The Use of Debt Covenants in Public Debt: The Role of Accounting Quality and Reputation

Transcription

1 The Use of Debt Covenants in Public Debt: The Role of Accounting Quality and Reputation Joy Begley Sauder School of Business University of British Columbia Sandra Chamberlain Sauder School of Business University of British Columbia 20 February 2005

2 1. Introduction The use of accounting numbers in lending agreements is considered to be an important part of the demand for accounting. The accounting choice literature hypothesizes that managers make accounting choices, such as manipulating net income, to avoid tripping accounting-based loan covenants. Much of this body of research presumes that debt covenants are frequently used and that the form of the covenants is standard boiler-plate. However, recent research by Begley and Freedman [2004] calls into question the prevalence and importance of loan covenants. They examine the use of accounting-based covenants in senior public debt issued in three different time periods (the late 1970s, and ), reporting that covenant use has declined sharply over the last three decades. In addition, they find that the form of debt covenants appears to be shifting away from measures that can be affected by accounting choices to measures that are closer to cash flows. While Begley and Freedman uncover some interesting data regularities, they do not investigate explanations for this dramatic change. This study is concerned with understanding what factors influence the choice of debt covenants in public lending agreements and what has changed over time to alter the observed distribution of covenants. Our analysis endeavors to explain cross-sectional covenant use with traditional agency problem proxy variables, plus variables to indicate reputation and accounting quality effects. We build on an existing literature that uses agency theory to explain debt covenant use. 1 This literature proposes that debt covenants are used to control conflicts of interest between stockholders and bondholders. One possible explanation for why covenant use 1 See for example Smith and Warner (1979), Begley (1994), Begley and Feltham (1999) and Nash, Netter and Poulsen (2003). 1

3 has changed over time is that the types of firms issuing public debt may have changed. There have been many innovations to corporate financing alternatives over the last three decades. In the late 1970s the high yield, junk bond market (Bond rated below BBB) came into vogue attracting many high risk companies to the public debt market. This continued into the 1980s with the onset of the LBO market. While in the 1990s there was an explosion in the use of asset securitizations. 2 It is therefore possible that, the earlier time periods saw riskier firms, with more opportunities for agency problems, raising capital in senior public debt markets, but a similar firm today is able to rely on other sources of financing 3. The market for public debt has therefore become much more specialized and increasingly sophisticated over time. After nearly three decades of dealing with high yield debt, investors are better able to understand its performance and risks relative to investment grade debt (bonds rated BBB and above). The decline in covenants reflects a movement away from contractual control of agency problems. If firms are returning frequently to debt markets to issue more public debt, then concerns about their future reputation and the interest rate they will have to pay on their return can be act as an indirect control on their behavior. Rating agencies play a role in this type of monitoring. Firms that maintain a high debt rating are able to borrow at more attractive 2 Dechow, Myers and Shakespeare (2004) report that the market for asset securitizations has grown from $2.4 trillion in 1995 to $6.8 trillion in They note that manufacturing and retail firms often rely on securitization of receivables as a means of financing. This type of financing is typically raised through the use of a special purpose entity, moving the asset and the debt off the balance sheet and outside of the legal control of the original company. 3 Smith and Warner (1979) suggest that issuing secured debt can be considered a substitute for using covenants to control agency problems. Asset securitization creates a claim on the securitized assets that is similar to the claim of secured debt. The protection to lenders is even greater than that of secured debt because the cash flows from the assets are not intermingled with the cash flows of the rest of the firm. 2

