Who Uses Interest Rate Swaps? A Cross-Sectional Analysis GNANAKUMAR VISVANATHAN”

Empirical analysis in this study examines factors that explain the use of interest rate swaps by nonfinancial firms in the Standard & Poor’s 500. Consistent with asymmetric information and agency cost theories, firms with signifcant expected financial distress costs use swaps to transform short-term debt into long-term fixed-rate debt. Debt maturity structure, but not interest rate sensitivity, is sign@cant in the decision to use a swap. Credit quality differentials or expectations of improving financial prospects are not signifcant in distinguishing among swap users.

1. Introduction This study focuses on the motives and characteristics of firms that use interest rate swaps. The use of financial instruments, including interest rate swaps, has increased in recent years. The dramatic increase in and the consequences of the use of such instruments have become a focus of regulatory concern (GAO [ 19941; SEC [ 19951). Regulators are concerned about the use of risky financial instruments; in particular, whether such instruments are used for speculative purposes. Illustrating this concern, a study by the United States General Accounting Office (GAO [ 19941) states: “Congress, federal regulators, and some members of the industry are concerned about these products and the risks they may pose to the financial system, individual firms, investors, and U.S. taxpayers. These concerns have been heightened by recent reports of substantial losses by some derivatives end-users.’’I A burgeoning literature, both theoretical and empirical, has emerged to address the choice of derivative instruments and the characteristics of firms that are likely *Accounting Department, The George Washington University. This paper is based on a portion of my dissertation completed at New York University. 1 wish to thank the members of my dissertation committee Paul Zarowin, Joshua Livnat, Steve Ryan, Eli Bartov, and Steve Figlewski for their guidance. The paper has benefited from comments of workshop participants at University of California at Irvine, George Washington University, Georgetown University, University of Maryland, State University of New York at Buffalo, Baruch College, University of Illinois at Chicago, and the 1997 JAAF-KPMG Conference. I am indebted to N. R. Prabhala and Lany Wall for valuable comments. Comments of Cathy Schrand (discussant) and Jeff Callen (the editor) have significantly improved the paper. 1. Several published reports have drawn attention to the significant losses reported on speculative derivatives by corporations such as Procter and Gamble and Gibson Greetings (Insriturional Invesror [1994]). These firms allegedly used some of their interest rate swaps to speculate on the direction of interest rate movements.

I73

174

JOURNAL OF ACCOUNTING, AUDITING & FINANCE

to use such instruments. This study contributes to the literature by examining in detail the determinants of the use of one instrument, the interest rate swap, which has not been the primary focus of other studies. A rich set of theories deals with interest rate risk management. Theoretical explanations range from asymmetric information and agency cost explanations to hedging motivations. As shown subsequently, the primary derivative instrument used by nonfinancial firms for interest rate risk management is the interest rate swap. Accordingly, theoretical predictions are tested on the nonfinancial firms in the Standard & Poor’s Index of 500 firms for the years 1992 and 1993, using disclosures on interest rate swaps (and other derivatives) required under Statement of Financial Accounting Standards (SFAS) No. 105 (FASB [ 1990al) and SFAS No. 107 (FASB [ 1991]).’Cross-sectional analysis is conducted for firms that disclose the use of a specific type of interest rate swap, fixed or variable, and for firms that do not report any swaps. Results are consistent with predictions of theories that hypothesize the use of ‘ ‘fixed-rate swaps” (make fixed-rate payments and receive payments based on a variable rate) to the extent that firms that use these swaps have higher expected financial distress costs. Predictionsregarding credit quality, as represented by credit ratings for firms that use “variable rate swaps” (make variable rate payments and receive payments based on a fixed rate) are mixed. Results also indicate that interest rate sensitivity is not asignificantdeteminant in the use of interest rate swaps. Evidence in studies of financial instruments is broadly consistent with hedging theories. Two early descriptive studies are Nance, Smith, and Smithson (1993) and Francis and Stephan (1993). Geczy, Minton, and Schrand (1997); Mian (1996); and Guay (1 997) consider large samples and use time periods that require the disclosure of derivative instruments. Geczy et al. examine determinants of the use of currency derivatives by considering a sample of Fortune 500 firms with ex ante foreign exchange rate exposure and test several hedging theories. Mian tests hedging theories with a large sample of firms that use currency and interest rate derivatives. Guay uses a sample of first-time users of derivatives and finds that firm risk declines in the first year of derivative use. In an industry-specific study, Tufano (1996) examines a sample of gold mining firms to test hedging theories and the extent of hedging. All these papers find their results to be supportive of some, if not all, hedging theories, although some of the results differ on specific variables. Fenn et al. (1996), using a sample of nonfinancial firms for one year, show a cross-sectional relation between a firm’s swap position and the amount of variable rate debt in the capital structure, but find no support for a relation between a firm’s swap position and interest rate sensitivity.’

2. Three accounting standards issued during 1991-95 concern the disclosure of off-balance-sheet instruments, including derivatives. These were statements numbered 105, 107, and 119. SFAS No. I19 (FASB [1994]) became effective for most entities in 1994; this standard requires firms to distinguish between derivatives held for trading and held for purposes other than trading. 3. Their study considers about 4,000 nonfinancial firms in the year 1994 and, in the empirical tests, uses a final sample of about 130 firms that report swaps. They also use the notional value of the swap (scaled by total assets) as the independent variable, with variable rate swaps taking negative values.

INTEREST RATE SWAPS

175

This study differs from others in several respects. First, this study examines the use of one derivative in detail, which is important because incentives to use derivatives are likely to differ across the type of instrument used. Second, because the decision to use a swap is linked to debt maturity choice, the issue of endogeneity with respect to the choice of debt maturity structure and the use of a specific type of swap is empirically tested. Hedging explanations are considered in conjunction with debt maturity structure theories. Results offer support for the conjecture that firms consider debt maturity structure in using interest rate swaps. Third, in contrast to studies that address the use of swaps in general, this study distinguishes the type of swap (as variable or fixed) in testing theoretical predictions about signaling, differences in credit ratings, financial distress costs, and mismatches in asset and debt mat~rities.~ The remainder of the paper is organized as follows. Section 2 discusses hypotheses based on various explanations advanced for the use of interest rate swaps. Section 3 describes the data sources and classification. Section 4 presents the empirical results and Section 5 summarizes the paper.

2. Hypotheses Development-Alternative Explanations for Swaps5 A swap contract obligates two parties to exchange, or swap, cash flows. In an interest rate swap, cash flows to be swapped are specified in terms of interest rate differentials. Two types of interest rate swaps are used frequently: fixed rate and variable rate. Typically, in a fixed (variable) rate swap, periodic payments at an agreed fixed (variable) rate are made, and in turn variable (fixed) rate payments are received based on an agreed index. In practice, only the difference between the two amounts is paid or received. In the following sections, the expression “fixed (variable) interest rate swap” is used when the firm is making fixed (variable) rate payments and receiving payments based on a variable (fixed) rate under the swap. Theoretical explanations relevant to this study are applicable to interest rate risk management in general. Swaps are one of the means by which interest rate risk management can be achieved, and several theoretical studies focus primarily on swaps. Litzenberger (1992) and Kuprianov (1994) discuss several of the reasons firms engage in interest rate swaps. These theories are used to develop the hypotheses.

4. Fenn et al. (1996) do not distinguish and test theories that are specific to different types of swaps. Also, they do not consider the role of predictions based on hedging theories. Their analysis is predicated on the debt maturity structure theories alone. 5. It is assumed that the swap is a choice variable for the companies. Exceptions to this were found in the footnotes to two companies’ annual reports, which state that the bankers for the firms require them to enter into swaps (Interlake and Stone Container Corporation).

176

JOURNAL OF ACCOUNTING, AUDITING & FINANCE

2.1 Incentives to Borrow Short-Term and Swap into a Long-Term Obligation

A set of theories explains the use of swaps by addressing incentives for firms to borrow short-term debt and then use interest rate swaps to create “synthetic” long-term fixed-rate debt. In these settings, firms desire long-term fixed-rate debt, but can obtain such debt only at a high interest rate because of agency costs or asymmetry of information. These theories of the use of swaps are discussed in Section 2.1.1, with a focus on the types of costs that create incentives for using short-term debt in combination with swaps rather than long-term debt. Theories on swaps make the implicit assumption that firms prefer long-term debt to short-term debt. Therefore, the theories suggest that firms consider debt maturity structure in the decision to use a swap. Empirically, the simultaneity of these two decisions is considered. Thus, in Section 2.1.2, variables that are related to debt maturity choice are discussed. In this context, it is useful to note that there is some overlap between the variables predicted here and those used to test the theories of swaps use. This is addressed in the empirical analysis. The assumption that firms prefer long-term fixed-rate debt to short-term debt has two additional implications for the analysis of swaps use. First, this assumption implies that short-term debt in fact creates more interest rate risk than long-term debt. However, this assumption is true only if the other cash flows of the firm, unrelated to debt, are fixed with respect to interest rate changes. For some firms, cash inflows may be very sensitive to interest rates, in which case short-term debt that varies with interest rates reduces the overall interest rate exposure of the firm as a whole. The role of interest rate sensitivity on the analysis is discussed in Section 2.1.3. Finally, the theories of swaps used assume that firms do not want the interest rate risk that is created by choosing short-term debt rather than long-term debt. Underlying this assumption is the notion that interest rate risk is costly to the firm. Theories of optimal hedging in general have provided explanations for the costs associated with cash flow volatility such as volatility resulting from interest rate exposure. These theories are discussed in Section 2.1.4. 2.1.1 DECISION TO USE A SWAP

