Corporate use of interest rate swaps: Theory and evidence

Journal of Banking & Finance 27 (2003) 1511–1538 www.elsevier.com/locate/econbase Corporate use of interest rate swaps: Theory and evidence Haitao Li...
Author: Britton Kennedy
14 downloads 2 Views 296KB Size
Journal of Banking & Finance 27 (2003) 1511–1538 www.elsevier.com/locate/econbase

Corporate use of interest rate swaps: Theory and evidence Haitao Li b

a,*

, Connie X. Mao

b,1

a Johnson Graduate School of Management, Cornell University, Ithaca, NY 14853, USA Fox School of Business and Management, Temple University, Philadelphia, PA 19122, USA

Received 14 April 2000; accepted 22 November 2001

Abstract We develop a simply theory on interest rate swaps based on the difference between bank loans and public debts. While restrictive covenants of bank loans help reduce agency costs, banks also have natural disadvantages in bearing interest rate risk due to their floating liabilities. A firm that wants a fixed-rate loan can borrow a floating-rate loan from a bank and enter an interest rate swap to hedge the interest rate risk. Consistent with our theory, we find empirically that fixed-rate swap payers generally have lower credit ratings, higher leverage ratios, higher percentages of long-term floating-rate loans, and are more likely to use bank loans than floating-rate swap payers.  2003 Elsevier Science B.V. All rights reserved. JEL classification: G21; G32 Keywords: Interest rate swap; Agency costs; Information asymmetry; Bank loan; Comparative advantage

1. Introduction Interest rate swaps are one of the most widely used financial derivatives for managing interest rate risk. 2 While swaps have become extremely important, the

*

Corresponding author. Tel.: +1-607-255-4961. E-mail addresses: [email protected] (H. Li), [email protected] (C.X. Mao). 1 Tel.: +1-215-204-4895. 2 For detailed reviews of swap contracts and the history of the development of swap markets, see Smith et al. (1986, 1988) and Wall and Pringle (1988). Duffie and Huang (1996) and Li (1998) study the pricing of credit risk in interest rate swaps. 0378-4266/03/$ - see front matter  2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0378-4266(02)00275-3

1512

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

economic rationales behind these transactions are not fully understood (see Litzenberger, 1992). This paper contributes to the literature by providing a simple theory and supporting evidence on corporate use of interest rate swaps. Existing theories on interest rate swaps (such as Wall, 1989; Arak et al., 1988; Titman, 1992; Minton, 1994) focus on firmsÕ choices between commercial paper and long-term fixed-rate debt. The basic idea of the above models is that firms can benefit from their constantly changing credit quality by rolling over short-term debts at frequent intervals. For example, in one of the earliest models of interest rate swaps, Wall (1989) argues that short-term debts can mitigate agency cost. Titman (1992) argues that short-term debts allow firms to exploit their private information about their future credit situation. The disadvantage of short-term debts, however, is that they expose firms to the fluctuation of interest rates. Interest rate swaps allow firms to benefit from borrowing short-term debts and avoid the associated interest rate risk. The existing theories provide important insights on interest rate swaps, but also have certain limitations. For example, by focusing on debt maturities, they can only explain the combination of short-term loans and interest rate swaps, but cannot explain why firms choose to borrow long-term floating-rate loans and enter interest rate swaps. In this paper, we develop a simply theory on interest rate swaps which is based on the difference between bank loans and public debts. Our theory extends the agency cost/information asymmetry argument by pointing out that banks are good at discerning the credit quality of their borrowers, and more importantly, that restrictive covenants of bank loans is another important way to mitigate agency cost (see Diamond, 1984; Leland and Pyle, 1977; Ramakrishnan and Thakor, 1984). If the borrowerÕs credit quality deteriorates significantly, banks can demand their money back and effectively accelerate the maturity of the debt. Thus, the distinction between long-term and short-term bank loans is not important for mitigating agency problems; restrictive covenants make bank loans short term with respect to their credit risk regardless of their stated maturity. While restrictive covenants of bank loans help reduce agency costs, banks have natural disadvantages in bearing interest rate risk due to their floating liabilities. 3 A firm that wants a fixed-rate loan can borrow a floating-rate loan from a bank and enter an interest rate swap to hedge the interest rate risk. From the borrowerÕs perspective, there maybe certain advantages of borrowing long-term floating-rate loans than rolling over a sequence of short-term loans. One advantage is that there is much higher transaction costs associated with rolling over short-term loans. Another advantage is that a long-term fixed-rate loan can be accelerated only if the firm violates the covenants. This provides a level of protection to the firm against inefficient liquidation of the loan by the bank. Thus, our theory 3

By claiming that banks are not good at bearing interest rate risk, we do not mean that they do not have the ability of managing interest rate risk. In fact, banks are very sophisticated at this task. What we mean is that due to their floating liability, they are at a natural disadvantage of lending fixed-rate loans. All things being equal, banks would charge a higher premium for a fixed-rate loan than would some other investors who have fixed-rate liability.

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1513

provides a natural explanation of the combination of long-term floating-rate loans and interest rate swaps, a phenomenon that is not studied in the existing literature. Given that our model extends and complements existing theories, it makes similar empirical predictions on certain aspects of the swap markets. Therefore, in our empirical study, we focus on testing those predictions that are unique to our theory. One direct implication of our analysis is that fixed-rate swap payers (firms that make fixed payments in a swap) would adopt more bank loans than floating-rate payers (firms that make floating payments in a swap). Our theory is also more relevant in cases where firms borrow long-term floating-rate loans and swap for fixed-rate loans. To test our theory, we empirically examine the financial characteristics of fixedrate and floating-rate swap payers using data published by Swaps Monitor in 1993 and 1994. Consistent with our theory, we find that fixed-rate swap payers have lower credit ratings and higher leverage ratios, and are more likely to borrow from banks. Fixed-rate swap payers are also shown to have significantly higher percentages of long-term floating-rate debt than do floating-rate payers, which highlights the importance of our theory. On the other hand, we find no significant difference in terms of debt maturities and percentage of short-term debts between the two groups of swap users. Overall, the empirical findings appear to support our theory but are not completely consistent with prior models. The paper continues as follows. Section 2 discusses the basic economic model. Section 3 considers firmsÕ borrowing decisions with and without interest rate swaps. Section 4 provides empirical tests of different theories of interest rate swaps, and Section 5 concludes the paper.

2. The model In this model, we consider a three-date ðt ¼ 0; 1; 2Þ, two-period economy in which all investors and firms are risk neutral. We will first discuss how interest rate evolves, then describe the borrowers and investors in the economy. For simplicity, we assume that the risk-free interest rate during the first period is zero. One plus the risk-free interest rate during the second period is a random varie , with possible outcomes of RH and RL , where RH > RL . The probable, denoted as R ability that the high interest rate occurs is pH and pH 1  pH 1 þ ¼ : R RH RL When there is no uncertainty in the interest rate, one plus the risk-free rate during the second period equals R. There are three types of firms in the economy: The super-good (SG), good (G) and bad (B) firms. At t ¼ 0, each firm needs to borrow a fixed amount of money––one dollar––for a project that generates cash flow at t ¼ 2. They can borrow either fixed-rate or floating-rate loans. For the fixed-rate loan, a firm promises to pay back

1514

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

a fixed amount D at t ¼ 2 which includes both the principal and coupon payment. e plus the coupon at t ¼ 2. 4 For the floating-rate loan, a firm promises to pay back R The super-good firm represents the highly rated firms in the real world, and is assumed to be publicly known. It has one great project with extremely high net present value (NPV), which makes it default-free. The good and bad firms represent middle rated firms with different qualities, and the set of projects available to them is their private information. The proportion of each type of firm in the economy (fG for the good and fB for the bad) is common knowledge, and fG þ fB ¼ 1. The good firm has access to two mutually exclusive projects: A safe and a risky project. The safe projectÕs return, which is fixed, equals G at t ¼ 2. The risky projectÕs e , with outcomes of H and L. The probareturn is a random variable, denoted as B bility that H occurs is p, and L is 1  p. The expected cash flow B equals pH þ ð1  pÞL. The bad firm has only one risky project which is identical to the risky project the good firm has. e and B e are independent, and that H > RH > G > B > R > We assume that R RL > L. The second assumption implies that (1) the safe project has higher expected NPV than the risky one, since G > B; (2) the risky project has positive expected NPV, since B > R; (3) the good firm is subject to default risk if it borrows a floating-rate loan, since G < RH ; and (4) the bad firm is subject to default risk regardless of whether it borrows a fixed-rate or a floating-rate loan, since R > L and RL > L. Managers of each firm work for the best interests of their shareholders. Since the types of projects available are fixed for both the super-good and the bad firm, their managers basically decide how to raise fund. However, the manager of the good firm must decide which project to choose and how to finance it. By taking an action a 2 fg; bg, the manager of the good firm chooses the safe ða ¼ gÞ or the risky ða ¼ bÞ project. There is a cost of X associated with the bankruptcy of the firm. Given the amount of debt D, the goal of the manager of the good firm is to maximize the following objective function: ( ) e ðaÞ  D; 0

max½ C e ðaÞ < Dg X ; max E  Prf C e a2fg;bg R R e ðaÞ represents the cash flow from the project, C e ðgÞ ¼ G and C e ðbÞ ¼ B e . The where C first term in the above equation is the expected present value of the equity, and the second term is the expected present value of the cost of financial distress. The easiest way to interpret this objective function is to consider a case in which the manager and the equityholder of the firm are the same person. If the cash flow from the project cannot satisfy the liability, the manager/equityholder will not only 4 For simplicity, we only consider zero-coupon bonds in the model. Although the argument can be applied to coupon bonds, it only increases the complexity of the analysis and provides no additional insights. Since we do not differentiate between rolling over a short-term loan and a long term, floating-rate loan, the argument here can be applied to both cases.