4 interest rates than firms in lower rating categories. Rating agencies monitor firms and adjust ratings downward as default risk increases - indicating agency problems of debt have become more severe. Arguably, the incentive to maintain a high debt rating will help to control managements incentive to increase the risk of existing debt and this should be more important for firms that frequently return to debt markets for financing. We investigate whether a firm s frequency of issuing debt can be used to explain the absence of covenants. Accounting quality has received considerable attention within the accounting literature, particularly with respect to its impact on cost of equity and to a lesser extent the cost of debt. 4 Watts (2003) argues that debt holders value accounting conservatism as it provides a lower bound, liquidation value, on which to contract. This view, that covenants written using conservative accounting figures are of higher value, implies that for a healthy firm it is relative low cost to renegotiate a binding covenant. We explore an alternative view on accounting quality. The accounting ratios used in debt covenants provide a signal to lenders about firm health. If these signals become too noisy in either direction (both aggressive and conservative accounting), we expect their use to decline. We explore this issue with a balance sheet measure of firm-specific signal to noise ratio and investigate its ability to explain covenant use. We use the same debt covenant sample as in Begley and Freedman (2004), but expand 4 Ahmed et al. (2002) and Zhang (2005) examine the relation between various measures of accounting quality and the interest rate paid by a firm on its borrowing. Zhang (2005) examines whether a Basu (1997) measure of conservatism can be used to explain lower yield spreads in private debt and finds evidence that more conservative reporting is associated with lower spreads. Ahmed et al. (2002) examine two measures of discretionary conservatism, an accrual measure based on Givoly and Hayn (2000) and an adjusted market-to-book measure based on Beaver and Ryan (2000). They find that debt ratings (which they use as a proxy for interest cost) are more favorable for firms with more discretionary conservatism. 3

5 the final sample to cover a longer time period. We find evidence that traditional agency problem proxy variables such as cash flows and leverage continue to explain covenant use. When we add our reputation variable we find consistent evidence that firms that value reputation are less likely to include covenants in the last two periods. The results regarding our accounting quality hypotheses are mixed. In the first two periods accounting quality does not appear to impact covenant choice in the manner we predict, while in the final period there is evidence that covenants are more frequently used when accounting is less conservative, i.e. market values are closer to book values. This is counter to Watts (2003) prediction. This paper contributes to the literature on debt contracting in a number of ways. Nash, Netter, and Poulson (2003) examine how cross-sectional variation in covenant use can be explained by firms desiring financial flexibility. In this paper we are able to examine this hypothesis over time and to examine the added role of reputation and accounting quality. This is the first study we are aware of that considers the impact of reputation on debt contracting. Prior studies that have looked at accounting quality and debt look at whether firms with conservative accounting are rewarded with lower cost debt, but they do not examine whether the public debt of such firms is more likely to include covenants. In this study we are able to explore this question. In the next section we look at how the use of debt can result in agency problems and the tradeoffs in using restrictive covenants to control those problems. The empirical model is presented in section three and the sample is described in section four. In section five we test the hypotheses and section five provides our conclusions and directions for future work. 4

6 2. Background and Hypothesis Development 2.1 Agency Problems and Debt In this study we are interested in the role of accounting based debt covenants in public debt. These covenants are predominantly restrictions on the payment of dividends and on new borrowing. To understand when these covenants are likely to be used we must first consider the broader agency setting in which they arise. The existence of risky debt in a firm s capital structure motivates management, acting on behalf of shareholders 5, to take actions that are not in the best interest of the firm as a whole. For example, once risky debt is outstanding shareholders would prefer that managers pay higher dividends, invest in riskier assets and use more leverage than what they would have preferred if the firm were all equity financed. 6 These actions increase the riskiness of the outstanding debt, causing it to decline in value. In this way, equity holders do not bear all of the increased risk associated with these actions. The magnitude of potential agency costs varies based on the characteristics of the firms assets. A firm with its assets already in place is likely to be less subject to agency problems compared to a firm that has yet to fully determine its investment strategy. In addition, the option like nature of equity means that shareholders have a greater incentive to take value reducing, risk increasing actions as the risk of bankruptcy increases. Companies with high risk assets will tend to favor equity financing over debt financing to 5 Begley and Feltham (1999) relax the assumption that managers act as if they hold all of the equity in their firm. The alignment of management and shareholder interests is typically assumed by researchers examining conflicts of interest between debt and equityholders. 6 Smith and Warner (1979). 5