A conventional explanation for the use of interest rate swaps points to imperfections in credit markets that could be exploited to achieve reductions in cost of debt (Goodman [ 19931). Bicksler and Chen (1986) argue that segmented markets provide an explanation of why interest rate swaps exist, citing the presence of ‘‘quality spread differentials” to support this explanation. Quality spread differentials are the differences in spreads between what a lower-quality credit must pay over a higher-quality credit for funds of the same denomination and maturity. Such spreads are observed to be typically increasing in maturity (Wall and Pringle [ 19891). Bicksler and Chen argue that quality spread differentials provide an in-

INTEREST RATE SWAPS

177

centive for low-credit-quality firms to use interest rate swaps to create a synthetic fixed-rate debt at a lower rate than it would have obtained by borrowing long-term debt dire~tly.~.’ This view suggests that firms that use fixed interest rate swaps typically have low credit ratings. To test this proposition, two variables are used following Barclay and Smith (1 995) and Stohs and Maw (1 996). The first is a numeric variable for S&P bond rating, and the second is a dummy variable that represents the availability of bond ratings. These variables are described in Table 1 Recent models that consider the decision to borrow short term and use a swap can be broadly classified into asymmetric information and agency costs models. Flannery (1986) and Diamond (1991) show that borrowers with private information may prefer short-term loans when they possess favorable information about their future prospects. When prospects improve, they will benefit from lower interest costs-a benefit that would have been lost had they borrowed long term at a fixed rate. Note that this analysis does not consider interest rate uncertainty. Titman ( 1992) extends Flannery’s analysis by considering interest rate uncertainty and financial distress costs. In Titman’s model, when firms have favorable information about future prospects, borrowing short term exposes them to interest rate risk that increases the likelihood of financial distress. Titman shows that firms can solve this problem by borrowing short term and swapping for long-term fixed-rate obligation through an interest rate swap, thus minimizing the costs of financial distress. Similar reasoning is employed by Arak, Estrella, Goodman, and Silver (1988), who suggest that it is advantageous for a firm that expects its credit rating to improve in the future to enter into a short-term debt and a swap concurrently rather than into a long-term debt with higher fixed interest rates. Myers (1977) argues that firms with high growth options may not fully invest in positive net present value projects in all states of the world (the “underinvestment” problem), and Jensen and Meckling (1976) argue that shareholders may invest in high-risk projects in certain states of the world, to the detriment of creditors who lend long term. Relying on this reasoning, Wall (1989) argues that, given the riskiness of long-term fixed-rate debt, lenders would require a higher rate to compensate for the risk. Borrowers can reduce this agency cost by taking on short6. The following stylized example illustrates this explanation: A company rated AAA can obtain (or variable) rate at LIBOR (London Inter-bank Offer Rate) + 0.25 percent. A BBB rated company can get a fixed rate loan at 12 percent and floating rate at LIBOR + 0.75%. The quality spread differential in this example is 0.70 percent (the difference in the cost of funds to the AAA company and the BBB company in raising fixed-rate funds less than the difference in costs in raising floating rate funds; i.e., 1.20 percent less 0.50 percent). These two companies can achieve savings by swapping. The BBB company can obtain a floating rate loan and then swap it for a fixed-rate loan, paying 10.9 percent to the AAA company, which pays LIBOR to the BBB company. Both save 0.35 percent. While this example seems to suggest an arbitrage opportunity, Smith, Smithson, and Wakeman (1986) discuss a variety of hidden costs such as callability premiums in a fixed-rate long-term debt that are not taken into consideration in this setting. 7. How default risk and credit quality affect the value of the swap is not a focus of the study. For a detailed theoretical treatment of this issue, see Cooper and Mello (1991) and Duffie and Huang (1996). a fixed rate loan at 10.8 percent and floating

TABLE 1 Description of Variables Description of variables used in the study. All variables are measured as of the beginning of the fiscal year unless mentioned otherwise. Variable DE STM LTM FABE FEPR BRAT DRAT DMAT IS

AMAT

CPR

VPR INDD DER LMV MB BETA NOL DOWN INTF

PRED

Description Debt-to-equity ratio, measured as the ratio of long-term debt plus short-term debt divided by the total of long-term debt, short-term debt, and market value of equity. Short-term debt as a proportion of market value of the firm. Short-term debt is debt (including long-term debt) that matures within a year. Long-term debt as a proportion of market value of the firm. Abnormal earnings measured as earnings per share at time r + I less earnings per share at time r divided by price per share at time 1. Cumulative abnormal earnings over the three-year period beginning after the current year. Numeric variable for S&P bond rating. This is set equal to I for rating of AAA through 27 for CCC or below and is set equal to 28 for firms that are not rated. Dummy variable set equal to 1 for firms with no S&P bond rating; equals 0 otherwise. Debt that matures in more than two years as a proportion of total (short- and long-term) debt. Coefficient on interest rates from a regression of quarterly per share earnings before interest and taxes on the interest rate on the 10-year T-bond for that quarter, for each firm over the 1984-1992 period (for the 1993 sample). Asset maturity is the (book) value-weighted average of the maturities of current assets and net property, plant, and equipment. Maturity of current assets is measured as current assets divided by cost of goods sold. Maturity of net property, plant, and equipment is that amount divided by annual depreciation expense (Stohs and Maur [ 19961). Dummy variable that takes the value of I if the difference between actual and predicted DMAT is greater than the median value for the sample. CPR is set equal to 0 for values below median. Dummy variable that equals I if AMAT is below the median and DMAT is above the median for that sample; VPR is set equal to 0 otherwise. Dummy variable that takes the value of 1 if the firm belongs to a regulated industry (primarily utilities) and 0 otherwise. Dummy variable that equals I if the firm uses derivatives other than interest rate swaps such as forward contracts, options, and futures. Log of market value of firm used to represent firm size. This is measured as market value of equity plus book value of long-term and short-term debt. Ratio of market value of assets to book value of assets. Market value of assets is estimated as book value of assets minus book value of equity plus market value of equity. Common stock beta, measured over the five-year period prior to the current year. Net operating losses outstanding scaled by market value of equity. Log of market value of stock held by insiders in the firm as reported in the Value Line Investment Survey. = (Controls) + u , Q,, + a2 Represents coefficient a3 from the following regression: (fK),, (S/K),, + u3 (CFIK),,, where I is capital expenditure in year r ; Q is a proxy for Tobin’s Q measured as (market value of common stock + long-term debt + current 1iabilities)hotal assets; S is sales in year t ; CF represents income before extraordinary items plus depreciation and amortization; K is a proxy for capital stock measured as total assets at the beginning of the year; Controls are firm and year dummies to control for firm specific and time specific effects. Eighteen years of data (1975-92 for the 1993 sample) are used to estimate this equation. Predicted value of DMAT from the following regression: DMAT, = a, + u2 MB, + u3 FABE, + u4 INDD, + a, LMV, + a6 DE, + u, INTF, + uI NOL, + a, DOWN, + a,, IS,.

INTEREST RATE SWAPS

179

term debt, which reduces agency costs, and protecting themselves from interest rate risk by entering into a swap. Note that while Tit-man uses arguments based on asymmetric information and Wall uses arguments based on agency costs explanation, both their analyses result in a similar outcome, namely, incentives to borrow short term and swap for a long-term fixed-rate borrowing. The following empirical inferences result from the theoretical models. From the Titman analysis, it follows that firms that borrow short term and swap for long-term fixed-rate obligations are likely to be firms with better expected future prospects. Such firms will use swaps, however, only if exposure to interest rate risk could result in financial distress costs. Accordingly, in comparison to firms that do not use interest rate swaps, firms that use fixed interest rate swaps are expected to have higher expected financial distress costs. Following Barclay and Smith (1995) and Stohs and Maur (1996), subsequent period abnormal earnings is used to measure expected improvement in future prospects.8 To represent expected costs of financial distress, a measure of firm leverage is used (Press and Weintrop [ 19901show that firms with high debt levels are more likely to encounter debt covenant violations).

2.1.2 DEBT MATURITY STRUCTURE Barclay and Smith (1995) consider debt maturity choice theories. Following Myers (1977), they argue that a solution to the underinvestment problem for a firm with growth options is to shorten its effective maturity of debt. They test this by examining the relationship between debt maturity choice and investment opportunity set, regulation, firm quality, and term structure of interest rates. Their results show that debt maturity structure is negatively associated with market-to-book ratio (a proxy for growth) and abnormal earnings and is positively associated with size, regulation, and term structure. Stohs and Maur (1996) also test the determinants of debt maturity structure by constructing a detailed measure of debt maturity. Following Diamond (1991), they argue that, because liquidity risk increases with leverage, firms with higher leverage would be expected to use more long-term debt, other things being equal. Stohs and Maur suggest that the significance of marketto-book ratio in the results of Barclay and Smith may be the result of omission of leverage. Accordingly, in addition to the variables used by Barclay and Smith, leverage is also considered. To address the simultaneity of debt maturity structure decision and the decision to use a swap, the predicted value of debt maturity structure is used as an additional variable in testing the characteristics of firms that use fixed interest rate swaps, and the predicted probability of using a fixed interest rate swap (from the empirical model, which does not include debt maturity struc8. To test an ex ante prediction it is necessary to know the time at which the swap transaction began. Such data, however, are disclosed in only a few instances by firms in the sample. In those few cases, disclosures suggest that the swap has been entered into in the current year or within the preceding two years. Accordingly, rather than use one-year-ahead abnormal earnings, cumulative abnormal earnings for the three-year period subsequent to the reporting year are used in the empirical results.