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1515

lose all the equity but also suffer personal distress. His/her reputation as an entrepreneur might be damaged as well. This idea is similar to that of Hirshleifer and Thakor (1992), who argue that managers may have an incentive to pursue relatively safe projects out of concerns for their reputation. Ross (1977) and many others use this kind of modeling in signalling literature. The basic goal is to show that if the firmÕs cash flow becomes too low, it will suffer some sort of distress costs. There are different ways to formulate this idea. However, the linear form adopted here makes the derivation easier and still captures the idea that we are trying to convey. In this paper, we take the firmÕs decision of using debt as a given and do not address the more fundamental question of why firms still use debt, given the agency problem and bankruptcy costs of debt financing. There are two types of lenders in the economy, public investors and commercial banks. We assume that the public investors cannot distinguish between the good and bad firm. They can only tell if a firm is in default or not. Hence if a firm is not in bankruptcy, the public investors cannot tell whether its cash flow is H or G, i.e., whether the firm has taken the risky project but is lucky, or whether it has taken the safe one. We assume that banks can distinguish the good from the bad firm, based on their expertise in credit analysis. Due to their large investments for individual borrowers, banks are also in a better position to monitor and enforce loan agreements. In contrast, public investors with relatively small holdings of debt securities face monitoring costs that can be very high. 5 Therefore, it is natural to assume in this analysis that the public investors do not have the ability to monitor their borrowersÕ behavior, while banks can monitor their borrowersÕ actions effectively. Banks will charge a price, c, for providing such services. Due to their floating liabilities, in an environment of volatile interest rates, banks prefer to lend short-term or floatingrate loans in order to match their assets and liabilities. 6 If they lend long-term fixedrate loans, they would demand a higher premium for bearing the interest rate risk. We use S to represent this premium which equals zero if interest rate is constant. For ease of exposition, we will also use the following notations in the next two sections: G0 ¼ G=R, B0 ¼ B=R, c0 ¼ c=R, S0 ¼ S=R and X0 ¼ X=R. 3. Borrowing with and without interest rate swaps 3.1. Borrowing without swaps: the case of fixed interest rate In this section, we examine the borrowing decisions of the good and the bad firm when the interest rate is constant in the economy. The basic financing decision facing each firm is whether to borrow from public investors or banks. 5 The advantage of bank loans is also reflected in renegotiations of covenant violations. It is much easier to renegotiate with a few well-informed institutional investors, than a large number of public investors, who not only try to improve their collective position versus the debtor but also their position relative to each other (see Berlin and Mester, 1992, for further details). 6 In order to justify such decisions for banks in a risk-neutral economy, we need to introduce the cost of financial distress for the banks too.

1516

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

Suppose that the good firm decides to raise one dollar from the public investors and promises to pay back D at t ¼ 2. According to the above assumptions, it will choose the risky project if ( )   e  D; 0

max½ B X e < Dg > E G  D ;  Prf B E e e R R R where R 6 D, since the public investors would require at least a return of R for their investment. Let D represent the present value of the difference between the expected payoff from the risky and the safe project, which equals ðð1  pÞðD  XÞ þ pH  GÞ=R. D is an increasing function of D and a decreasing function of X, since oD=oD ¼ ð1  pÞ=R > 0 and oD=oX ¼ ð1  pÞ=R < 0. As leverage increases, the good firm is better off by taking the risky project. However, the cost of financial distress might prevent it from doing that. Ultimately, the project that the good firm chooses depends on the leverage ratio and the cost of financial distress. For example, if X > pðH  GÞ=ð1  pÞ, we have pH  G  ð1  pÞX < 0; R pðH  GÞ  ð1  pÞX DjD¼G ¼ < 0: R The implication is that the good firm will never choose the risky project, since the cost of financial distress offsets the gain to the equity by increasing the volatility of the underlying asset. On the other hand, if X < pðH  GÞ=ð1  pÞ, then DjD¼0 < 0 and DjD¼G > 0. Since D is a continuous and increasing function of D, there exists a level of debt D such that D ¼ 0, i.e., ( )   e  D ; 0

max½ B X e < D g ¼ E G  D : E  Prf B e e R R R DjD¼0 ¼

If R > D , then the good firm will always choose the risky project, since the gain to the equity is higher than the cost of financial distress. This requires that D ¼ ðð1  pÞX þ G  pH Þ=ð1  pÞ < R, or equivalently X < R  ððG  pH Þ=ð1  pÞÞ. Next we consider each firmÕs borrowing decisions under the conditions of X > X and X < X , where X ¼ min½pðH  GÞ=ð1  pÞ; R  ððG  pH Þ=ð1  pÞÞ . In the first case where X > X , the good firm will never choose the risky project. Assuming no information asymmetry, the good firm should get the risk-free rate if it borrows from the public investors. However, this is not possible since the public investors cannot distinguish between the good and the bad firm. To compensate for the potential default risk of the bad firm, they will charge any firm that borrows from them a default premium, d1 , which is defined as e ; R þ d1 g þ fG min½G; R þ d1

fB Efmin½ B ¼ 1: R

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1517

Solving the above equation, we get d1 ðfB Þ ¼ fB ð1  pÞðR  LÞ=ð1  ð1  pÞfB Þ, an increasing and continuous function of fB , the proportion of bad firms in the economy. Therefore, by borrowing from the public investors, the good firm will be pooled with the bad and charged a premium of d1 ðfB Þ above the risk-free rate. The present value the good firm gets at t ¼ 0 is G0  1  d10 ðfB Þ, where d10 ðfB Þ ¼ d1 ðfB Þ=R. Instead of borrowing from the public investors, the good firm can borrow from a bank that can differentiate the good from the bad firm by collecting and analyzing information of its borrowers at a cost of c. Since the good firm will always choose the safe project for X > X , it should get the risk-free rate from the bank once the information asymmetry is removed. Borrowing from the bank, the good firm gets G0  1  c0 at t ¼ 0. In equilibrium, the good firm will borrow from the bank if the benefit of distinguishing itself from the bad firm is higher than the cost of the bank loan, i.e. d10 ðfB Þ > c0 , which happens if fB > fB , where fB is defined such that d10 ðfB Þ ¼ c0 . Regardless of whether the bad firm borrows from banks or public investors, its status will be revealed. Banks of course screen their borrowers, and rational public investors would deduce that the firms coming to them must be the bad ones. In either case, the bad firm will be found out and charged a higher rate. The cost c associated with the bank loan will make it stay in the public market, which is probably the junk bond sector of the corporate bond market. Of course in an economy with no welldeveloped junk bond markets, it might be prohibitively expensive for the bad firm to borrow from the public market, in which case it could borrow from the bank as well. However, how the bad firm borrows is not essential here and will not affect the main conclusions of this paper. In the second case where X < X , the good firm might choose the risky project if R > D , which would hurt the bondholders by making their loans more risky. To protect themselves from the potential moral hazard problem, the public investors would charge a premium d2 , which is defined as e ; R þ d2 g Efmin½ B ¼ 1: R Solving the above equation yields d2 ¼ ð1  pÞðR  LÞ=p. As a result, if the good firm borrows from the public investors, it has to pay a premium d2 above the risk-free rate. The present value that the good firm receives at t ¼ 0 is B0  1  ð1  pÞX0 ; the present value of the expected cash flow from the risky project is B0 ¼ ðpH þ ð1  pÞLÞ=R minus the one dollar borrowed; and the present value of the expected cost of financial distress is ð1  pÞX0 . However, the good firm can do better by borrowing from the bank, which would monitor and enforce loan agreement more effectively. We assume that the covenants of the bank loan specify that the firm can only take the safe project. The manager of the good firm takes an action (g or b) to select the project it wants, and then the bank gathers information about the firm at a cost of c. If the bank discovers that the risky project has been taken, the covenant allows the bank to liquidate early, which yields