7 completely avoid the agency costs associated with debt. To the extent that debt financing has associated benefits, and agency problems are not excessive, some firms will choose to use debt and deal with the agency problems that arise. One possibility is that debt is issued at a price that fairly reflects anticipated agency costs of management s uncontrolled behavior. If shareholders can establish a reputation for acting in the interest of the firm as a whole, rather than on their own behalf, then a desire to maintain this valuable reputation for fair dealing could control ex post opportunism. 7 Alternatively, agency costs can be reduced through a variety of debt contracting features. 8 Agency costs associated with debt can be reduced (but not necessarily eliminated) by issuing shorter term debt, issuing secured debt, transferring ownership rights to the lender (i.e. a sale and lease back arrangement) and moving the debt and the assets that will be used to service the debt to a separate company. Restrictive debt covenants, the focus of this study offer another means to reduce agency costs associated with debt. 9 In general, these mechanisms are not likely to reduce agency costs entirely. However, they can be used in combination to minimize the problem. 2.2 Covenants as a Specific Control for Debt-related Agency Problems Holding constant the other control mechanisms, we now turn to consideration of what determines debt holders demand for covenants and management s willingness to agree to include 7 There are numerous references in the literature to the value of reputation in controlling agency costs. For example, Milgrom and Roberts (1992) on page 257 cite Adam Smith s Lecture on Justice, Police, Revenue, and Arms, Edwin Cannan, ed. (New York: Augustus M. Kelley, 1964) which argues the greater a merchant s volume of business the greater the incentive to act honestly to protect his or her reputation. 8 Jensen and Meckling [1976] and Smith and Warner [1979]. 9 This is not an exhaustive list of control mechanisms. For example, bank monitoring and restrictive covenants in private debt issues may provide indirect protection to all lenders. 6

8 them. This discussion will ultimately lead us to a discussion of the role of accounting quality in debt contracting and the choice of control variables to include in our analysis. Two factors that are expected to impact supply and demand for debt covenants are (i) the inherent magnitude agency costs implicit in the firm s assets; and (ii) the ability of the covenants to detect or miss agency problems when they exist (i.e. the likelihood of a Type I error); versus, the likelihood that a covenant will be triggered when an agency problem does not exist (i.e. a Type II error) and the implication for restricting management s actions. The first factor indicates which firms are expected to benefit more from including a covenant. The second factor incorporates the tradeoff between Type I and Type II errors and the cost each imposes on the firm. These factors are not independent. Firms that are more likely to face agency problems stand to benefit more from the inclusion of covenants, however, they may also face higher costs due to the imprecision of the covenants. With regard to the first factor, the inherent magnitude of agency costs, prior studies have examined the choice of accounting based debt covenants based on anticipated agency problems. Future agency problems are likely to be larger the greater the uncertainty about the future payoffs to the firm and the associated higher likelihood of corporate failure. Studies have included variables such as volatility of cash flows, indicators of high growth and bankruptcy predictors as proxy variables for when agency problems are likely to be more severe and covenants more beneficial. The importance of Type 1 and Type 2 errors depends on the type debt (public vs. private), and the type of firm that is borrowing the money. Public debt covenants are typically written to restrict management s actions (e.g. the payment of a dividend or the issuance of 7

9 additional debt) if a particular accounting condition arises. 10 Ideally a covenant only becomes binding, when conditions have shifted such that without the covenant restriction management would prefer the restricted action which reduces overall firm value. Covenants based on accounting numbers are likely to be blunt instruments for identifying and controlling these agency costs. On the one hand there is the danger that a covenant will fail to identify some cases of agency problems (Type I error). This could occur either because GAAP accounting is unable to capture the shift in the underlying economics or because management discretion over reported figures allows strategic avoidance of default. Note that theoretically the likelihood of these Type I errors occurring should be priced by debt holders. Studies that view conservative accounting rules as preferable for debt contacting (Watts (2003), Zhang (2005)) are focused on a setting where the goal is to minimize Type I errors. As Type I errors are reduced Type II errors are likely to increase. Recall that Type II errors occur when agency problems are not present, but the covenant becomes unintentionally binding and prevents management from taking firm value increasing actions. Type II errors will be costly if they force management to take actions that are not in the interest of the firm as a whole. For example, a binding dividend restriction prohibits dividends, even if returning cash to shareholders is of greater value than retaining cash in the firm. A binding additional borrowing covenant will prevent the firm from raising additional debt. This could be extremely costly to a firm facing positive NPV investment projects and with insufficient internal funds or new equity available to finance the projects. Type II errors will be costly if they restrict management 10 This is substantially different from the type of covenant observed in private debt, where a covenant default gives the lender the option to accelerate debt repayment. 8