180

JOURNAL OF ACCOUNTING, AUDITING & FINANCE

ture) is used in the debt maturity structure model to ascertain whether the debt maturity structure choice takes into consideration the decision to use a swap. Debt that matures in more than two years as a proportion of total debt is used as the measure of debt maturity structure. This measure is similar to the one used by Barclay and Smith?

2.1.3 INTEREST RATE EXPOSURE In considering debt maturity structure, an additional variable that is likely to be relevant is the interest rate exposure of the firm’s operating activities. Specifically, assuming firms would like to minimize variance of income (net of interest payments), firms with income before interest and taxes that are positively (negatively) correlated with changes in interest rates would obtain variable (fixed) rate financing (Morris [1976]).10Testing this proposition requires an ex ante measure of interest rate exposure. This, however, is complicated because earnings or stock returns are affected by the use of interest rate swaps. Present accounting rules require the effect of the swap to be adjusted to the interest expense. The use of derivatives such as swaps is also likely to alter the stock price interest rate sensitivity (Schrand [1997]). Accordingly, a measure of interest rate exposure that is based on earnings before interest and taxes is used.” For this purpose, quarterly per share earnings before interest and taxes are regressed on the interest rate on the 10-year T-bond for that quarter, for each firm during the 1984-1992 period (for the 1993 sample). The coefficient on the interest rate from this regression is the measure of interest rate exposure.’2

2.1.4 INCENTIVES TO HEDGE Whether or not a firm desires to manage its interest rate exposure, if any, by borrowing short term or long term at fixed rate depends on the firm’s incentives to hedge. Theoretically, financial policies such as hedging are value relevant only in the presence of transaction costs. Smith and Stulz (1985) and Duffie (1993) 9. Where Compustat data are unavailable, this information is collected from the footnotes. Under the provisions of SFAS No. 47 (FASB [19Sl]), repayment of debt over the next five years is disclosed in footnotes, but not for specific years beyond the fifth year. 10. The following excerpt from the annual report of AMR Corporation is illustrative of sensitivity to interest rates and the use of swaps: “Because American’s operating results tend to be better in economic cycles with relatively high interest rates and its capital instruments tend to be financed with long term fixed-rate instruments, interest rate swaps in which American pays the floating rate and receives the fixed rate are used to reduce the impact of economic cycles on American’s net income.” 11. In a study on interest rate risk management by banking firms, Ahmed et al. (1997) find that such firms focus on managing interest rate sensitivity of net income rather than on the interest rate sensitivity of stock returns. 12. Several other measures are used in the hedging literature to measure exposure. Gezcy et al. (1997) use ratio of foreign income to total sales, ratio foreign sales (assets) to total sales (assets), and other related ratios to measure a firm’s exposure to foreign exchange rate risk. Hentschel and Kothari (1995) relate stock returns to exchange rates and interest rates to develop a measure of exposure.

INTEREST RATE SWAPS

181

suggest three types of frictions as determinants of hedging policy: bankruptcy costs, taxes, and agency ~ 0 s t s . IAlso, ~ Froot, Scharfstein, and Stein (1993) suggest dependence on internal funds as another determinant of hedging. Each of these is discussed next.14 Costs ofjinancial distress: Smith and Stulz (1985) argue that hedging reduces costs of financial distress by reducing the variance of firm value. The probability of financial distress depends on the amount of fixed claims, such as debt, that exist on the assets of the firm. As the proportion of fixed claims increases, so do the probability of financial distress and the incurrence of costs associated with distress. Debt-to-equity ratio is used as a measure of expected costs of financial d i s t r e s ~ . ' ~ Underinvestment and dependence on internal funds: Froot, Scharfstein, and Stein (1993) using the underinvestment argument of Myers (1977), conclude that hedging is beneficial to firms with high external financing costs. They argue that underinvestment could occur when firms do not have sufficient internal funds to undertake all their investments, given costly external financing. Assuming decreasing marginal returns to investment, they show that hedging adds value to the extent that i t helps ensure that a corporation has sufficient internal funds to take advantage of attractive investment opportunities. An implication of the Froot, Scharfstein, and Stein hypothesis is that firms whose investment depends more on internally generated cash flows (as against externally raised financing) are more likely to find hedging useful. To test this proposition, a regression model that estimates investment by using Tobin's Q, sales, and cash flows, advanced in Fazzari, Hubbard, and Petersen (1988) and Hoshi, Kashyap, and Scharfstein (1991), is used.I6 The coefficient on cash flows is used as the measure of dependence on internal funds. Growth options: Nance, Smith, and Smithson (1993) argue that firms that have more growth options are likely to find hedging useful, as the underinvestment problem is likely to be severe for such firms (Myers [1977]). Accordingly, such firms are more likely to hedge to ensure availability of sufficient funds in all states. To measure the availability of growth options, market-to-book ratio and research and development expenditures are used. Taxes: Advantages to hedging can arise to the extent that progressivity in the tax code discourages some tax payers from taking on risky investments (Scholes and Wolfson [ 19921). This progressivity, typically arising out of net operating

13. Santomero (1998) provides a summary of these explanations. 14. Derivatives such as swaps can be used to hedge or to speculate on the movement of interest rates. In contrast to published instances of speculation in the general press, Dolde (1993). based on a survey of Fortune 500 firms, concludes that the primary motive for corporate use of derivatives such as swaps is hedging. 15. As an alternative to debt to equity, Altman's 2 ratio is also used (Altman [1968]). Results for this specification are quite similar to results using debt to equity and are not reported. 16. While a large body of economic literature exists that supports the investment-cash flow sensitivity measure, Kaplan and Zingales (1997) criticize investment cash flow sensitivities as poor measures of financing constraints. Their conclusion relies to a large extent on the result in their sample that shows the lowest financially constrained group exhibiting the highest sensitivity to cash flows, while the rest of the groups exhibit a positive and increasing relationship.

182

JOURNAL OF ACCOUNTING, AUDITING & FINANCE

losses and tax credits, provides an incentive for firms with NOLs or other tax credits to hedge. The tax variable is measured as net operating losses scaled by market value of equity. Ownership: Managerial risk aversion and management compensation policies also influence hedging decisions (Smith and Stulz [ 19851). For a firm with a high management ownership and an owner who is not well diversified, minimizing variance of firm cash flows is desirable. This implies that firms with a high managerial ownership are more likely to hedge. Ownership is measured by the log of market value of insider holdings. Size: Francis and Stephan (1993) and Nance, Smith, and Smithson (1993) establish that larger firms use hedging instruments more than smaller firms. It may be that size proxies for costs to hedge; that is, larger companies negotiate better prices. On the other hand, larger companies tend to be more diversified and, as such, the benefits of hedging are less obvious. This is treated as an empirical issue. Size is measured as the log of market value of the firm.

2.2 Incentives to Use Variable Interest Rate Swaps In contrast to theories that explain incentives to borrow short term and swap for long-term fixed-rate debt, theories that explain borrowing long term and swapping for a variable rate obligation tend to be brief. Wall (1989) and Titman (1992) conjecture that firms that engage in variable interest rate swaps are high-creditquality firms and that they share in the gains of firms that use fixed interest rate swaps. This hypothesis is supported by evidence in Wall and Pringle (1989), who show that firms that report variable interest rate swaps have better credit ratings than firms that use fixed interest rate swaps. These explanations and evidence suggest that firms that report variable interest rate swaps are expected to be firms of higher credit quality and have lower expected financial distress costs than firms that use fixed interest rate swaps. An alternative explanation for why firms use variable rate swaps could be that the firm’s duration of liabilities is greater than the duration of its assets. In this setting, the firm loses when interest rates increase. To hedge against this loss, the firm can modify its debt maturity structure by using a variable interest rate swap. Testing this explanation requires measures of both asset and debt maturities. Measurement of debt maturity structure was discussed earlier. For asset maturities, a measure based on Stohs and Maur (1996) is used. Because both these measures are merely proxies for the underlying maturities, they are not directly comparable. Accordingly, the empirical tests employ a dummy variable specification to capture the mismatch in asset and liability maturities.

INTEREST RATE SWAPS

183

3. Data 3.1 Sources and Description Data are collected on the non-financial companies in the S&P 500. Use of S&P 500 companies both yields a reasonable sample size of companies and encompasses a variety of industries. The focus is on nonfinancial firms because financial firms use swaps both for trading and for hedging purposes and the distinction is neither consistently nor clearly disclosed. This restricts the overall sample size to 443 firms.” Excluding foreign firms and firms that have merged since 1992 results in a sample size of 410 for 1993.IX While many companies disclosed the use of derivatives before 1990, this paper uses the disclosures required by SFAS No. 105 and SFAS No. 107 on notional and fair values of financial instruments. Companies use four major types of derivatives: forward contracts, interest rate and currency swaps, options, and futures. Most often used are interest rate swaps and forward contracts.” Data are collected on all types of derivatives reported. Few firms in the sample (fewer than 10 in each category) use interest rate derivatives other than swaps. These instruments are “swaptions” (options on swaps) and caps and floors that limit the exposure to interest rate movements. Given the insignificant number of these firms and the limited disclosures on these instruments, the study focuses only on interest rate swaps. Most companies disclose the information regarding swaps as part of the footnote on debt, but some companies disclose the information in a separate footnote on financial instruments or as part of contingent liabilities. Ownership data are from Value Line Investment Survey Reports; stock returns are from CRSP; and other company-specific financial variables are from Compustat.