1518

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

the lender the present value of the one dollar loaned and the borrower zero. 7 Hence if the good firm borrows from the bank, the punishment is severe enough to prevent it from choosing the risky project. We would like to point out that in reality, loan covenants typically specify a series of financial ratio tests that the firms must not violate, and rarely seek to dictate project choices by firms. In our model, we abstract from this reality based on the observation that if a firm chooses risky projects, then the restrictions on financial ratios are more likely to be violated. Our argument holds as long as the covenants exert a significant effect on project choice. Similarly, we also assume perfect monitoring, which means that the information gathered ascertains which project has been selected. By making this simple assumption, we ignore many details in the relation between the bank and its borrower. For example, what is the optimal way for the bank to monitor? What if the bank cannot observe exactly what happened? How can the contract be constructed to provide the bank with the right incentives to do the monitoring? All these concerns deserve separate detailed study and cannot be covered here. Yet the assumption of perfect monitoring does not invalidate our first-order approximation of the fact that banks are better monitors and loan enforcers than general public investors are. By these assumptions, the good firm will always choose the safe project and pay the bank R þ c. The present value the good firm receives at t ¼ 0 is G0  1  c0 . If G0  1  c0 > B0  1  ð1  pÞX0 , or equivalently c0 < G0  B0 þ ð1  pÞX0 , the good firm would prefer to borrow from the bank. For the same reason, the bad firm has no interest in borrowing from the bank. The above analysis leads to the following results. 8 Proposition 1. With constant interest rate in the economy, we have (i) Suppose X > X and c < ð1  pÞðR  LÞ=p. If fB > fB , there exists a separating equilibrium in which the good firm borrows from the bank and the bad firm borrows from the public investors; if fB < fB , there is a pooling equilibrium in which both firms borrow from the public investors. fB is defined such that d10 ðfB Þ ¼ c0 . (ii) If X < X and c0 < G0  B0 þ ð1  pÞX0 , there exists a separating equilibrium in which the good firm borrows from the bank and the bad firm borrows from the public investors. Proof. See Appendix A. The above results establish the borrowing decisions of both firms in an economy with constant interest rate. The good firmÕs decision to borrow from the public investors or from the bank depends on the relative costs of asymmetric information and 7

We follow Diamond (1991) in using this assumption. The case in which c is big is not very interesting for our purposes, since in such a situation no firms would borrow from banks, and there is no need for banks to exist. So we only consider the situations in which c < ð1  pÞðR  LÞ=p or c < G  B þ ð1  pÞX. 8

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1519

the bank loan. If the cost of asymmetric information is too high, then it is beneficial for the firms to borrow from the bank. 3.2. Borrowing without swaps: The case of volatile interest rate Given volatile interest rates, the good firm might not find the bank loan very attractive. Due to its floating liability, the bank would find it easier to lend short-term or floating-rate loans, but for fixed-rate loans, it would demand a higher premium for bearing the interest rate risk. Borrowing from the bank, the good firm can take either a fixed-rate loan with an extra premium S or a floating-rate loan. 9 However, in the latter case, the good firm is subject to default risk because the fixed cash flow generated by the safe project is not enough to pay off the debt if interest rate goes up. If the probability of high interest rate occurring is high, then the cost of financial distress might prevent the good firm from borrowing the floating-rate loan from the bank. Now the good firm faces a dilemma: It can borrow a fixed-rate loan from the public investors but pay a premium due to the perceived credit risk; it can borrow a fixed-rate loan from the bank but pay a premium of S due to the interest rate risk; or it can borrow a floating-rate loan from the bank, which exposes it to the risk of financial distress. For X > X , if the good firm borrows from the public investors, it will be charged R þ d1 ðfB Þ, and the present value it gets at t ¼ 0 is G0  1  d10 ðfB Þ. If it borrows a fixed-rate loan from the bank, in addition to the cost c, it has to pay a premium S, and the present value it receives at t ¼ 0 is G0  1  c0  S0 . If instead it borrows a floating-rate loan from the bank, since it might default when the interest rate goes up, it will be charged a default premium d3 defined as ( ) e þ d3

min½G; R c E ¼1þ : e R R Solving the above equation, we get d3 ¼ ðpH ðRH  GÞRL =ð1  pH ÞRH Þ þ ðRL c= ð1  pH ÞRÞ. The present value the good firm gets at t ¼ 0 from borrowing the floatig-rate bank loan is G0  1  c0  pH X0 , the present value of the equity minus the expected cost of financial distress. In equilibrium, the good firm will borrow from the bank if the benefits of distinguishing it from the bad firm are higher than the cost of the bank loan, which equals c0 þ S0 for the fixed-rate, or c0 þ pH X0 for the floating-rate bank loan (all in present value terms). This happens when d10 ðfB Þ > c0 þ S0 or d10 ðfB Þ > c0 þ pH X0 . Let fB1 be such that d10 ðfB1 Þ ¼ minfc0 þ pH X0 ; c0 þ S0 g. Since d10 ðfB Þ is an increasing function of fB and both pH and S0 are non-negative, we have fB1 P fB . The good firm 9

In our single period world, we do not explicitly model the advantages of borrowing a long-term fixedrate loan over rolling over a series of short-term loans from a bank. One advantage is that there is much higher transaction costs associated with rolling over short-term loans. Another advantage is that a longterm fixed-rate loan can be accelerated only if the firm violates the covenants. This provides a level of protection to the firm against inefficient liquidation of the loan by the bank.

1520

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

will borrow either a fixed-rate or a floating-rate loan from the bank if fB > fB1 . Otherwise, it will stay in the public market. The bad firm will borrow from the public market due to the additional cost associated with the bank loan. If X < X , the good firm will choose the risky project if it borrows from the public investors, and it will get B0  1  ð1  pÞX0 . Borrowing from the bank, which provides monitoring service, the good firm will choose the safe project and get G0  1  c0  S0 if it opts for a fixed-rate loan or G0  1  c0  pH X0 if it wants a floating-rate loan. Let pH be such that G0  1  c0  pH X0 ¼ B0  1  ð1  pÞX0 , and let S0 be such that G0  1  c0  S0 ¼ B0  1  ð1  pÞX0 . When the interest rate is volatile, the good firm no longer benefits uniformly from the bank loan. Hence it wants to borrow from the bank only if pH < pH or S0 < S0 , that is, if either the probability for the interest rate is high or the interest rate risk premium demanded by the bank is sufficiently low. Proposition 2. With volatile interest rate in the economy, we have (i) Suppose X > X and c < ð1  pÞðR  LÞ=p. If fB > fB1 , there exists a separating equilibrium in which the good firm borrows from the bank, and the bad firm borrows from the public investors; if fB < fB1 , there exists a pooling equilibrium in which both firms borrow from the public investors. fB1 is defined such that d10 ðfB1 Þ ¼ minfc0 þ pH X0 ; c0 þ S0 g. (ii) Suppose X < X and c0 < G0  B0 þ ð1  pÞX0 . If pH < pH or S0 < S0 , there exists a separating equilibrium in which the good firm borrows from the bank, and the bad firm borrows from the public investors; if pH > pH and S0 > S0 , there exists a pooling equilibrium in which both firms borrow from the public investors. pH and S0 are defined such that G0  1  c0  pH X0 ¼ B0  1  ð1  pÞX0 and G0  1  c0  S0 ¼ B0  1  ð1  pÞX0 respectively. Proof. See Appendix A. The above results show that an increase in interest rate volatility will raise the cost of the bank loan for the good firm and affect its borrowing decisions. While restrictive covenants of bank loans help mitigate agency costs, banks also have natural disadvantages of lending fixed-rate loans, given their floating liabilities. As a result, the firm cannot reap the benefits of bank loans without being exposed to interest rate risk, given the existing instruments in the economy. 3.3. Borrowing with interest rate swaps The main problem facing the good firm in the economy described above is that it must bear the cost of either credit risk or interest rate risk. Interest rate swap resolves this problem by making it easier to transform a fixed-rate loan into a floating-rate loan or vice versa. Fig. 1 illustrates two different ways that interest rate swap can help reduce the good firmÕs borrowing costs. In the first case, to get the fixed-rate loan it desires,

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1521

Fig. 1. Two possible swap transactions: (a) The good firm borrows a floating-rate loan from the bank and enters an interest rate swap with the super-good firm, which borrows a fixed-rate loan from the public investors. (b) The bank lends a fixed-rate loan to the good firm and enters an interest rate swap with the super-good firm, which borrows a fixed-rate loan from the public investors.

the good firm borrows a floating-rate loan from the bank and simultaneously enters into an interest rate swap with a counterparty that is willing to take the interest rate risk. In the second case, the bank lends a fixed-rate loan to the good firm and enters into a swap itself. In both cases, the obvious candidate for the counterparty is the super-good firm. We assume that in the swap contract, either the good firm or the e , where dS is bank pays R þ dS to the super-good firm in exchange for receiving R 10 the swap spread and dS 2 ð0; G  R  c . Since the bank in the second case has

10 Due to the same modeling considerations mentioned previously, the interest rate swap in this section has only one period and thus is similar to a forward or futures contract. However, in reality, swaps are preferable to forwards or futures because the transaction cost of arranging a swap is much less than that of a package of forward or futures contracts. Also, the swap market is very liquid for maturities up to 10 years, but the liquidity of forward or futures markets is very poor for maturities beyond two or three years. In this model the swap is an agreement between the two counterparties, for the sake of simplicity. But in reality swap transactions are usually intermediated by swap dealers. For the roles played by swap dealers, see Campbell and Kracaw (1991).

