greater usage of calls. Nevertheless, call usage drops in the 1990s. A plausible explanation is that falling 26

27 supply-side barriers makes off-the-shelf derivatives more accessible in the 1990s, resulting in more firms using these alternatives instead of callable bonds. We provide additional evidence on the declines in supply-side barriers in interest rate hedging by testing whether there is an observable increase in the usage of interest rate derivative products over the period when call usage declined. We obtain data on interest rate derivatives from the Bank for International Settlements (BIS) website. The BIS dataset gives the notional amount of derivatives classified by market and product type. Ideally, we would like to have data from 1981, when our sample period begins, but the earliest data we could obtain starts in Figure 5 plots the annual growth in notional amount of interest-rate options over our sample period starting in There is clear evidence of significant growth in the late 1980s and early 1990s. For instance, growth exceeds 100% in 1988 and remains above 40% in each successive year until Growth tails off towards the latter part of the sample period. Interestingly, we do not see similar growth in derivatives usage in another commodity, currency, which is also depicted in Figure Robustness of specification We examine the robustness of the probit results in Table 5 to alternative specifications that incorporate nonlinearity in interest rates lagged changes in interest rates as determinants of hedging. The first variable tests an alternative missing variable interpretation of the time period dummy variable, while the latter tests for an alternative market timing interpretation of the callable bond issuance decisions. 27

28 The probit model estimated in Table 5 specifies the latent variable governing firms choices between callable and noncallable bonds as being linear in interest rates. However, hedging benefits could be nonlinear in interest rates. It is particularly important to rule out nonlinearity in view of the significance of the 1990s time period dummy variable in the probit model. While we interpret the significance of the time period dummy variable as evidence of a structural shift away from callable bonds due to declining in supply-side barriers, it might instead merely reflect an omitted nonlinearity in interest rates. Columns 1 through 3 in Table 6 report the results of a specification with the squared interest rate included as an extra explanatory variable. Neither the squared term nor the interest-rate term is significant. Importantly, introducing the nonlinear term has little effect on the other coefficients particularly that for the time period dummy. Thus, we find little evidence that the time period dummy variable reflects an omitted nonlinear dependence of callable bond issuance on interest rates. We include lagged interest rate changes to test a market timing hypothesis. Surveys such as Graham and Harvey (2001) and the Wharton survey of risk management practices (Bodnar, Hayt, and Marston, 1998) suggest that managers do incorporate market views into their decisions while taking derivatives positions. As Bodnar, Hayt, and Marston report, "...32% of firms that use derivatives reported that their market view of exchange rates leads them to `actively take positions' at least occasionally. A similar result is found for the market view of interest rates." Bodnar et al. report that about twothirds of the firms alter the timing of a transaction because of a view on interest rates. Thus, speculative intent could partly explain callable-debt usage, and given the evidence in the Wharton survey, we include a proxy to test this timing hypothesis. 28

29 Testing the timing hypothesis requires an assumption and a specification. The assumption is that managers believe that any trend they spot is not priced in the callable bond market. Additionally, we need a specification that tells us what perceived trends would lead to a greater probability of callable bond issues. If we assume that managers are trend chasers, recent changes in interest rates should influence the decision to issue callable bonds. If managers extrapolate recent increases in interest rates to mean that interest rates will continue to rise, protection against declining interest rates is less useful, so a callable bond issue should be less likely. Thus, timers are less likely to issue callable bonds when interest rates have risen. In unreported results, we find that the coefficient for the interest-rate change over a six-month period prior to issue is not significant in the full period or the first subperiod, and is 10% significant in the second subperiod. However, the sign of the significant coefficient is positive, contrary to the prediction of the timing hypothesis. We then consider the possibility that callable bond issuance is driven by managers attempts to time the slope of the term structure rather than levels of interest rates. To test this possibility, we include in the probit specification the change in the term spread over the six-month period prior to a debt issue. We define the term spread as the difference between ten- and one-year Treasury rates. The last three columns of Table 6 report probit estimates that include the proxy for changes in the term structure of interest rates. The coefficient for Change in Term Spread is negative and significant for the full period and both subperiods, indicating that managers are indeed more likely to issue callable debt when the yield curve is flatter. None of the other coefficients are, however, affected. This finding suggests that while the strategic decision to hedge is made in a 29