3.2 Classification Issues2” The sample is partitioned into the following categories: 1. Companies that do not report interest rate swaps. This includes firms that

use non-interest-rate derivatives, such as foreign currency forward contracts, but excludes firms that use any type of interest rate derivative. 17. Two firms identified as undergoing bankruptcy proceedings are excluded because the arguments for financial distress costs are not applicable to such firms. 18. The 1992 sample size is smaller (338 firms) because SFAS No. 107 became effective only for fiscal years ending after December 15, 1992. To verify whether this introduces a bias in the sample, analyses for 1993 are repeated using only firms with December 31 as the fiscal year-end. This produces results similar to the ones reported. 19. Disclosures on swaps have some attractive features in contrast to other financial instruments, such as foreign currency forward contracts. Firms rarely disclose whether they are long or short in a specific currency, and because disclosure is required only as of the end of the fiscal year, the disclosed data may not be representative of the use of forward contracts during the year. Several firms do disclose the specific type of swap, and because swaps are used for longer periods than forward contracts, disclosures at the end of the year are representative of the usage during the year. 20. Companies do not always specify reasons for entering into swaps. When stated, the reason most often given is the management of interest rate exposure.

JOURNAL OF ACCOUNTING, AUDITING & FINANCE

184

2. Companies that report interest rate swaps. This group in turn is classified by : 2a. Companies that report fixed rate swaps 2b. Companies that report variable or floating rate swaps, and 2c. Companies that report both fixed and variable rate swaps but do not provide notional values by type of swap or merely indicate the existence of swaps without mentioning the type of swap. These are some of the largest firms in the sample (e.g., General Motors and General Electric). Groups 2a and 2b include firms that report a specific type of swap only (in 1993, 65 firms report only fixed-rate swaps and 42 report only variable rate swaps) and firms that report both types of swaps but the notional value of one type exceeds the notional value of the other type (in 1993, for 19 [I21 firms the notional value of fixed [variable] rate swap exceeds the notional value of variable [fixed] type swaps). As none of the theories discussed offers explanations for the simultaneous use of both types of swaps, firms in group 2c are considered separately in the reported empirical results.

4. Empirical Results 4.1 Univariate Tests

A description of variables used in the study is included in Table 1. Descriptive statistics and industry distributions are reported only for the year 1993. Results for 1992 are similar unless otherwise indicated. The mean notional value of swaps as a proportion of book value of total debt is 24 percent for all firms that report swaps (not reported). Table 2 presents the sample distribution by industry. Of the 410 firms in the sample, 196 report the use of interest rate swaps, and 214 report no swaps. In 1993, 84 (76) firms report fixed rate swaps, and 54 (42) firms report variable rate swaps. Utilities (a rate-regulated industry) and rails, trucking, and toy industries have the least swap users. Telephone and chemical and drug industry firms have the most swap users. In the specific swap category, utility firms report the highest proportion of fixed interest rate swaps (six out of nine users), whereas only one firm in the mining, gold, and oil industries use fixed interest rate swaps. The proportion of firms that use variable rate swaps is the least among the groups with none of the firms in the rails, trucking and toys group reporting any. Table 3 provides descriptive statistics. Panel A provides comparisons of the variables discussed in the hypotheses development across the three groups: firms that report fixed interest rate swaps (represented as FS firms), firms that report variable interest rate swaps (represented as VS firms), and firms that report no interest rate swaps (represented as No Swap firms). In addition, descriptive statistics are provided for firms that report both types of swaps or do not provide specific information about swaps (Both Swaps). Both FS and VS firms have significantly higher debt-to-equity ratio, short-term debt, and long-term debt compared to No

INTEREST RATE SWAPS

185

TABLE 2 Sample Distribution by Industry Sample consists of 410 S&P 500 firms in 1993. Sample is classified into firms that report interest rate swaps (indicated as Swap) and firms that do not report any type of interest rate swaps (indicated as No swap). Firms that report swaps are classified into firms that report predominantly fixed-rate swaps (indicated as FS), firms that report predominantly variable rate swaps (indicated as VS), and firms that report both fixed and variable rate swaps (indicated as Both). The proportion of firms that report interest rate swaps to total number of firms in the sample is indicated as SwapiTotal.

Industry Mining, gold, and oil Food and beverages Paper and printing Chemicals and drugs Petroleum refining Rubber, steel, and glass Computers and equipment Autos and instruments Rails, trucking, and toys Telephone Utilities Department stores, software, and miscellaneous Total number of firms

SIC Codes

FS

VS

Both

Swap

10-16 20-25 26-27 28 29 30-34 35-36 37-38 39-45 48 49

1 13 6 9 I 6 10 12 2 3 6

4 4 7 9 5 6 3 0 2 3

4 2 2 10 5 5 15 9 2 4 0

9 19 15 28 7 16 31 24 4 9 9

SO-87

15 84

10 54

0 58

25 196

1

No Swap 10

15 14 17 6 18

Swapl Total (%) 47 56 52 62 54

4 29

47 54 59 27 69 24

47 214

35 48

26 17 11

Swap firms.” FS and VS firms are also bigger and have better credit ratings (note that lower values of BRAT represent higher credit ratings). No significant differences exist between VS and FS groups, except that FS firms have somewhat (10 percent significance level) higher short-term debt. However, in 1992 (results not reported), FS firms have significantly higher debt to equity and short-term debt than VS firms. Neither the debt maturity measure (DMAT) nor the asset maturity measure (AMAT) are significantly different across the two groups. Also, the groups do not differ significantly in their interest rate sensitivity measure (IS). Firms that are classified as both swaps have higher debt-to-equity ratios, short-term debt, and asset maturity but significantly lower debt maturity than firms that use no swaps. They are also larger and have better credit ratings. Last, results for the hedging variables indicate that firms using swaps have significantly higher dependence on internal funds (INTF) than firms that do not use swaps. No significant differences exist between FS and VS firms in terms of dependence on internal funds, net

21. Significant differences are discussed. A cut off of 5 percent is used for significance level unless otherwise indicated.

L

a

CQ

BRAT DMAT AMAT LMV MB BETA 1s INTF NOL DOWN

FEPR

DE STM LTM

Variable

0.33 0.06 0.26 0.0 I 12.50 0.73 5.19 8.53 1.65 1.15 0.10 0.29 0.06 3.69

Mean

0.3 1 0.04 0.23 0.01 10.00 0.77 4.37 8.53 1.33 1.10 0.01 0.15 0.00 4.09

Median

FS (N=84)

0.30 0.05 0.25 0.02 12.56 0.73 5.50 8.52 I .74 1.24 0.14 0.47 0.01 3.83

Mean

0.26 0.03 0.19 0.01 9.00 0.74 4.95 8.49 1.47 I .22 0.06 0.20 0.00 4.0 I

Median

VS (N=54)

0.22 0.03 0.19 -0.0 I 16.10 0.72 5.54 7.90 1.89 1.13 0.09 0.05 0.05 3.84

0. I8 0.02 0.16 0.01 11.00 0.8 I 3.84 8.06 I .46 1.18 0.03 0.09 0.00 4.05

Median

0.28 0.1 I 0.23 0.02 10.45 0.63 6.03 9.30 1.74 I .ox 0.15 0.29 0.01 4.67

Mean

Both Swaps (N=58)

0.00 0.00 0.01 0.26 0.00 0.42 0.26 0.00 0.03 0.38 0.36 0.04 0.36 0.28

FS and No Swap

0.00 0.47 0.47 0.00 0.16 0.05 0.12 0.0 I 0.00 0.49

0.0 1 0.05 0.03 0.19

VS and No Swap

0.21 0.07 0.31 0.21 0.48 0.43 0.33 0.47 0.28 0.1 1 0.22 0.33 0.07 0.37

vs

FS and

p-Values for I Statistics-Comparing

that report a specific type. of swaps and firms that report no swaps

No Swap (N=214) Mean

Panel A: Comparison of firm characteristics across various groups-firms

0.03 0.00 0.30 0.41 0.00 0.01 0.05 0.00 0.17 0.24 0.08 0.02 0.05 0.01

Both and No Swap

Mean Values

Characteristics of sample firms that report a specific type of swap in 1993 and firms that do not report any type of interest rate swap in 1993 are presented.