1522

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

hedged its interest rate risk, it would not charge the risk premium due to the interest rate risk but would pass the swap spread to the good firm. In both cases the good firm receives G0  1  c0  dS0 at t ¼ 0; where dS0 ¼ dS =R. As shown before, in both scenarios where X > X and X < X , the good firm will always choose the safe project if it borrows from the bank, and since G > R þ c þ dS , the good firm is actually e þ c for the floating-rate loan or R þ c for the default-free. Thus it pays the bank R 11 fixed-rate loan. Since the super-good firm pays public investors R for its loan, it wants to participate in this transaction because it effectively obtains the floating-rate financing at a e  dS , which is lower than the riskless rate. So the super-good firm rate that equals R benefits from the swap. Interest rate swap can simultaneously enable the good firm to reduce its financing costs due to the adverse selection problem and the interest rate risk. This cost reduction is possible in cases where X > X , if dS0 is small enough, i.e., dS0 < min½S0 ; pH X0 , and if fB > fB2 , where fB2 is defined such that d10 ðfB2 Þ ¼ c0 þ dS0 . In both swap transactions in Fig. 1, the good firm gets G0  1  c0  dS0 , a higher NPV than that produced by the firmÕs three other alternatives: (1) borrowing from the public investors, which yields G0  1  c0  d10 ðfB Þ; (2) borrowing a fixed-rate loan from the bank, which yields G0  1  c0  S0 ; and (3) borrowing a floating-rate loan from the bank, which yields G0  1  c0  pH X0 . For X < X , if dS0 is small enough, i.e., dS0 < min½S0 ; G0  B0  c0 þ ð1  pÞX0 , the interest rate swap also helps the good firm to reduce the agency cost and the interest rate risk if pH > pH1 , where pH1 X0 ¼ dS0 . Again the payoff the good firm gets after the introduction of the interest rate swap in Fig. 1, G0  1  c0  dS0 , is better than the payoff of its three alternatives: (1) borrowing from the public market, which yields B0  1  ð1  pÞX0 ; (2) borrowing a fixed-rate loan from the bank, which yields G0  1  c0  S0 ; (3) borrowing a floating-rate loan from the bank, which yields G0  1  c0  pH X0 . The above analysis leads to the following results. Proposition 3. With volatile interest rate in the economy, we have (i) Suppose X > X and c < ð1  pÞðR  LÞ=p. If dS0 < minfS0 ; pH X0 g and fB > fB2 , then the introduction of interest rate swaps would lower the good firm’s borrowing costs. fB2 is defined such that d10 ðfB2 Þ ¼ c0 þ dS0 . (ii) Suppose X < X and c0 < G0  B0 þ ð1  pÞX0 . If dS0 < minfS0 ; G0  B0  c0 þ ð1  pÞX0 g and pH > pH1 , then the introduction of interest rate swaps would lower the good firm’s borrowing costs. pH1 is defined such that pH1 X0 ¼ dS0 .

11

The swap spread, denoted as dS , is part of firm GÕs savings of financing cost and is paid to firm SG to entice it to enter a pay-floating-receive-fix swap with either G or a bank. In a competitive swap market, the bank must offer a spread of dS to SG as well, otherwise SG would have no incentive to enter a swap with it. However, the bank will not have to bear this cost itself, but it will pass it onto G. Therefore the ultimate e þ c in both Fig. 1A and B. lending rate of the bank is R

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1523

Proof. See Appendix A. The above analysis shows that both counterparties can benefit from an interest rate swap. It reduces the good firmÕs borrowing costs due to agency costs/information asymmetry, and interest rate risk, and it helps the super-good firm to obtain floating-rate financing at a rate lower than the risk-free rate. While the analysis in this paper is partial equilibrium in nature, our argument does suggest that the introduction of interest rate swaps could improve the welfare of an economy with volatile interest rates and widespread information asymmetries. 12 Our model extends existing theories on interest rate swaps based on agency cost and information asymmetry. Those theories (such as Wall, 1989; Titman, 1992) implicitly assume that firms face a choice between commercial paper and long-term bond. Short-term loan has the advantage that it can mitigate agency problem or information asymmetry, but the disadvantage that it exposes firms to interest rate risk. Interest rate swaps allow firms to transform short-term loans into long-term fixedrate loans so as to eliminate the interest rate risk. These theories, however, ignore the possibility that for certain firms, it might be very expensive or even impossible to borrow from public market. Our theory fills in this gap by considering the choice between bank loans and market debts. Bank loans generally contain restrictive covenants that allow the bank to accelerate the maturity of the loan if there is material decline in the firmÕs financial condition. Therefore bank loans, irrespective maturity, help mitigating information asymmetry and agency problem. By considering a class of borrowers excluded from prior studies, our theory is able to explain why firms use long-term floating-rate bank loan and interest rate swaps. The specialness of bank loans, especially the fact that commercial banks have natural advantages in dealing with adverse selection and moral hazard problems, and natural disadvantages in bearing interest rate risk, plays an important role in our theory. 13 As a result, we can provide more precise empirical predictions on users of swaps. For example, one immediate implication of our model is that fixed-rate swap payers are more likely to use bank loans than fixed-rate swap payers. 4. Empirical evidence 4.1. Testable hypotheses In Table 1, we present empirical predictions of different theories on interest rate swaps. Based on the principal of comparative advantage, Bicksler and Chen 12

The basic argument of how our theory works in both scenarios as described in Fig. 1. In later part of the paper, we present further discussion on the two scenarios and empirically examine the first swap arrangement. 13 Simplified assumptions are made here to illustrate our point. For example, we assume that banks can perfectly monitor their borrowers, and that the super good firms are default-free so that they can bear all the interest rate risk. While reality is so clear-cut, our argument is valid as long as there is difference in investorsÕ abilities of bearing the two different risks.

1524

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

Table 1 Empirical predictions on differences between fixed- and floating-rate swap payers Financial characteristics of swaps users

Bicksler and Chen (1986)

Wall (1989)

AEGS (1988), Titman (1992)

Our model

Credit rating Likelihood of using bank loans Debt maturity Short-debt/debt Agency costs (debt/equity) Asymmetric information (size)

FIX < FLT NA

FIX < FLT NA

FIX < FLT NA

FIX < FLT FIX > FLT

FIX < FLT FIX < FLT NA NA

FIX < FLT FIX > FLT FIX > FLT NA

FIX < FLT FIX > FLT NA FIX < FLT

NA NA FIX > FLT FIX < FLT

Summary of the predictions of finance theories on corporate use of interest rate swaps. Each column contains the implications of a specific theory. FIX represents fixed-rate swap payers (making fixed payments in a swap contract), and FLT represents floating-rate swap payers (making floating payments in a swap contract). ‘‘NA’’ means the theory does not have any implications on the specific firm characteristic.

(1986) argue that typically, the spread between AAA-rated commercial paper and Arated commercial is somewhat less than the spread between AAA-rated five-year obligation and A-rated obligation. This provides an incentive for an A-rated firm to use the swap market to create a synthetic fixed-rate obligation with an initial spread over AAA rate that is equal to the short-term spread rather than the long-term spread. This discussion suggests that firms with lower (higher) credit ratings are likely to pay fixed (floating) in swaps, and fixed-rate swap payers would use more short-term debt and have shorter debt maturity than floating-rate swap payers. Wall (1989) and Wall and Pringle (1989) propose that firms can reduce their agency costs by taking on short-term debts and protect themselves from interest rate risk by entering into interest rate swaps to make fixed payments. This theory predicts that fixed-rate swap payers should have higher agency costs (proxied by credit rating 14 and debt to equity ratio) and shorter debt maturities, and use more short-term debts. Titman (1992) and Arak et al. (1988) suggest that it is advantageous for a firm that expects a better credit rating to enter into a short-term debt and a swap rather than into a long-term fixed-rate debt with a currently higher interest rate. These models suggest that fixed-rate swap payers should have lower credit ratings, 15 use more short-term debts, have shorter debt maturities, and a higher degree of information asymmetries (proxied by firm size) than would floating-rate swap payers. In our model, we study the impact of interest rate swaps on a firmÕs decision to borrow bank loans and publicly placed debts. We demonstrate that firms with high agency costs and severe information asymmetries would borrow variable-rate loans (either short-term or long-term floating) from banks to reduce their financing costs. 14 Lower rated firms are more likely to have agency problems, since shareholders benefit more from choosing risky projects. 15 Lower rated firms may stand to gain more from an improvement in their credit rating, since the credit spreads are larger and the room for a rating increase is greater.