10 flexibility to take firm value increasing actions. These Type II error costs are likely to be greater for growth firms with greater uncertainty about their future prospects, than for firms that already have the majority of their assets in place. 11 Private debt may be less subject to these costs because the lender has the ability to monitor the firm and grant a default waiver if a covenant default occurs when the firm is healthy and not subject to agency problems of debt. However, the covenants in public debt cannot be waived as easily because the lenders are disperse and the contract typically requires that a covenant change be approved by at least 2/3 of the outstanding debt. An important insight is that the cost of lost flexibility that is associated with minimizing Type I errors implies that conservative accounting may not be ideal for all firms writing debt contracts. The sum of Type I error costs and Type II error costs are the total agency costs of debt. Accounting based debt covenants, and other control mechanisms will be used if they can reduce this total. The accounting methods that are best able to reduce this total are not necessarily the ones the accounting literature has labeled as high quality accounting. 2.3 Accounting Quality and Debt Contracting Francis, LaFond, Olsson and Schipper (2004) examine seven earnings attributes that have been characterized as desirable features of earnings. They look at four accounting-based attributes: accrual quality, persistence, predictability, and smoothness; and three market-based attributes: relevance, timeliness and conservatism. These measures are all based on accounting numbers from the income statement, however, the value of the income statement in debt contracting is not clear. To the extent that agency problems are anticipated when firm value 11 Begley (1994) and Nash, Netter, and Poulson (2003) 9

11 declines and the likelihood of corporate failure becomes a major concern then debt covenants will be more valuable if they can identify this condition. The earnings number reported in a single year is not likely to be as important in identifying potential agency problems as is the stock of wealth that shareholders have invested in the firm. As shareholders interest in the firm declines, agency costs of debt are more likely to arise. This implies that a debt covenant based on the ratio of debt to market value of equity might provide a better indicator of agency problems. We do not observe such a measure in debt contracts, possibly because the market value of equity is subject to fluctuations that are outside of the control of the firm s managers. If a debt covenant that tracks a firm s market based leverage ratio is better able to reduce agency costs, and if debt contracting cannot be based on market values directly, then accounting numbers will be better suited for debt contracting when the ratio of market value to book value of equity is close to one. 2.4 Hypotheses and Measurement The discussion in this section, leads to a number of hypotheses. Firms with different types of assets are more or less likely to experience agency problems of debt due to the inherent nature of their assets: H1: The use of accounting-based debt covenants is positively related to the level and volatility of a firm s cash flows and to its likelihood of corporate failure. We use the standard deviation of EBITDA to proxy for the volatility of cash flows and we use leverage to measure the risk of corporate failure. We also include the collateral value of fixed assets to measure assets-in-place, which is assumed to indicate a lower risk of debt default 10

12 and corporate failure. If Type I errors are more costly to firms than Type II errors, as is assumed in studies that examine the effect of accounting quality on debt pricing (Zhang (2005) and Ahmed et al. (2002)), then firms using more conservative accounting will be able to write more efficient covenants. This leads to the following prediction: H2: The use of accounting-based debt covenants is positively related to the extent to which a firm s accounting is conservative. We use the natural log of market value of equity over book value of equity to proxy for accounting conservatism. 12 A higher market-to-book ratio indicates that accounting book values are conservative relative to the market s valuation the equity. If accounting quality for debt contracting is based on proximity of accounting book value to market value rather than on conservatism then this implies a different predication: H3: Firms with a market to book value ratio closer to one will be more likely to include accounting based debt covenants. This hypothesis differs makes the same prediction as hypothesis H2 when book values are aggressive (i.e. when market value of equity is less than book value) but makes an opposite prediction when accounting is conservative (i.e. when market values exceed book values). In order to test hypothesis H2 and H3 jointly, we estimate separate coefficients on the log of market-to-book value of equity depending on whether market-to-book is above or below one. If conservative accounting makes contracting less costly then both coefficients should be positive. 12 The analysis in this paper only examines one measure of accounting quality. We are in the process of considering other measures for inclusion. We are currently collecting the data to estimate the Basu (1997) measure of conservatism. 11