Univariate Statistics and Comparisons for 1993

TABLE 3

4

m

-

62 (74%) 8 (9.5%) 14 (16.5%) 84

FS

8.45 8.50

9.15 9.00 8.49 8.00

9.20 9.00

I990

8.59 8.00

9.35 9.00

1991

54

8.73 8.00

9.43 9.00

I992

39 (72.2%) 4 (7.4%) I I (20.4%)

vs

8.76 8.00

9.43 9.00

I993

8.63 8.00

9.59 9.00

8.42 8.00

9.64 9.00

1995

47 (81%) 3 (5.2%) 8 (13.8%) 58

125 (58%) 21 (10%) 68 (32%) 214

1994

Both Swaps

No Swap

Note: A BRAT value of 8.00 represents an S&P credit rating of A and a BRAT value of 9.00 represents an S&P credit rating of A-, both of which are considered investment grade.

Mean Median

vs

FS Mean Median

1989

Punel C: S&P credit ratings of firms that repon swaps in 1993 over time Mean and median values OF BRAT are reported over 1989-1995.

Total

Investment grade AAA to BBBInvestment grade BB+ and below Nonrated

S&P Credit Rating

Punel B: S&P credit ratings of 1993 sample firms Number of firms in each category are reported and the number of firms in a specific credit rating category as a percentage of the total number of firms are reported in parentheses.

JOURNAL OF ACCOUNTING, AUDITING & FINANCE

188

operating loss carryforwards (NOL) and measure of managerial stock ownership (DOWN) variables.22 Panels B and C of Table 3 provide additional details on the credit ratings of sample firms. About three quarters of the firms that use swaps are of investment grade credit quality, whereas about 10 percent of these firms are rated below investment quality. Some of the sample firms are not rated, in particular the No Swap firms. Both Barclay and Smith (1995) and Stohs and Maur (1996) cite evidence that firms that are not rated tend to be firms with little debt or with debt that is private. To verify whether such firms are of low or high credit quality the procedure employed by Barth, Beaver, and Landsman (1996) is used.*’ Results (not reported) indicate that a majority of these sample firms that are not rated would be of investment grade quality. To consider the prediction of Titman (1992) that FS firms’ credit ratings are expected to improve over time, panel C reports the ratings over a seven-year horizon. Results do not indicate any systematic changes over this horizon for either FS or VS firms. 4.2 Logistic Model Results

Correlations among key explanatory variables are reported in Table 4 for the entire sample. Several variables are significantly correlated and the impact of such collinearity on the empirical results is discussed later. 4.2.1 RESULTS FOR THEORIES THAT PREDlCT THE USE OF FIXED INTEREST RATE SWAPS A two-stage estimation procedure is used to consider simultaneity in the choice of debt maturity structure and the decision to use a swap (Maddala [ 1983, 19911). For debt maturity structure, the following model is used based on modifications to Barclay and Smith (1995) and Stohs and Maur (1996):

DMAT, = a,

+ a2 MB, + u3 FABE, + u4 INDD, + us LMV,

+ us DE, + a, INTF, + a, NOL, + ay DOWN, + a,,

IS,,

(1)

where DMAT is debt maturing after two years as a proportion of total debt, MB is market-to-book ratio, FABE is abnormal earnings reported in the subsequent year, INDD is a dummy variable for regulated industries (primarily utilities industry), LMV is log of market value of firm, and DE is debt-to-equity ratio. Barclay and Smith also include term structure of interest rates as an additional variable; it 22. Results (not reported) for other hedging variables used in hedging literature such as Altman’s Z , research and development expenditures, quick ratio, and dividend yield are broadly consistent with the hedging theories in terms of differences between firms that report swaps and firms that do not report swaps. 23. Barth et al. estimate bond ratings for firms that are not rated by using a model that considers a firm’s total assets, return on assets, debt to total assets, and a dummy for whether the firm pays dividends.

INTEREST RATE SWAPS

189

TABLE 4 Correlations among Key Explanatory Variables Correlations are reported for the entire sample for 1993. Pearson (Spearman) correlations are reported below (above) the diagonal with p-values in parentheses. DE I .oo (0.00)

DE DMAT BRAT DRAT FEPR AMAT IS

DMAT

BRAT

DRAT

FEPR

AMAT

IS

0.27 (0.00)

0.14 (0.00) 0.14 (0.00)

-0.18 (0.00) -0.07 (0.14) 0.78 (0.00)

0.30 (0.00) 0.1 I (0.04) 0.01 (0.93) -0.18 (0.00)

0.22 (0.00) 0.10 (0.04) -0.28 (0.00)

I .oo

0.08 (0.14)

0.11 (0.03) 0.05 (0.36) -0.07 (0.20) -0.06 (0.18) 0.17 (0.00) 0.06 (0.30) 1.oo

0.30

I .oo

(0.00)

(0.00) 0.01 (0.48) -0.1 I (0.03) 0.03 (0.52) 0.13 (0.01) 0.02 (0.73)

0.01 (0.47) -0.14 (0.00) 0.07 (0.20 0.27 (0.00) 0.15 (0.00)

I .oo (0.00) 0.94 (0.00) -0.03 (0.31) -0.16 (0.00) -0.05 (0.29)

I .oo (0.00) -0.01 (0.90) -0.17 (0.00) - 0.06 (0.23)

(0.00) 0.01 (0.80) 0.08 (0.10)

-0.32 (0.00) I .oo (0.00) 0.10 (0.06)

(0.W

See Table 1 for a description of each variable

is not included here because this is a cross-sectional regression. DE is included based on Stohs and Maur. Barclay and Smith do not include DE. The predicted sign on LMV, DE, and INDD is positive, and the predicted sign on FABE and MB is negative. INTF, NOL, and DOWN are included based on hedging theories. Note that other variables considered by hedging theories such as debt-to-equity ratio, market-to-book ratio, and size are already included based on debt maturity structure theories. Last, a measure of interest rate sensitivity IS is included. Predicted values of DMAT are introduced as an additional explanatory variable in the logistic regression that tests theories of incentives for firms to use fixed interest rate swaps. The following logistic model is used to test theories of incentives for firms to use fixed interest rate swaps:

+ a, PRED, + u3 DE, + u4 CPR, * PRED, + u5 BRAT, + us DRAT, + u, FEPR, + u8 DER,,

2, = a ,

(2)

where Z takes the value of 1 if the firm reports a fixed interest rate swap and 0 otherwise, PRED is the predicted value of DMAT from model 1, DE is debt-toequity ratio, CPR is a dummy variable that takes the value of 1 if the difference between actual and predicted DMAT is greater than the median value for the sample and is 0 for values below median, BRAT is a numeric variable for S&P bond rating (lower the value of BRAT higher the credit rating), DRAT is a dummy variable that is set equal to 1 for firms with no S&P bond rating and is 0 otherwise,

JOURNAL OF ACCOUNTING, AUDITING & FINANCE

190

TABLE 5 Logistic Model Results that Estimate the Likelihood of Reporting a Fixed Interest Rate Swap Sample consists of firms that report only fixed interest rate swaps and firms that report no interest rate swaps Coefficients along with marginal probabilities that measure the change in the probability of using a fixed rate swap for a change in the independent vanable are reported Panel A The logistic model is Z, = a, + a, PRED, + a, DE, + u, CPR, * PRED, + a, BRAT, + u6 DRAT, Z = 1 if firm I reports a fixed interest rate swap and equals 0 otherwise

+ a, FEPR, + a,

1992 Variable

Predicted Sign

Intercept PRED DE CPR * PRED BRAT DRAT FEPR DER Pseudo R’ = 0.18

?

-

+ -

+ ?

+ ?

Coefficient 6.58

- 11.82 6.59 -0.54 -0.10 1.01 0.24 0.80

1993

Marg. Prob.

t Ratio

I .24 -2.22 1.24 -0.10 -0.02 0.19 0.04 0.15

2.38 -2.85** 4.36** - 1.75* - I .43 0.72 0.2s 2.29**

Coefficient

1.55 -5.16 7.29 - I .29 -0.06 0.36 -2.73 I .25 Pseudo R’

Predicted Dependent Variable Actual dependent variable 0 I Total

0 125 19 I44

1

44 22 66

DER,

Marg. Prob. 0.30 -0.99 I .39 -0.24 -0.01 0.07 -0.52 0.24 = 0.21

t Ratio

0.42

- I .70* 2.85** -2.80** - I .oo 0.26 - 1.38 3.67**

Predicted Dependent Variable

Total 169

0 I56

41 -

15 -

210

171

1 41 32 73

Total 197 41 244

FEPR is cumulative abnormal earnings for the three-year period subsequent to the current year, and DER is a dummy variable that is set equal to 1 if the firm uses derivatives other than interest rate swaps and 0 otherwise. Three-year cumulative abnormal earnings are used (FEPR) to capture the effect of a change in earnings over a period of time. DER is included to control for firm characteristics that relate to the use of derivatives in general. The predicted signs on DE, BRAT, and FEPR are positive, and the predicted signs on PRED and CPR are negative. The predicted value of DMAT (PRED) is used as an additional variable in eq. (2), and the predicted probability from the logistic model that estimates the probability of using a fixed interest rate swap is used as an additional variable in eq. ( I ) as in a two-stage estimation procedure (Maddala [ 19831). Results are reported in Table 5 for 1992 and 1993. Panel B reports results for model 1. The debt maturity structure model exhibits low explanatory power with a subset of the var-

191

INTEREST RATE SWAPS

TABLE 5 (continued) Panel B . Vanable PRED in panel A IS the predicted value of DMAT from the following regression DMAT, = a , a2 MB, + a, FABE, + a4 INDD, + a, LMV, + a, DE, + a, INTF, + u8 NOL, + a, DOWN, + a,,, IS,

+

1992

Variable

Predicted Sign

Intercept MB FABE INDD LMV

DE INTF NOL DOWN IS

? -

+ + + ? ? ? ?