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1525

These firms would also enter into interest rate swaps as fixed-rate payers to eliminate their exposure to interest rate risk. As shown in Table 1, the model predicts that fixed-rate swap payers should have lower credit ratings and higher agency costs and information asymmetries than would floating-rate payers. They are also more likely to use bank loans. The above discussion shows that different theories tend to make similar predictions. However, there are two specific predictions that distinguish ours from existing theories. First, we can examine whether fixed-rate swap payers use more bank loans. If that is true, then it would support our theory. Second, we can see whether fixedrate swap payers tend to borrow more long-term floating-rate loans. If that is true, then our theory is more relevant than previous theories, which cannot explain this phenomenon. 4.2. The data The information on swap outstanding for interest rate swap users is derived from the ‘‘Database of Users of Derivatives’’, published by Swaps Monitor. The database contains a digest of information about the use of both OTC and exchange-traded derivative instruments by corporations, banks, thrifts, insurance companies, government agencies and other entities in the United States. The information in the database has been derived largely from public sources – principally from annual reports, but also from filings with regulatory agencies. Of course, no outstanding can be shown for companies which use derivatives but make no disclosures of that fact. In this study, we focus on the users of interest rate swaps in 1993 and 1994. There are several empirical studies on interest rate swaps in the existing literature. In one of the earliest studies, Wall and Pringle (1989) test theories on interest rate swaps based on information contained in the footnotes of Annual Reports for the year of 1986 and find supporting evidence for agency theory. Samant (1996), using the COMPACT DISCLOSURE database from June 16, 1990 to June 15, 1992, also finds evidence that supports the agency theory of Wall (1989) and information asymmetry explanation of Titman (1992). Harper and Wingender (2000) test the agency theory using the COMPACT DISCLOSURE database for the years of 1986–1991. With a small sample of 24 observations, they find a significant positive relationship between the risk of the firm and the reduction of agency costs, which are consistent with WallÕs hypothesis and other theories on swaps. While some companies voluntarily disclosed the use of derivatives prior to 1990, only in the beginning of last decade has the Financial Accounting Standard Board (FASB, 1981, 1990, 1991a,b) issued accounting standards concerning the disclosure of off-balance-sheet instruments including derivatives. For example, Statement of Financial Accounting Standard (SFAS) nos. 105 and 107 on notional and fair values of financial instruments, became effective only after 1992. Compared to the previous studies, the data used in this paper is more up-to-date and detailed and lack of sample selection bias, as the data is collected during a period with increased disclosure on

1526

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

derivative uses. Therefore, we have a unique opportunity to study the use of these instruments by non-financial firms in the US. 16 In 1993 and 1994, the total number of non-financial firms that used interest rate swaps and disclose their types (fixed or floating) is 465 and 486 respectively. If the sample is restricted to those firms that are contained in COMPUSTAT, there are 345 and 347 users of swaps for 1993 and 1994 respectively. The sample size here is thus much larger than that used in previous papers. 17 Table 2 contains the summary information of interest rate swap users which includes both the size and direction of their swap positions. Panel A shows that more than 40% of total swap users pay fixed, about 18–20% of them pay floating and another 18% use both types. There are about 21.45% (in 1993) and 17.87% (in 1994) of firms only report the notional amounts of swaps without disclosing their types. Panel B of Table 2 contains information about the outstanding notional values of swap users. The mean swap notional values of all users are $322.54 and $536.47 million in 1993 and 1994 respectively. Panel C shows the summary statistics of swap users with respect to their book values of assets. We find the mean asset value of all swap users is $6971.27 million in 1993 and $6991.82 million in 1994. Panel D shows the industry distribution of all swap users. The financial information on swap users is extracted from COMPUSTAT, which includes a variety of balance sheet and income statement items, detailed debt structure information, and firmsÕ S&P senior debt ratings. We construct the following variables for each swap user from COMPUSTAT to test the hypotheses developed in Section 4: 1. MATY, the proportion of debt that has a maturity of more than three years in total debt; 18 2. SHTD, the ratio of short-term debt to total debt; 3. LFTD, the ratio of long term, floating-rate debt 19 to total debt; 4. DE, the ratio of book value of long-term debt to market value of equity plus book value of long-term debt; 5. SIZE, market value of the equity plus book value of long-term debt; 6. CRR, a dummy variable that equals one if none of a firmÕs credit ratings (senior debt rating, subordinated debt rating, and commercial paper rating) is available and zero otherwise.

16 Due to limited disclosure, this database unfortunately reports only the amounts of swaps for commercial banks without any information on their fixed/floating types. Therefore, we can only focus on non-financial firms for our empirical analysis. 17 For example, the number of swap users in the sample of Visvanathan (1995) is only 168 and 196 in 1992 and 1993 respectively, which is half of the sample size here. 18 This measure is used by Barclay and Smith (1995) to assess debt maturity in analyzing the determinants of corporate debt maturity structure. 19 This is data item 148 in COMPUSTAT, named as long-term debt tied to prime. According to the definitions in COMPUSTAT, item 148 includes long-term debt tied to various floating rates, such as LIBOR rate, Eurodollar rate, and brokerÕs call rate, etc.

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1527

Table 2 Descriptive information of interest rate swap users Panel A: The distribution of the types of swap users Fixed payersa Floating payersb Users of bothc Unknown typed

1993

1994

146 62 63 74

152 70 63 62

(42.32%) (17.97%) (18.26%) (21.45%)

1993 ($ million) Panel B: Summary statistics of the outstanding swap notional values Mean 322.54 Standard deviation 681.35 Minimum 0.30 Maximum 5295.00 Panel C: Summary statistics of the asset values of swap users Mean Standard deviation Minimum Maximum

Panel D: The industry distribution of swap users Agriculture production and services Mining Construction Food and tobacco Textiles and apparel Lumber and furniture Paper and printing Chemicals Petroleum, rubber and plastics Primary and fabricated metals Mechanical and electrical machine Transportation equipment Measuring instruments and machine Transportation and utilities Communication Electrics and gas Wholesale trade Retail trade Financial service Hotels Services Total a

1994 ($ million) 536.47 3039.49 0.10 52700.00

6971.27 22300.36 22.40 251506.00

6991.82 20499.65 54.40 219354.00

1993

1994

2 17 4 19 9 4 22 39 21 19 37 16 11 14 17 22 18 22 8 2 22

2 18 3 18 7 3 21 39 19 17 40 16 13 18 18 27 14 19 9 2 24

345

347

Firms that only use fixed-rate swaps (making fixed payments in a swap contract). Firms that only use floating-rate swaps (making floating payments in a swap contract). c Firms that use both types of swaps. d Firms that only report the existence of swaps without disclosing their types. b

(43.80%) (20.17%) (18.16%) (17.87%)

1528

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

4.3. Empirical findings In this subsection, we test the implications of different theories on interest rate swaps. Since the predictions of each theory are not mutually exclusive, we do not test them separately. Instead, we look at the financial characteristics of different types of swap users to see if they are consistent with the predictions of each theory. The whole sample of swap users is partitioned into three categories: 20 • fixed-rate swap payers––companies that report only fixed-rate swaps (i.e. firms that make fixed payments in a swap contract); • floating-rate swap payers––companies that report only floating-rate swaps (i.e. firms that make floating payments in a swap contract); and • both––companies that report both fixed and floating swaps. We first look at the S&P senior debt rating differences among swap users. As shown in Table 3, users of both types of swaps have the highest credit ratings, and floating-rate swap payers have better credit ratings than fixed-rate swap payers. In both 1993 and 1994, the proportion of users of both types of swaps whose ratings were above investment grade was more than 70%, while the proportions for floating and fixed swap payers were about 54% and 30% respectively. We also notice that the proportion of swap users whose credit ratings are not available is significantly different among the three groups of users. We compute the proportion of users with credit ratings above investment grade among those for whom credit ratings are available. We find that these proportions are 60.3%, 76.2% and 86.8% in 1993 (66.2%, 77.6%, and 100% in 1994) of fixed payers, floating payers and both types users respectively. Wall and Pringle (1989) find similar evidence on credit ratings between fixed and floating payers. The evidence is consistent all theories that fixed-rate swap payers should have lower credit ratings than floating-rate swap payers. 21 Another interesting phenomenon observed is that there is consistently a higher proportion of fixed-rate swap payers whose credit ratings are not available than of the two other groups of swap users. The proportion of fixed-rate swap payers whose credit ratings are not available is more than 50% in both 1993 and 1994, while for floating swap payers and users of both types of swaps, the proportions are about 32% and 16% in 1993 (30% and 24% in 1994) respectively. According to Barclay and Smith (1995), firms that do not have credit ratings are those that use private funds, mainly bank loans. Therefore, lack of credit rating can serve as an indication of using bank loans. This implies that fixed-rate swap payers are more likely to borrow from private sources, most likely banks. This is consistent with the predictions of our model. 22 20

We delete swap users that do not disclose their types. It is worth pointing out that, to test the financial arbitrage argument for interest rate swaps, this evidence is not sufficient. The information on the rates each firm is charged for their swap and debt positions is needed. 22 We find that firms with no senior debt ratings do not have S&P subordinated debt or commercial paper ratings either. 21