13 If proximity of market values to book values makes contracting less costly then the coefficient on conservative accounting will be negative and the coefficient on aggressive accounting will be positive. The ratio of market-to-book is also likely to reflect a firm s growth opportunities. For growth firms, accounting-based covenants are expected to be costly because of the reduction in financial flexibility. However, growth firms typically face an uncertain future, increasing their likelihood of becoming distressed and making the benefits of including accounting-based covenants potentially greater. It is unclear which effect will dominate. Nash et al. (2003) find that growth firms use less covenants, indicating the desire for financial flexibility dominates. If this is true in our sample as well, then we expect conservatism to interact with growth. In these cases conservatism is no longer desirable and H2 is less likely to hold. We include R&D expense to total assets and average sales growth over the past five years to proxy for growth opportunities. As discussed earlier agency costs can be controlled by a firms desire to maintain a reputation for honest dealings with debt holders. A firm that returns regularly to public debt markets to raise more financing stands to gain more from maintaining its reputation and thereby being able to sell its debt at a higher price (pay a lower coupon) compared to another firm that issues public debt infrequently. If the value a firm places on its reputation helps to control agency problems then firms that return regularly to debt markets are less likely to use debt covenants to control agency problems, leading to the following prediction: H4: Firms that are regularly issuing public debt are less likely to include debt covenants. We measure the number of other public debt issues (non-subordinated, non-convertible, senior 12

14 public debt) the firm makes during the same sample period to indicate how regularly the issuing firm returns to public debt markets. 3. Methodology Consistent with prior studies we use a combination of default risk variables and growth measures to predict covenant use. We add to these variables, proxies for reputation and accounting quality to investigate their incremental explanatory power. The dependent variable, covenants, is a zero-one dichotomous variable, therefore we use a logit regression model to investigate the relation of covenants with our hypothesized explanatory variables. Based on the discussions and hypotheses in section 2 our model takes the following form: where: Collat = Net PPE / Total assets; R&D = Research and Development Expense / Total assets; SalesGrwth = (Sales t - Sales t-1 ) / Sales t-1 EBITDA = Operating income before depreciation / Total assets; Volatility = Standard deviation of Operating income before depreciation / Total assets; BVLev = Total Debt / Total assets; ActgCons = Ln(MVE/BVE) * Dummy indicating if MVE/BVE >1.0; ActgAggr = Ln(MVE/BVE) * Dummy indicating if MVE/BVE <1.0; Reputation = 1 if the issuing firm made more than one debt issue during the sample period, 0 otherwise. All variables other than Volatility and Reputation are measured as the arithmetic mean of the variable over five years prior to the debt issue date (or as many years as there are available on COMPUSTAT). 13