Coefficient 0.73 -0.01 -0.16 0. I4 -0.01 0.17 0.01 -0.01 0.0 I

0.04

Adjusted R' = 0.07

1993 p-Value for t Statistics 0.00 0.34 0.08 0.01 0.23 0.05 0.48 0.48 0. I9 0.29

Coefficient

p-Value for t Statistics

0.60 -0.01 -0.03 0.1 1 -0.01 0.44

0.02 -0.02 0.01 -0.02 Adjusted R' = 0. I3

0.00 0.40 0.38 0.02 0.37 0.00 0.05 0.37 0.18 0.37

*Significant at the 10% level. **Significant at the 5% level. See Table I for a description of each variable

iables being significant. The dummy variable for regulated industries (INDD) and debt-to-equity ratio (DE) are consistently significant in both years. The signaling variable FABE is weakly significant in 1992 but insignificant in 1993. Although the significance of these variables is consistent with Barclay and Smith (1995) and Stohs and Maur (1996), the explanatory power of the model is lower than the explanatory power shown in those two studies. One explanation for this is the smaller sample size (only firms that report fixed-rate swaps and firms that report no swaps in S&P 500) in the present context as against the entire population of Compustat firms and a longer time series used in the other studies. Other hedging variables included in the model (with the exception of INTF in 1993) are not significant. More important, interest rate sensitivity is not significant, in contrast to the theoretical predictions. Results for model 2 support the predictions on the relevance of financial distress costs, debt maturity structure, and the presence of short-term debt in the decision to use a swap. The results are reported in panel A. Firms that report fixed interest rate swaps and firms that report no swaps are used in the sample. Marginal probabilities that denote the change in probability of using a fixed interest rate swap for a per unit change in the explanatory variables are also reported. Consistent with the predictions of Titman (1992) and Wall (1989), expected financial distress costs, as measured by debt to equity, are significant in both years. A 1 percent increase in the debt-to-equity ratio results in an increase of 1.39 (1.24)

192

JOURNAL OF ACCOUNTING, AUDITING & FINANCE

percent increase in the 1993 (1992) sample in the probability of using a fixed interest rate swap. The predicted value of debt maturity is significant (at the 10 percent level in 1993) and negative, supporting the conjecture that firms consider the debt maturity decision in deciding to use a fixed interest rate swap.24The variable CPR is used to measure whether the firm desires long-term fixed-rate debt (as represented by the predicted value of DMAT) but uses short-term debt instead. It takes the value of 1 (0) for firms that have greater (lesser) long-term debt than predicted, and it interacts with PRED. Results show that it is significant in 1993 and weakly significant (at the 10 percent level) in 1992.*’ The logistic model correctly predicts about 77 (70) percent of the observations in 1993 (1992), but the incorrect classifications are greater in identifying firms that use fixed interest rate swaps. The prediction that firms that expect their prospects and credit quality to improve have incentives to borrow short term and use fixed interest rate swaps is not supported. FEPR, a measure of cumulative abnormal earnings, is not significant in both periods. Measuring FEPR as one year ahead or two years of cumulative earnings produces similar results. These results also are consistent with the time series pattern of credit ratings shown in Table 3, which show no significant improvement. In addition, variables that denote credit ratings BRAT and DRAT are mostly not significant. DER, a dummy variable for other derivatives usage, is positive and significant in both years, indicating that firms that report swaps also report other types of derivatives as compared to firms that do not report swaps. Results that include the predicted probability in the debt maturity structure model indicate that the predicted probability of using a fixed interest rate swap is not significant in explaining the debt maturity structure decision. These results are not reported.

4.2.2 LOGISTIC RESULTS THAT COMPARE FIRMS THAT USE FIXED INTEREST RATE SWAPS AND FIRMS THAT USE VARIABLE INTEREST RATE SWAPS

To test theories that explain the use of variable interest rate swaps, firms that report predominantly fixed interest rate swaps and firms that report predominantly variable interest rate swaps are used in the sample. The logistic model consists of the credit rating variables, predicted value of DMAT (results not reported), debt to equity, and variables measuring asset maturity. Results in Table 6 do not support the hypothesis that credit ratings of firms 24. PRED and DE are significantly correlated (Pearson correlation of 0.60 or more). To consider the impact of collinearity, when PRED alone is considered in the model, its significance does not change in the specifications in Table 5 and in other empirical results. 25. Excluding the CPR variable from the model results in a change in the magnitude of PRED but not its significance. Without CPR in the model, coefficient estimates of PRED are - 11.79 in 1992 and -5.57 in 1993.

INTEREST RATE SWAPS

193

TABLE 6 Logistic Model that Estimates the Likelihood of Reporting a Fixed-Rate Swap-Sample of Firms that Report a Fixed or a Variable Interest Rate Swap Sample consists of firms that report predominantly fixed interest rate swaps and firms that report predominantly vanable interest rate swaps Coefficients along with marginal probabilities that measure the change in the probability of using a fixed-rate swap for a change i n the independent vanable are reported The logistic model is Z , = u , C u 2 PRED,+u, DE,+u, VPR,fa, AMAT,+u, BRAT,+u, DRAT,+u, FEPR, Z= 1 if firm reports a fixed interest rate swap and equals 0 otherwise

1992 Predicted Sign

Variable

? Intercept PRED DE + V PR ? AMAT BRAT DRAT ? FEPR Pseudo R' = 0.12

+ +

1993

Coefficient

Marg. Prob.

t Ratio

0.35 -2.00 3.73 I .06 -0.08 0.09 - I .24 -0 26

0.07 -0.41 0.76 0.22 -0.02 0.02 -0.25 -0.05

0.19 -0.67 2.14** I .28 - 1.22 0.77 -0.52 -0.15

Coefficient

0.33

0.07 -0.39 0.96 0.22 0.24 I .04 0.01 0.01 0.12 0.03 -0.60 -2.65 -1.51 -0.34 Pseudo R' = 0.06 - I .73

Predicted Dependent Variable Actual dependent variable 0

Marg. Prob.

t Ratio

0. I6 -0.49 0.63 1.38 0.05 I .20 - I .42 -0.73

Predicted Dependent Variable

I

0 II 25 -

13 52 -

Total 24 77 -

0 5 47 -

61 -

-

Total

36

65

101

52

63

I15

1

I 2

Total 7 108

*Significant at the 10% level. **Significant at the 5% level. See Table I for a description of each variable

that report variable interest rate swaps are likely to be higher than ratings for firms that report fixed interest rate swaps. The credit rating variable BRAT is positive but not significant, and the dummy variable DRAT is negative and insignificant. Because debt to equity and credit ratings are correlated, the logistic model is also estimated with only the credit rating variables BRAT and DRAT (results not reported). Both these variables are significant in I992 but not in I993 in this modified model, indicating limited support for the prediction that variable rate swap firms have higher credit ratings. This results, however, from the inclusion of firms that report both types of swaps and whose notional values of one type of swap exceed

194

JOURNAL OF ACCOUNTING, AUDITING & FINANCE

the other type. When only firms that report a specific type of swap are included, credit rating variables are not significant. Predictions about the relevance of the mismatch in asset and debt maturities are not supported. Two variables AMAT and VPR are used. The AMAT of firms that report variable rate swaps are not significantly different from asset maturities of firms that report fixed-rate swaps. VPR, a dummy variable that is designed to capture mismatches in asset and liability maturities, has the incorrect sign and is not significant. Because firms with a VPR value of 1 are firms that have low asset maturities and high debt maturities, the mismatch hypothesis predicts that such firms are more likely to report variable rate swaps. Results in Table 6 do not support this conjecture. These results are also consistent with evidence in Table 3, which shows no significant differences between fixed and variable rate swap firms in terms of AMAT and DMAT at the univariate level. Although these results provide little support to the asset-liability mismatch, it should be noted that the VPR variable does not measure the actual “gap” between asset and liability maturities, but instead reflects the relative length of these maturities. Accordingly, these results should be interpreted with caution. None of the other variables in Table 6, with the exception of DE in 1992, is significant. The analysis is extended by repeating the logistic procedure with a sample that includes only firms that report variable interest rate swaps and firms that report no swaps. These results also show no significant differences in credit ratings between firms that report variable interest rate swaps and firms that report no swaps. In summary, these results, along with results in Table 6, provide only limited support for the prediction that high credit quality is a significant feature of firms that report variable interest rate swaps. 4.2.3 LOGISTIC MODEL RESULTS THAT COMPARE FIRMS CLASSIFIED AS REPORTING “BOTH SWAPS” AND FIRMS THAT REPORT NO SWAPS

As discussed previously, a significant number of firms that report swaps use both types of swaps but provide inadequate information as to what type of swap is used. A distinguishing feature of these firms is that they are the largest among the overall sample; results in Table 3 show the mean value of LMV is significantly larger for these firms than the mean values of firms that use a specific type of swap and firms that use no swaps. Several of these firms have finance subsidiaries. These subsidiaries appear to account for the large number of swaps that these firms use. Because these firms provide no information on the type of swap and no specific theoretical predictions pertain to these firms, variables that are used to test hypotheses for both fixed and variable type swaps are considered in the empirical analysis. Results are reported in Table 7. The results show that firms that report “both swaps” have some similarities to firms that report fixed interest rate swaps. These firms have a higher debt-to-equity ratio than firms that report no swaps. As do firms that use fixed-rate swaps, they consider the debt maturity structure