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1529

Table 3 Credit ratings of swap users Ratings Year: 1993 AAA AA above BBB above BBþ to C NA Total Year: 1994 AAA AA above BBB above BBþ to C NA Total

Fixed payers 2 (1.4%) 6 (4.1%) 41 (28.1%) (60.3%)a 27 (18.5%) 78 (53.4%) 146

1 (0.7%) 7 (4.6%) 47 (30.9%) (66.2%)a 24 (15.8%) 81 (53.3%) 152

Floating payers

Both types users

2 (3.2%) 7 (11.3%) 32 (51.6%) (76.2%) 10 (16.1%) 20 (32.3%)

0 (0%) 4 (6.4%) 46 (73.0%) (86.8%) 7 (11.1%) 10 (15.9%)

62

63

2 (2.56%) 9 (5.13%) 38 (54.3%) (77.6%) 11 (15.7%) 21 (30.0%)

4 (5.71%) 10 (14.29%) 53 (75.71%) (100%) 0 17 (24.29%)

70

70

Summary of credit ratings of firms that used fixed, floating or both types of swaps in 1993 and 1994. Fixed or floating payers are those firms that use only fixed- or floating-rate swaps. Both types users are those firms that use both types. ‘‘NA’’ means a firmÕs credit rating is not available. Those firms usually use bank loans. a This row represents percentages among the users for which credit ratings are available.

We also test the implications of different theories on swaps by comparing fixed and floating swap payers with respect to the firm characteristics summarized in Table 1. The results of the univariate analysis is contained in Table 4. Unlike previous studies (see Visvanathan, 1995), we find no significant difference in terms of debt maturities between the fixed- and floating-rate swap payers in 1993 and 1994. The P-values of the Wilcox rank sum tests and parametric T tests are mainly insignificant. We also compare the proportion of short-term debt of the two groups of swap payers and find no significant difference. However, we find evidence (significant at the 5% level) that fixed swap payers have a higher proportion of long-term floating-rate debt than floating swap payers in both 1993 and 1994. The evidence obtained so far is not consistent with the existing theories, which imply that fixed-rate swap payers should have shorter debt maturities and use more short-term debt. The fact that fixed swap payers have significantly more long term, floating-rate loans reveals the importance of our theory. We compare the debt to equity ratios of the fixed and floating swap payers to test agency costs arguments. In both 1993 and 1994, we find that fixed swap payers have higher debt to equity ratios than floating swap payers (significant at the 5% level). This is consistent with the predictions of Wall (1989) and our model that fixed-rate swap payers have higher agency costs. The data also shows evidence that fixed-rate swap payers are smaller than floating-rate swap payers, but the difference is not significant.

1530

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

Table 4 Financial characteristics of fixed and floating-rate swap payers Variables

Median (Wilcox test)

Mean (T test)

Fixed

Floating

Difference

P-value

Fixed

Floating

Difference

P-value

0.6744 0.1116 0.3876 0.2861 982.6700 1

0.7390 0.2353 0.2035 0.2082 1345.9400 0

0.0646 0.1237 0.1841 0.0780 363.2700 1

(0.663) (0.245) (0.036) (0.049) (0.120) (0.005)

0.6256 0.2174 0.4175 0.3029 2947.9200 0.5374

0.6554 0.2118 0.2582 0.2437 3996.1100 0.3226

0.0298 0.0056 0.1593 0.0593 1048.1900 0.2148

(0.529) (0.900) (0.039) (0.042) (0.329) (0.004)

0.7418 0.1908 0.3703 0.3011 1215.8000 1

0.8227 0.0791 0.2164 0.2109 1408.6400 0

0.081 0.112 0.154 0.090 192.840 1

(0.154) (0.385) (0.044) (0.063) (0.205) (0.002)

0.6762 0.3034 0.4328 0.3157 3603.3200 0.5290

0.7304 0.1539 0.2949 0.2746 3478.3800 0.3099

0.0541 0.1495 0.1379 0.0411 124.9400 0.2191

(0.179) (0.092) (0.058) (0.077) (0.875) (0.002)

Year:1993 MATY SHTD LFTD DE SIZE CRR

Year:1994 MATY SHTD LFTD DE SIZE CRR

Summary statistics of the financial characteristics of firms using fixed or floating-rate swaps in 1993 and 1994. All variables are measured as of the end of each fiscal year. MATY represents the debt maturity; VARD represents the percentage of variable rate loans in total debt; SHTD and LFTD represent the percentages of short-term and long-term floating-rate loans in total debt respectively; DE represents the debt to equity ratio; SIZE represents the total value of a firm; CRR is a dummy variable that equals one if none of a firmÕs credit ratings (senior debt rating, subordinated debt rating, and commercial paper rating) is available and zero otherwise. The results of the two-sided Wilcox rank sum tests and parametric T tests are reported. P-values are in parentheses.

Since there are no data sources on the amount of bank loans adopted by corporations, we cannot compare the use of bank loans by the two types of swap users directly. According to Barclay and Smith (1995), firms that do not have credit ratings are those use private funds, mainly bank loans. Therefore, lack of credit rating would serve as an indication for using bank loans. Univariate analyses in Table 4 indicate that fixed swap payers are much more likely to be those without credit rating, i.e., they more likely use bank loans than do floating swap payers. Since CRR is a characteristic variable, the univariate test in Table 4 is not appropriate on CRR. To circumvent this problem, we perform an alternative statistical test on the existence and strength of any association between the type of swap users and CRR using contingency tables. As shown in Table 5, we compute the percentage of swap users falling into these 2 2 cells according to the type of swaps (fixed or floating) and the existence or not of credit rating (CRR ¼ 0 or CRR ¼ 1). In 1993 and 1994 respectively, about 40% and 37% of fixed-rate swap payers had no credit rating (i.e., were likely to use bank loans). The Chi-square test which examines the independence for these two variables (swap type and CRR) is highly rejected at the 0.4% and 0.2% levels in 1993 and 1994 respectively. It confirms that there exists a strong relationship between swap type and CRR, and that fixed swap payers are more likely to use bank loans. The evidence obtained so far suggests that fixed-rate swap payers are more leveraged and more likely to use bank loans. They also have more long-term floatingrate loans in their debt as well as lower credit ratings. We do not find evidence that fixed-rate swap payers have more short-term loans and shorter debt maturities than

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1531

Table 5 Contingency table analysis on the type of swap users Year: 1993 Fixed

Floating

Total Chi-square test (P-value) Year: 1994 Fixed

Floating

Total Chi-square test (P-value)

CRR ¼ 0

CRR ¼ 1

67 32.21% 45.89% 61.47%

79 39.98% 54.11% 79.80%

146 72.19%

42 20.19% 67.74% 38.53%

20 9.62% 32.26% 20.20%

62 29.81%

109 52.40%

99 47.60%

208 100%

70 31.53% 46.05% 58.82%

82 36.94% 53.95% 78.85%

152 68.47%

49 22.07% 70.00% 41.18%

21 9.46% 30.00% 21.15%

70 31.53%

119 53.60%

103 46.40%

222 100%

Total

0.004

0.002

We examine the correlation between two characteristic variables, type of swap users (fixed or floating) and CRR, which represents the likelihood that a firm uses bank loans. CRR equals one if a firmÕs credit rating is not available and zero otherwise. The percentages of firms falling in the corresponding cells are reported below. A likelihood ratio Chi-square test is performed to examine the independence of these two variables, and the P value of the null hypothesis is present.

floating-rate swap payers. The evidence supports our theory, which argues that one way firms can reduce their financing costs is to borrow variable-rate loans from banks (either short-term or long-term floating) and enter interest rate swaps as fixed-rate payers. Table 6 presents the results of logit estimations regressing a dichotomous variable that equals one for fixed swap payers and zero for floating swap payers, on the explanatory variables constructed above. For each year, we estimate different logit models with different combinations of explanatory variables. For both 1993 and 1994, a firmÕs debt to equity ratio and its use or not of private loans were important determinants of its decision to use a fixed-rate swap rather than a floating-rate swap. This can be seen from the regression coefficients for these

1532

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

Table 6 Logit regressions explaining the use of fixed vs. floating-rate swaps Variables

Predicted sign

Year: 1993 Intercepta DE

þ

CRR

þ

LFTD

þ

Log-likelihood ratio statisticb P-value

1.7973 (0.047) 0.8919 (0.005)

4.163 0.0413

Year: 1994 Intercepta DE

þ

CRR

þ

LFTD

þ

Log-likelihood ratio statisticb P-value

Pearson correlation coefficients for explanatory variables Year: 1993 CRR DE CRR DE LFTD SHTD

8.224 0.0041

14.820 0.0006

0.9170 (0.003)

1.5814 (0.056) 0.9221 (0.004)

1.8109 (0.022)

5.239 0.0241

2.3577 (0.015) 1.1015 (0.002)

9.618 0.0019

LFTD

10.572 0.0051

2.6048 (0.065) 4.238 0.0395

4.4624 (0.026) 0.8324 (0.039) 1.5700 (0.158) 9.690 0.0350

1.8203 (0.064) 3.829 0.0505

2.7545 (0.020) 0.9106 (0.029) 1.1338 (0.2809) 8.975 0.0460

SHTD

1 0.1862 (0.0259) 0.4553 (0.0001) 0.0422 (0.6931)

0.0962 1 (0.4647) 0.5107 0.0363 1 (0.0001) (0.8437)

CRR

DE

1

Year: 1994 CRR DE LFTD SHTD

1 0.17794 (0.0094) 0.28391 (0.0136) 0.0751 (0.6682)

LFTD

SHTD

1 0.07087 1 (0.5513) 0.4443 0.2000 1 (0.0109) (0.4748)

The dependent variable is equal to one if a firm uses fixed-rate swaps and zero if a firm uses floating-rate swaps. All variables are measured at the beginning of each fiscal year. DE represents the debt to equity ratio; CRR represents the likelihood that a firm uses bank loans. It equals one if a firmÕs credit rating is not available and zero otherwise. a Intercept is not reported since the coefficient is not significant in any regression. b The log-likelihood ratio statistic follows a Chi-squared distribution with degrees of freedom reported in the next row.