15 Hypothesis H1 predicts the coefficients on Collat and EBITDA will be negative and the coefficients on Volatility and BVLev will be positive. Hypothesis H2 predicts a positive coefficient on ActgCons and ActgAggr, while hypothesis H3 predicts a negative coefficient on ActgCons and a positive coefficient on ActgAggr. The coefficient on Reputation is predicted to be negative under H4. 4. Sample and Data We examine public debt contracts from three time periods: ; ; and restricting our analysis to new issues of straight debt 13 by industrial companies reported in the Directory of Corporate Financing Decade or in the electronic equivalent, Securities Data Corporation new issues data base. We hand collect debt contract information directly from indentures or prospectuses. The and samples contain both notes and debentures 14, while the 1970s sample, obtained from a previous study, contains only debentures. The 1970s sample was previously examined in Begley and Feltham (1999). They make an alphabetical list of all new public debentures issued between 1975 and 1979 and select every second debt issue and limit their sample to one debt issue per firm. This results in a final sample 13 Convertible, guaranteed, subordinated, and variable rate public debt are excluded because these features may act as substitutes for covenants. Excluding them from the sample avoids variation in these features over time and reduces the sample to a manageable size for hand collecting covenants. If these features substitute for covenants then the excluded debt issues are likely to use fewer covenants than the debt issues in the sample. 14 What distinguishes a note from a debenture is term to maturity of the debt issue. Debentures typically have an initial term to maturity of between 15 to 30 years. Notes typically have a term to maturity of ten years or less. Ideally, the 1970s sample would also include notes. However, contracts from the 1970s are not available electronically, making it extremely costly to collect the prospectuses governing these notes. 14

16 of 90 debenture issues. The sample includes one randomly-selected new public debt issue per firm for a total of 283 new issues (86 of which are debentures). The most current sample is selected in the same manner as the sample, but from the period , resulting in a sample of 250 debt issues (including 14 debentures). 15 Ideally we would select all three samples using a common procedure. But because the 1970s sample is drawn from an earlier study its selection procedure differs from the random-sampling procedure used in the two later samples. The main difference between the first sample and the two later samples is that it consists entirely of debentures. It is not clear whether notes or debentures are likely to use more covenants. The firms issuing public debt of the type examined here comprise only a small fraction of all publicly-traded companies during each of our sample periods. Our samples exclude nonindustrial firms and firms issuing public debt with specialty features like a conversion option or subordination. However, the low numbers also reflect that many firms rely entirely on private debt and equity to finance their operations. Debt issue data is obtained from SDC Platinum and accounting data is obtained from Standard and Poor s COMPUSTAT. Table 1 describes the debt issue characteristics by sample period. We find, as do Begley and Freedman (2004), the incidence of covenants is decreasing across the three sample periods. The existence of an accounting-based covenant restricting dividend payments (DIV) has dropped from 39% of debt issues in the late 1970's to 25% of issues 15 The first and second samples are the same as in Begley and Freedman (2004), the third sample is larger than their sample as it includes debt issues over a longer time period ( ) than they examine ( ). In randomly selecting the larger sample some of the original debt issues are replaced by other debt issues by the same firm in the later time period. 15

17 in the second period, and finally to only 9% of debt issues in the most recent sample. 16 The restriction on additional borrowing (DEBT) has seem a similar decline. The variable COVENANT is set to equal 1 if either the dividend or the additional borrowing covenant are present. Given the high degree of overlap between the DIV and DEBT covenants, we will perform our analysis on COVENANT rather than on the individual covenants. The mean time- to- maturity also declines across the three sample periods. A decline from period one to two is to be expected, because the first sample is made up entirely of debentures which have a longer term-to-maturity. The decline from period two to period three is also consistent with a further decline in debentures from 30% of sample two to 6% of sample three. Average debt ratings are similar in the second and third sample periods, but are significantly higher in the earliest period. The higher debt ratings imply a lower risk default in the 1970s sample period. The median debt issue size as a percentage of debt both prior to and following the issue shows weak evidence of a reduction in debt issue size over the three time periods. Table 2 provides descriptive statistics on the firms issuing the sample debt issues. Firm characteristics are measured based on their average value during the five years prior to the date of the debt issue. Leverage relative to total assets is higher in periods two and three than in the earliest period, but it does increase between periods two and three. In contrast, leverage relative to the market value of assets (which is defined as market value of common equity plus book value of liabilities and preferred stock) does not change between the first two periods but it declines significantly in period three. This difference in direction for leverage reflects the fact that market 16 Looking at debentures only the drop off is even more dramatic. Only 4% of debentures in the middle period have a dividend restriction and none of the debentures in the final period include this covenant. 16