195

INTEREST RATE SWAPS

TABLE 7 Logistic Model Results that Estimate the Likelihood of Reporting “Both Swaps” ~

~

~

~

~

~

~~~

Sample consists of firms that are classified as “both swaps Coefficients along with marginal probabilities that measure the change in the probability of reporting “both swaps” for a change in the independent vanable are reported The logistic model is Z, = a , a2 PRED, + a , DE, + a4 CPR, * PRED, + a, BRAT, t a, DRAT, t a7 FEPR, + a, AMAT, t ug VPR, + a,,, DER, Z = 1 if firm 1 is classified under “both swaps” and equals 0 otherwise ”

+

1992 Predicted Sign

Variable

Intercept PRED DE CPR * PRED BRAT DRAT FEPR AMAT v PR DER Pseudo RL = 0.34

? ? ? 1

? ? ? ? ? ?

Coefficient

Marg. Prob.

10.56 -20.00 5.58 -0.87 -0.07 0.80 0.22 0.0 I -0.14 2.14

1.16 -2.20 0.61 -0.10 -0.01 0.09 0.02 0.00 -0.02 0.24

1993 t Ratio

Coefficient

3.07** -3.5 I ** 3.12** - I .07 -0.88 0.4 I

8.75 -16.19 13.16 -0.11 -0.21 3.36 -2.17 -0.19 -0.96 - I .90 Pseudo R’ =

0.1 1

0.2s 0.20 3.89**

Predicted Dependent Variable Actual dependent variable 0

M arg . Prob. 0.94

- 1.73 1.41 -0.01 -0.02 0.36 -0.23 -0.02 0.10 0.20 0.36

t Ratio

2.31** -2.53** 3.99** -0.17 -2.39** 1.78* -0.90 -1.13 I .38 -4.13**

Predicted Dependent Variable 1

Total

32 -

0 I60 II -

21 31 -

42 -

I90

171

52

223

1

Total

22 22 -

1.58

1

0 136 10 -

Total

I46

44

181

*Significant at the 10% level. **Significant at the 5% level. See Table I for a description of each variable

(regression results for predicted values of DMAT are not reported) in using a swap. In 1993, these firms have better credit ratings than firms that do not report swaps. Other variables that are specific to fixed interest rate swaps (CPR) or specific to variable interest rate swaps (VPR) are not significant. Last, the explanatory power and the correct number of predictions are the highest for this sample. The results suggest that several of these swaps are likely to be fixed interest rate swaps, given the similarity to results in Table 5 . However, because variables that are specific to fixed interest rate swap firms (CPR and FEPR) are not significant, no definitive

196

JOURNAL OF ACCOUNTING, AUDITING & FINANCE

inferences could be made as to whether these are primarily fixed or primarily variable interest rate swaps.*‘

4.2.4 RESULTS FOR LOGISTIC MODEL THAT CONSIDERS FIRMS THAT ARE NOT INTEREST RATE SENSITIVE

Earlier analysis considers sensitivity of firm cash flows to interest rates in determining the debt maturity structure. To address the role of interest rate sensitivity more directly in the use of swaps, further analysis is conducted by examining a subsample of firms that have insignificant interest rate sensitivity (i.e., earnings before interest and taxes are not significantly associated with interest rates). This sample is classified into two groups: firms that report any type of interest rate swap and firms that do not report any swaps. Four explanatory variables that are common to all types of swaps are used in the analysis: debt-to-equity ratio (DE), predicted value of debt maturity (PRED), and the two credit rating variables (BRAT and DRAT). Results are shown in Table 8. Results indicate that interest rate sensitivity is not a critical factor in using swaps. This is supported by the evidence that in both years, nearly one half of the sample firms report interest rate swaps. On the other hand, a high debt-to-equity ratio, short debt to maturity, and high credit ratings distinguish firms that report swaps from firms that do not report swaps. Variables DE and PRED are significant in both years, indicating firms with high debt to equity and firms with lower debt maturity report swaps. Moreover, firms with higher credit ratings tend to report swaps (BRAT is negative and significant). These results suggest that reducing financial distress costs and modifying debt maturity structure are considered more important than sensitivity of firm cash flows to interest rates in using interest rate swaps.

4.3 Sensitivity Checks Measurement of debt maturily structure: Results are reported with debt maturity structure measured as debt maturing after two years as a proportion of total debt. Results are robust to the number of years (three to five) used as a cut off. Moreover, similar results are obtained when only short-term debt, that is, debt maturing within a year, as a proportion of total debt is considered as the measure of debt maturity structure. Also, as noted earlier, debt-to-equity ratio is included in the debt maturity structure model based on Stohs and Maur (1996), even though the Barclay and Smith (1995) model does not include debt to equity. To consider the robustness of the significance of PRED (predicted value of debt maturity) in 26. One possible explanation for the reporting of “both swaps” is earnings manipulation through the simultaneous use of different types of swaps. Given the paucity of information about the impact on accounting earnings from the use of specific swaps, testing this explanation is beyond the scope of this study.

197

INTEREST RATE SWAPS

TABLE 8 Logistic Model that Estimates the Likelihood of Reporting an Interest Rate Swap-Firms that Do Not Have Significant Interest Rate Exposure Sample consists of firms that report any type of interest rate swap (fixed, variable, and both) and firms that do not report any interest rate swaps. Firms that are interest-rate-sensitive are excluded. Coefficients along with marginal probabilities that measure the change in the probability of using swap for a change in the independent variable are reported. The logistic model is Z, = a , + u2 PRED, + a3 DE, + a4 BRAT, + a, DRAT, + u, DER, Z = 1 if firm i reports an interest rate swap and equals 0 otherwise. 1992 Variable Intercept PRED DE BRAT DRAT DER Pseudo R’

=

Predicted Sign

Coefficient

? ? ? ? ? ?

4.51 -6.73 3.81 -0.13 0.99 0.94

1993

Marg. Prob. 1.14 - 1.68

0.95 -0.03 0.25 0.24

Coefficient

2.37** -2.48** 2.86** -2.17** 0.72 2.54**

5.03 1.26 -9.35 -2.34 8.24 2.06 -0.1 I -0.03 1.36 0.34 I .06 0.27 Pseudo R’ = 0.19

0.16 Predicted Dependent Variable

Actual dependent variable 0 I Total

0 46 37 83

1

17 59 76

Marg. Prob.

t Ratio

Total 63 96 161

t Ratio

2.50** -2.37** 4.14** -2.01** I .oo

3.05**

Predicted Dependent Variable 0 59 40 99

1 18

74 92

Total 77 I14 191

-

~

*Significant at the 10% level. **Significant at the 5% level. See Table I for a description of each variable

the decision to use a swap, the debt maturity structure model is estimated without debt to equity ratio. In this specification, too, PRED remains significant. Logistic regression diagnostics: Analyses are performed based on Pregibon (1981). Deletion of influential observations that have C’s or CBAR’s (see Pregibon [1981]) greater than 1 does not have any material effect on the results.

5. Summary This study examines the determinants of the use of interest rate swaps and empirically tests theories that predict the use of fixed and variable interest rate swaps. Consistent with predictions, results show that firms that use fixed interest rate swaps have higher expected financial distress costs and consider debt maturity

198

JOURNAL OF ACCOUNTING, AUDITING & FINANCE

structure in the decision to use swaps. Specifically, firms with shorter debt maturity structures are more likely to report fixed rate swaps. Although firms that report swaps have better credit ratings than firms that do not report swaps, predictions about differences in credit ratings between firms that report fixed-rate swaps and firms that report variable interest rate swaps are (in contrast to earlier studies) supported only weakly in a subset of firms. Predictions about improvement in the financial prospects of firms that report fixed rate swaps are not supported. Some of the largest firms in S&P 500 report both types of swaps and provide inadequate information about the type of swaps used. Results show that these swap users have significantly higher financial distress costs and consider their debt maturity structure in using a swap. Last, results for the sample of firms that are not sensitive to interest rates indicate that the two aforementioned variables, financial distress costs and debt maturity structure, are significant in distinguishing firms that report swaps from those that do not, in this subsample. Overall, these results indicate that distress costs and debt maturity structure are the critical factors in the decision to use a swap and not interest rate sensitivity of the operations. The role of credit ratings appears to be significant in obtaining a swap but not necessarily in distinguishing between swap users. The results of the study, while partly consistent with the existing empirical literature on hedging, suggest that incentives to use different types of derivatives are not all alike and that there is incremental value in studying specific types of instruments. The results of the study are also relevant to the growing literature on debt maturity structure. One of the attractive features of interest rate swaps is that they can be used to alter the debt maturity structure. Both the measurement of debt maturity structure and the test of theories that examine debt maturity structure need to take into account the change in the debt maturity as a result of swaps.*’ A shortcoming of the study is the limited period of two years and the lack of entry and exit dates of swaps for a large portion of the sample. This renders conclusions about changes in financial prospects of the firms that use swaps difficult to interpret. Improving disclosures and a longer time series would further the analysis in this regard. This is a promising avenue for future research.