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1533

variables, which are positive and statistically significant. The variable of long term, floating-rate debt is positively significant in a univariate logit regression but becomes insignificant in a multivariate logit regression. These results offer supporting evidence for our theory, which argues that firms with high agency costs can reduce their financing costs by borrowing variable-rate loans from banks and entering a swap to pay fixed rate. Table 6 also presents the Spearman Correlation matrix for the explanatory variables. We find that CRR and DE are significantly negatively correlated, CRR and LFTD are significantly positively correlated. This suggests that firms without credit rating tend to have lower debt to equity ratios but use more long term, floating-rate debt. In contrast, there is no significant correlation between CRR and SHTD, suggesting that firms without credit rating do not use more short-term debt. This result fits well with our model. Firms without credit rating are those having difficulty or high costs in raising capital in public due to asymmetric information or agency problem. They tend to borrow from commercial banks to reduce the costs of financing because banks deal better with credit risk. However, due to their floating liabilities, banks are vulnerable to interest rate fluctuation and would prefer to lend at floating rate. Therefore, firms with high financing costs in the capital market would borrow long-term floating loans from commercial banks and enter fixed-payer swap contracts to hedge their interest rate risk. We do not find strong evidence that supports the prediction of existing theories that firms should borrow short-term loans and enter into swaps as fixed payers. For example, we estimate logit models with MATY and SHTD as explanatory variables, which measure debt maturity and the ratio of short-term loan to total debt respectively, and find that neither of them is statistically significant (results not shown). In summary, we examine the differences of financial characteristics between the fixed- and floating-rate swap payers. We find that fixed-rate swap payers are more leveraged and more likely to use bank loans, and that they have higher percentages of long term, floating-rate loans in their total debt. However, there is no significant difference between fixed-rate and floating-rate swap payers in terms of debt maturities and percentage of short-term debt. While the evidence obtained appears to support our theory, it is not totally consistent with previous theories. As shown in Fig. 1, there are two possible ways that the swap transaction can be arranged. In our empirical study, we focus on the first possibility in which a firm borrows a floating-rate loan from the bank and enters a swap. This is because we only have information on non-financial firms that use interest rate swaps, but no information on banks using swaps. While we do not address the issue that which form should prevail in the market, our empirical evidence clearly shows that the first arrangement is indeed an important way in which swap transactions are conducted. With respect to the second arrangement, Gorton and Rosen (1995), based on regulatory filings of commercial banks, show that swap activities are mainly concentrated in large commercial banks and many smaller banks do not engage in swaps. For example, they show that only 2.3% of 8596 commercial banks engaged in interest rate swaps in 1993. There was 97%, 52% and 0.21% of banks whose assets are bigger than $5 billion, between $1 and $5 billion, and between $500 million and $1 billion,

1534

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

respectively, use interest rate swaps. Explaining the patterns of commercial banks use of interest rate swaps, is an interesting issue that we leave for future research. 5. Conclusion In this paper, we have developed a simple theory and supporting evidence on corporate use of interest rate swaps. Unlike existing theories which focus on the difference between commercial paper and long-term fixed-rate debt, our theory is based on the difference between bank loans and public debts. While restrictive covenants of bank loans help reduce agency costs, banks have natural disadvantages in bearing interest rate risk due to their floating liabilities. A firm that wants a fixed-rate loan can borrow a floating-rate loan from a bank and enter an interest rate swap to hedge the interest rate risk. Our theory extends the existing literature by considering a class of borrowers excluded from prior studies and explains why firms use long-term floating-rate bank loans and interest rate swaps. Consistent with our theory, we find empirically that fixed-rate swap payers generally have lower credit ratings, higher leverage ratios, higher percentages of long-term floating-rate loans, and are more likely to use bank loans than floating-rate swap payers. On the other hand, we find little differences on debt maturities and percentages of short-term debt between the two groups of swap users, which is not consistent with existing theories. Acknowledgements We would like to thank Warren Bailey, Kenneth French, Roger Ibbotson, Jonathan Ingersoll Jr., N.R. Prabhala, Anjan Thakor, Sheridan Titman, Larry Wall, the anonymous referees, the associate editor, and seminar participants at Cornell University, London Business School, INSEAD, Penn State University, the University of Michigan, Emory University, the University of Oregon, University of California at Riverside, and Yale School of Management for helpful comments and discussions. We would like to thank the Finance Center at Yale School of Management for financial support in obtaining the data used in this study. Any errors are solely ours. Appendix A Proof of Proposition 1. (i) If X > X , the good firm will always choose the safe project whether it borrows from the public investors or the bank. If both firms borrow from the public market, they have to pay R þ d1 ðfB Þ at t ¼ 2 due to the adverse selection problem, where d1 ðfB Þ ¼ fB ð1  pÞðR  LÞ=ð1  ð1  pÞ fB Þ, an increasing and continuous function of fB . The good firm gets G0  1  d10 ðfB Þ at t ¼ 0, where d10 ðfB Þ ¼ d1 ðfB Þ=R. Since d1 ð0Þ ¼ 0 < c and d1 ð1Þ ¼ ð1  pÞðR  LÞ=p > c, there exists fB such that d1 ðfB Þ ¼ c. If fB > fB , the good firm will borrow from the bank and pay R þ c, and the bad firm will remain in the public market. The public investors, realizing this, will update

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1535

their beliefs that all firms in the public market are bad and charge them R þ ðð1  pÞðR  LÞ=pÞ. This is an equilibrium because, given the public investorsÕ beliefs about each firmÕs strategies, neither the good nor the bad firm wants to deviate from their equilibrium strategies; given each firmÕs borrowing strategies, the investorsÕ beliefs are indeed correct. For the good firm, the bank loan is better, since G0  1  c0 > G0  1  ðð1  pÞðR  LÞ=pRÞ. The bad firm is indifferent to the two choices, since in either case, it will be found out as bad. However, it would borrow from the public market because of the cost associated with the bank loan. If fB < fB , the good firm will not borrow from the bank, since the cost of adverse selection is lower than c. Under such conditions, a pooling equilibrium exists in which both firms borrow from the public market and are charged R þ d1 ðfB Þ. (ii) If X < X , the good firm will choose the risky project if it borrows from the public investors. Therefore, firms that borrow from the public market will be charged a default premium, d2 , which equals ð1  pÞðR  LÞ=p, and receive B0  1  ð1  pÞX0 at t ¼ 0. The good firm will choose the bank loan if the cost, c0 , is small enough, i.e., c0 < G0  B0 þ ð1  pÞX0 . The monitoring services provided by the bank force the good firm to take the safe project. As a result, the good firm pays the bank R þ c for the loan and receives G0  1  c0 at t ¼ 0. The bank loan is a better choices since G0  1  c0 > B0  1  ð1  pÞX0 . The bad firm will remain in the public market due to the same reasons mentioned before. Therefore, a separating equilibrium exists.  Proof of Proposition 2. (i) If X > X , the good firm will choose the safe project whether it borrows from the public investors or the bank. If both firms borrow from the public market, they will be charged a premium of d1 ðfB Þ, and the good firm will receive G0  1  d10 ðfB Þ. If the good firm borrows a fixed-rate bank loan, it has to pay R þ c þ S, where S is the interest risk premium charged by the bank, and it receives e þ d3 , G0  1  c0  S0 at t ¼ 0. If it borrows a floating-rate bank loan, it has to pay R where d3 is the default premium and equals ðpH ðRH  GÞRL =ð1  pH ÞRH Þ þ ðRL c=ð1  pH ÞRÞ. This is because the good firm might default if interest rate goes up. In this case, the good firm receives G0  1  c0  pH X0 at t ¼ 0. Define fB1 such that d1 ðfB1 Þ ¼ minfc þ pH X; c þ Sg. If fB > fB1 , there exists a separating equilibrium in which the good firm borrows from the bank and the bad firm borrows from the public investors. Investors believe that those firms that borrow from the public market are bad. This is an equilibrium, since neither firm wants to deviate from the equilibrium strategies given the investorsÕ beliefs; given each firmÕs borrowing strategies, the investorsÕ beliefs are correct. For the good firm, the bank loan is a good choice, since either G0  1  c0  pH X0 > G0  1  d10 ðfB Þ or G0  1  c0  S0 > G0  1  d10 ðfB Þ. As argued before, the bad firm will not be interested in borrowing from the bank because of the cost of the bank loan. If fB < fB1 , there exists a pooling equilibrium in which both firms borrow from the public market. The good firm will not be interested in seeking either a fixed-rate or a floating-rate bank loan, since the cost of bank loan, which equals c0 þ S0 for the