18 values are much higher at the end of the1990s than in earlier times. The higher market values for sample three is due, in part, to a change in the nature of firms issuing debt in period three and partly due to an increase in market-to-book ratios across all firms including our sample firms. Firm size, measured by total assets adjusted for inflation 17, is also higher in period three, but it does not change significantly in the first two periods. There is very little change in ability to generate cash flows, measured by EBITDA to assets or collateral value of assets, measured by Net PPE to assets, across the three sample periods. However Volatility of cash flows declines between period one and two, indicating a decrease in default risk. R&D spending as a percentage of assets and sales growth does not change significantly between the second and third sample periods but the median declines between periods one and two, suggesting that firms in the later two periods are not growing as fast as in the first period. This result is in sharp contrast to the increase in market-to-book ratios across the three sample periods. The higher market-to-book ratios are also leading to higher conservatism measures between periods one and two and between two and three, while aggressive accounting appears to have declined from what it was in period one. Finally reputation appears to be more important in the later two periods compared to the earlier period. In the earlier period about two thirds of companies had only one debt issue during the sample period, but in the later two periods over half of the companies made multiple debt issues. Rank correlations are reported in Table 3. Group 1 correlations are reported in Panel A. In the first period sales growth and R&D spending are significantly positively correlated with market-to-book in this period suggesting they are all capturing similar aspects of growth. Panel B 17 The GNP index is used to adjust total assets to dollars in

19 reports period two correlations. Sales growth and market-to-book are also significantly positively correlated in this period, but R&D spending is not. In period three R&D spending, sales growth and market-to-book are not significantly correlated, indicating that the market is valuing something other than rents from R&D spending and an extrapolation of historical growth rates. 5. Logistic Analysis The logit results are reported in Table 4. For the earliest sample period (reported in Panel A) a number of base model variables are significant in explaining covenants. Sales growth, and a book value measure of leverage are significantly positively related to covenants, while collateral value of assets and EBITDA are negatively related. These relations are in the direction predicted. The positive correlation with size suggests that firms experiencing growth are subject to higher agency problems causing them to use more covenants. The negative relation with collateral value of assets and EBITDA indicates that profitable firms and those with their assets in place are less concerned about agency problems of debt. When the accounting quality and reputation variables are added to the base model, results for the base model variables are very similar and the additional variables are not significant. Results for the second sample period are reported in Panel B of Table 4. In the base model leverage is significantly positively related to covenants, while EBITDA and R&D spending are significantly negatively related to covenants. The significant negative relation with covenants suggests that firms growing through R&D spending value their financial flexibility and therefore less likely to use covenants. In the second sample period, the reputation variable is significantly negative as predicted, indicating that reputation may be substituting for covenants for firms that are making multiple debt issues. The accounting quality variables and not 18

20 significant when added to the base model. When all variables are included in the final model, aggressive accounting is significantly negative, counter to our prediction. In the third sample period volatility of EBITDA and leverage are significantly positively related to covenants and EBITDA is significantly negatively related to covenants, consistent with our predictions regarding risk of corporate failure. When reputation is added to the model it is again significantly negative as predicted. When the two accounting quality measures are added, conservative accounting is significantly negatively related to covenants. This result does not support H2, the prediction that conservative book values are preferred for debt contracting, but rather it does support H3, the prediction that book values closer to market value are more suitable for debt contracting. However, the negative relation with conservative accounting, is also consistent with this variable capturing the demand for financial flexibility in high growth firms. Sales growth is not significant in the period three base model, but it becomes significantly positive in the full model. Overall the regression results indicate that underlying firm characteristics that proxy for agency problems of debt continue to related to covenant use in all three periods. Significant reputation effects are found in the second and third sample periods suggesting concerns about maintaining bond ratings is substituting for covenants. The results for our accounting quality measures are not strong, although there is some evidence in the third period that high market-tobook ratios (i.e. conservative book values) are not suitable for debt contracting. 6. Conclusion We have examined the determination of the existence of debt covenants in senior, public debt issues over the period , , and This analysis is motivated in 19