REFERENCES

Ahmed, A. S., A. Beatty, and C. Takeda. 1997. “Evidence on Interest Rate Risk Management and Derivatives Usage by Commercial Banks.” Working Paper, University of Rochester. Altman, E. I . 1968. “Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy.” The Journal of Finance 23 (September): 589-609. Arak, M., A. Estrella, L. Goodman and A. Silver. 1988. “Interest Rate Swaps: An Alternative Explanation.” Financial Management 17 (Summer): 12-18. Barclay, M. J., and C. W. Smith. 1995. “The Maturity Structure of Corporate Debt.” The Journal .f Finance 50 (June): 609-631. Barth, M. E., W. H. Beaver, and W. R. Landsman. 1996. “Valuation Characteristics of Equity Book 27. Guedes and Thompson (1995) point out the importance of considering the impact of interest rate swaps in tests that examine the choice between fixed and adjustable rate debt.

INTEREST RATE SWAPS

199

Value and Net Income: Tests of the Abandonment Option Hypothesis.” Working Paper. Stanford University. Bicksler, J., and A. H. Chen. 1986. “An Economic Analysis of Interest Rate Swaps.” Journal of Finance 41 (July): 645-655. Cooper, 1. A,, and A. S. Mello. 1991. “The Default Risk of Swaps.” The Journal ofFinance 46 (June): 597-620. DeMarzo, P., and D. Duffie. 1992. “Corporate Incentives for Hedging and Hedge Accounting.” Working Paper, Northwestern University. Diamond, D. W. 1991. “Debt Maturity and Liquidity Risk.” The Quarterly Journal ofEconomics 106 (August): 709-737. Dolde, W. 1993. “The Trajectory of Corporate Risk Management.” Journal of Applied Corporate Finance 4 (Fall): 3 3 4 1 . Duffie, D. 1993. Futures Markets. Englewood Cliffs, N.J.: Prentice-Hall. Duffie, D., and M. Huang. 1996. “Swap Rates and Credit Quality.” The Journal of Finance 51 (July): 921-949. Fazzari, S.. R. G. Hubbard, and B. Petersen. 1988. “Financing Constraints and Corporate Investment.” Brookings Papers on Economic Activiry 1:141-195. Fenn, G. W., M. Post, and S. A. Sharpe. 1996. “Debt Maturity and the Use of Interest Rate Derivatives by Nonfinancial Firms.” Working Paper, Federal Reserve Board, Washington, D.C. Financial Accounting Standards Board (FASB). I98 1. Statement o j Financial Accounting Standards No. 47: Disclosure of Long-Term Obligations. Stamford, Conn.: FASB. -. 1990a. Statement ($Financial Accounting Standadv No. 105: Disclosure of Information about Financial Instruments with Or-Balance-Sheet Risk and Financial Instruments with Concentrations of Credit Risk. Stamford, Conn.: FASB. -. 1990b. Statement of Financial Accounting Standards No. 106: Employers’ Accounting f o r Postretirement Benefits Other thtrn Pensions. Stamford, Conn.: FASB. -. 1991. Stutement of Financiul Accounting Standards No. 107: Disclosures about Fair Value of Financial Instruments. Stamford, Conn.: FASB. -. 1994. Statement (~fFinancial Accounting Standards NO. I1 9: Disclosure About Derivative Financial Instruments and Fair Value o j Financial Instruments. Stamford, CT: FASB. Flannery, M. J. 1986. “Asymmetric Information and Risky Debt Maturity Choice.” Journal ofFinance 4 1:19-37. Francis, J., and J. Stephan. 1993. “Characteristics of Hedging Firms: An Empirical Examination.” In Advanced Strategies in Financial Risk Manugement, edited by R. J. Schwartz and C. W. Smith, 615-635. New York: New York Institute of Finance. Froot. K., D. S. Scharfstein, and J. C. Stein. 1993. “Risk Management: Coordinating Corporate Investment and Financing Policies.” Journal ($ Finance 48 (December): 1629-1658. GAO. 1994. Financial Derivatives-Actions Needed to Protect the Financial System. Washington, D.C.: U S . General Accounting Office. Geczy, C., B. A. Minton, and C. Schrand. 1997. “Why Firms Use Currency Derivatives.” The Journal of Finance 52 (September): 1323-1354. Goodman, L. S. 1993. “The Use of Interest Rate Swaps in Managing Corporate Liabilities.” In Advanced Strategies in Financial Risk Management. Edited by R. J. Schwanz and C. W. Smith, 103-127. New York: New York Institute of Finance. Guay, W. R. 1997. “The Impact of Derivatives on Firm Risk: An Empirical Examination of New Derivative Users.” Working paper, University of Rochester. Guedes, J., and R. Thompson. 1995. “Tests of a Signaling Hypothesis: The Choice between Fixed and Adjustable-Rate Debt.” The Review of Financial Studies 8 (Fall): 605436. Hentschel, L., and S. P. Kothari. 1995. “Life Insurance or Lottery: Are Corporations Managing or Taking Risks with Derivatives?” Working Paper, University of Rochester. Hoshi, T., A. Kashyap, and D. Scharfstein. 1991. “Corporate Structure, Liquidity and Investment.” Quarterly Journal of Economics 106 (February): 33-60. Institutional Investor. 1994. “Companies’ Derivatives Trading: Complexity and Hypocrisy.” May: 1112. Jensen, M. C., and W. H. Meckling. 1976. “Theory of the Finn: Managerial Behavior, Agency Costs and Ownership Structure.” Journal of Financial Economics 3 (October): 305-360. Kaplan, S. N., and L. Zingales. 1997. “Do Investment-Cash Flow Sensitivities Provide Useful Measures of Financial Constraints?” The Quarterly Journal of Economics 112 (February): 169-215.

200

JOURNAL OF ACCOUNTING, AUDITING & FINANCE

Kuprianov, A. 1994. “The Role of Interest Rate Swaps in Corporate Finance.” Economics Quarterly 80 (Summer): 49-68. Litzenberger, R. H. 1992. “Swaps: Plain and Fanciful.” The Journal of Finance 47 (July): 831-850. Maddala, G . S. 1983. Limited-Dependent and Qualitative Variables in Econometrics. New York: Cambridge University Press. -. 1991. “A Perspective on the Use of Limited-Dependent and Qualitative Variables Models in Accounting Research.” The Accounting Review 66 (October): 788-807. Mian, S. L. 1996. “Evidence on Corporate Hedging Policy.” The Journal of’Financia1 and Quantitative Analysis 31 (September): 4 1 9 4 3 9 . Moms, J. R. 1976. “On Corporate Debt Maturity Policies.” Journal of Finance 31 (March): 29-37. Myers, S. C. 1977. “Determinants of Corporate Borrowing.” Journal i f Financial Economics 5 (November): 147- 175. Nance, D. R., C. W. Smith, and C. W. Smithson. 1993. “On the Determinants of Corporate Hedging.” Journal of Finance 48 (March): 267-281. Opler, T., and S. Titman. 1996. “The Debt-Equity Choice.” Working Paper, The Ohio State University. Pregibon, D. 198 I , “Logistic Regression Diagnostics.” Annals of Statistics 9:705-724. Press, E., and J. Weintrop. 1990. “Accounting-Based Constraints in Public and Private Debt Agreements: Their Association with Leverage and Impact on Accounting Choice.” Journal of Accounting and Economics 12 (January): 65-95. Santomero, A. M. 1998. “Financial Risk Management: The Whys and Hows.” S h n Manugenierit Review (forthcoming). Scholes, M., and M. Wolfson. 1992. Tuxes and Business Strategy-A Planning Approach. Englewood Cliffs, NJ: Prentice-Hall. Schrand, C. M. 1997. “The Association between Stock-Price Interest Rate Sensitivity and Disclosures about Derivative Instruments.” The Accounting Review 72 (January): 87-109. Securities and Exchange Commission (SEC). 1995. Proposed Amendments to Require Disclosure cf Accounting Policies ,for Derivatives. Washington, D.C.: SEC. Smith, C., and R. Stulz. 1985. “The Determinants of Firms’ Hedging Policies.” Journal of Finuncial and Quantitative Analysis 20:39 1405. Smith, C., C. W. Smithson, and L. M. Wakeman. 1986. “The Evolving Market for Swaps.” Midland Corporate Finance Journal 3 (Winter): 20-32. Stohs, M. H., and D. C. Maur. 1996. “The Determinants of Corporate Debt Maturity Structure.” Journal Uf Business 69 (July): 279-312. Titman, S. 1992. “Interest Rate Swaps and Corporate Financing Choices.” Journal of Finance 47 (September): 1503-15 16. Titman, S., and R. Wessels. 1988. “The Determinants of Capital Structure Choice.” The Journal of Finance 43 (March): 1-19. Tufano, P. 1996. “Who Manages Risk? An Empirical Examination of Risk Management Practices in the Gold Mining Industry.” The Journul of Finance 5 1 (September): 1097-1 137. Wall, L. 1989. “Interest Rate Swaps in an Agency Theoretic Model with Uncertain Interest Rates.” Journal of Banking and Finance 13 (May): 261-270. Wall, L., and J. Pringle. 1989. “Alternative Explanations of Interest Rate Swaps: A Theoretical and Empirical Analysis.” Financial Management 18 (Summer): 59-73.