1536

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

fixed-rate loan and c0 þ pH X0 for the floating-rate loan, is higher than the cost of adverse selection d10 ðfB Þ. (ii) If X < X , the good firm will choose the risky project if it borrows from the public investors. Firms that borrow from the public market will be charged a default premium, d2 , which equals ð1  pÞðR  LÞ=p, and receive B0  1  ð1  pÞX0 . If the good firm borrows a fixed-rate bank loan, it receives G0  1  c0  S0 ; if it borrows a floating-rate bank loan, it receives G0  1  c0  pH X0 . Define S0 and pH such that G0  1  c0  S0 ¼ B0  1  ð1  pÞX0 , and G0  1  c0  pH X0 ¼ B0  1  ð1  pÞX0 . If S < S0 or pH < pH , then either G0  1  c0  S0 > B0  1  ð1  pÞX0 or G0  1  c0  pH X0 > B0  1  ð1  pÞX0 . Thus the good firm will borrow from the bank. As before, the bad firm will remain in the public market, and a separating equilibrium exists. If S > S0 and pH > pH , then both G0  1  c0  S0 < B0  1  ð1  pÞX0 and G0  1  c0  pH X0 < B0  1  ð1  pÞX0 , and both firms borrow from the public market since the cost of bank loan is higher than agency cost. Thus, a pooling equilibrium exists.  Proof of Proposition 3. (i) In the two swap transactions in Fig. 1, the good firm has a fixed-rate liability of R þ c þ dS and gets G0  1  c0  dS0 , where dS0 ¼ dS =R. Since the good firm always takes the safe project for X < X and G > R þ c þ dS , it is default-free. As a result, it can get the fixed-rate and the floating-rate bank loan at e þ c, respectively. Hence the swap transactions are indeed viable. R þ c þ dS and R 0 Choose dS such that 0 < dS0 < minfS0 ; pH X0 g and define fB2 such that d10 ðfB2 Þ ¼ c0 þ dS0 , then fB < fB2 < fB1 . If fB > fB2 , the three financing alternatives available to the good firm without swaps all yield worse payoffs than what the good firm gets in the two swap transactions in Fig. 1: (1) borrowing from the public market, the good firm gets G0  1  d10 ðfB Þ, which is lower than G0  1  c0  dS0 for fB > fB2 ; (2) borrowing a fixed-rate bank loan, the good firm gets G0  1  c0  S0 , which is smaller than G0  1  c0  dS0 ; (3) borrowing a floating-rate bank loan, the good firm gets G0  1  c0  pH X0 , which is also smaller than G0  1  c0  dS0 . Therefore, if there is a sufficiently high proportion of bad firms in the economy and the swap spread is small enough, the introduction of interest rate swaps will make the good firm better off. For the same reason as before, the bad firm will borrow from the public market, and a separating equilibrium exists. (ii) The two swap transactions in Fig. 1 again are viable if dS 2 ð0; G  R  c . Borrowing from the bank, the good firm is forced to choose the safe project. Since e þ c and R þ c þ dS for G > R þ c þ dS , the good firm is default-free, and it can get R a floating-rate and a fixed-rate bank loan respectively. In both cases the good firm gets G0  1  c0  dS0 . Choose dS0 such that 0 < dS0 < minfS0 ; G0  B0  c0 þ ð1  pÞX0 g and define pH1 such that pH1 X0 ¼ dS0 . If pH > pH1 , the three financing alternatives available to the good firm without swaps all yield worse payoffs than what the good firm gets in the two swap transactions in Fig. 1: (1) borrowing a fixed-rate loan from the public market, the good firm will choose the risky project and get B0  1  ð1  pÞX0 , which is smaller than G0  1  c0  dS0 ; (2) borrowing a fixed-rate loan from the bank, the

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

1537

good firm gets G0  1  c0  S0 , which is smaller than G0  1  c0  dS0 ; and (3) borrowing a floating-rate loan from the bank, the good firm gets G0  1  c0  pH X0 , which is also smaller than G0  1  c0  dS0 , for pH > pH1 . Therefore, if the probability of high interest rate occurring is high, and if the swap spread is small enough, the introduction of interest rate swaps makes the good firm better off. The bad firm will choose to borrow from the public market for the same reasons given before, and a separating equilibrium obtains. 

References Arak, M., Estrella, A., Goodman, L., Silver, A., 1988. Interest rate swaps: An alternative explanation. Financial Management 17, 12–18. Barclay, M.J., Smith, C.W., 1995. The maturity structure of corporate debt. Journal of Finance 50, 609– 631. Berlin, M., Mester, L., 1992. Debt covenants and renegotiation. Journal of Financial Intermediation 2, 95–133. Bicksler, J., Chen, A.H., 1986. An economic analysis of swaps. Journal of Finance 41, 645–655. Campbell, T., Kracaw, W., 1991. Intermediation and the market for interest rate swaps. Journal of Financial Intermediation 1, 362–384. Diamond, D., 1984. Financial intermediation and delegated monitoring. Review of Economic Studies 51, 393–414. Diamond, D., 1991. Monitoring and reputation: The choice between bank loans and directly placed debt. Journal of Political Economy 99, 688–721. Duffie, D., Huang, M., 1996. Swap rates and credit quality. Journal of Finance 51, 921–949. FASB, 1981. SFAS no. 47, Disclosure of long term obligations. Financial Accounting Standard Board, Norwalk, CT. FASB, 1990. SFAS no. 105, Disclosure of information about financial instruments with off-balance-sheet and financial instruments with concentrations of credit risk. Financial Accounting Standard Board, Norwalk, CT. FASB, 1991a. SFAS no. 107, Disclosure about fair value of financial instruments. Financial Accounting Standard Board, Norwalk, CT. FASB, 1991b. SFAS no. 107, Disclosure about derivative financial instruments and fair value of financial instruments. Financial Accounting Standard Board, Norwalk, CT. Gorton, G., Rosen, R., 1995. Banks and derivatives, NBER Working paper #5100. Harper, J.T., Wingender, J.R., 2000. An empirical test of agency cost reduction using interest rate swaps. Journal of Banking and Finance 24, 1419–1431. Hirshleifer, D., Thakor, A., 1992. Managerial conservatism, project choice and debt. Review of Financial Studies 5, 437–470. Leland, H.E., Pyle, D.H., 1977. Informational asymmetries, financial structure, and financial intermediation. Journal of Finance 32, 371–387. Li, H., 1998. Pricing of swaps with default risk. Review of Derivatives Research 2, 231–250. Litzenberger, R., 1992. Swaps: Plain and fanciful. Journal of Finance 47, 831–850. Minton, B., 1994. Interest rate derivative products and firmÕs borrowing decisions: The case of interest rate swaps and short-term interest rate futures contracts, Manuscript, Ohio State University. Ramakrishnan, R., Thakor, A., 1984. Information reliability and a theory of financial intermediation. Review of Economic Studies, 415–432. Ross, S., 1977. The determination of financial structure: The incentive-signalling approach. Bell Journal of Economics, 23–40. Samant, A., 1996. An empirical study of interest rate swap usage by nonfinancial corporate business. Journals of Financial Service Research 10, 43–57.

1538

H. Li, C.X. Mao / Journal of Banking & Finance 27 (2003) 1511–1538

Smith, C.W., Smithson, C., Wakeman, L.W., 1986. The evolving market for swaps. Midland Corporate Finance Review, 20–32. Smith, C.W., Smithson, C., Wakeman, L.W., 1988. The market for interest rate swaps. Financial Management, 34–44. Titman, S., 1992. Interest rate swaps and corporate financing choices. Journal of Finance 47, 1503–1516. Visvanathan, G., 1995. Who uses interest rate swaps? A cross sectional analysis, Working Paper, George Washington University. Wall, L.D., 1989. Interest rate swaps in an agency theoretic model with uncertain interest rates. Journal of Banking and Finance 13, 261–270. Wall, L.D., Pringle, J., 1988. Interest rate swaps: A review of the issues. Economic Review of the Federal Reserve Bank of Atlanta, 22–40. Wall, L., Pringle, J., 1989. Alternative explanations of interest rate swaps: A theoretical and empirical analysis. Financial Management 18, 59–73.

Suggest Documents