21 part by the sharp decline in the use covenants in public debt over these periods. We examine both the explanatory power of traditional agency cost measures, and the explanatory power of two previously unexplored factors: the role of reputation in implicitly controlling agency costs associated with debt, and the role of changes in the quality of earnings. In all three time periods we document that debt covenant use is related to traditional variables that capture the riskiness of assets in place, and the likelihood of corporate failure. Growth opportunities, as measured by research and development expenses and historical sales growth, are positively related to covenant use in the earliest sample, are negatively related to covenant use in the middle sample, and are unrelated to covenant use in the third period. Since there are good theoretical arguments for growth to relate to covenant use with either a positive or negative sign, our documented changes in the sign and significance of the coefficient over the sample periods are not necessarily surprising. We find that the relation between firm reputation (as measured by the frequency of debt issues by our sample) and covenant use is zero in our first period and is negative in our second two periods. This evidence is consistent with an increasing preference on the part of borrowers to rely on implicit rather than explicit control mechanisms to mitigate agency costs in debt contracting. This, we believe, is an intriguing result that warrants further investigation using refined measures of reputation, and perhaps a supplemental analysis of the ability of debt rating agencies to signal potential agency costs through ratings. Our results on accounting quality appear to contrast with some accepted intuition on the benefits of accounting conservatism. Most notably, across all three periods we find no evidence that covenants are more likely when accounting is conservative (i.e., covenants are not more 20

22 likely when market to book ratios exceed 1.) This finding contrasts with Watts (2003) who suggests that conservatism is beneficial in debt contracting. The finding also appears to contrast with results in Ahmed et. al (2002) and with Zhang (2005) that find conservatism lowers the interest rate charged to borrowers. We have argued that in public debt contracts, borrowers and lenders are unlikely to be able to engage in low-cost renegotiation when a debt covenant is tripped, and that Type II Errors can be costly to firms. Therefore, we do not view conservatism as necessarily a favorable attribute for the sorts of debt issues we study, and we are not surprised that conservatism is not associated with covenant use. However, there is obviously work to be done to reconcile our findings with these prior studies which use different measures of conservatism than we do and which investigate debt pricing rather than covenant use. 21

23 BIBLIOGRAPHY Ahmed, A., B. Billings, R. Morton and M. Stanford-Harris (2002), The Role of Accounting Conservatism in Mitigating Bondholder-Shareholder Conflicts over Dividend Policy and in Reducing Debt Costs, Accounting Review, Vol 77 (No 4) pp Basu, S. (1997) The Conservatism Principle and the Assymetric Timeliness of Earnings, Journal of Accounting and Economics, Vol 24 (No 1), pp Begley, J. and G. Feltham (1999) An Empirical Examination of the Relation Between Debt Contracts and Management Incentives, Journal of Accounting and Economics, Vol 27, pp Begley, J. and R. Freedman (2004) The Changing Role of Accounting Numbers in Public Lending Agreements, Accounting Horizons, Vol. 18 (No. 2) pp Francis, J., R. LaFond, P. Olsson and K. Schipper (2004) Costs of Equity and Earnings Attributes, Accounting Review, Vol 79 (No 4) pp Nash, R, J. M. Netter, and A. B. Poulson (2003) Determinants of Contractual Relations Between Shareholders and Bondholders: Investment Opportunities and Restrictive Covenants, Journal of Corporate Finance, Vol. 9, pp Smith, C. and J. Warner (1979) On Financial Contracting: An Analysis of Bond Covenants, Journal of Financial Economics, Vol 7 pp Watts, R. (2003) Conservatism in Accounting Part I: Explanations and Implications, Accounting Horizons, Vol 17, (No 3), pp Zhang, J. (2005) Efficiency Gains from Accounting Conservatism: Benefits to Lenders and Borrowers, Working Paper, Massachusetts Institute of Technology. 22

24 TABLE 1 Characteristics of Debt Contracts for Senior Public Debt Issued in (Group 1), (Group 2) and (Group3) Mean and (Median) Characteristics Comparisons Across Groups b Group2less Group3less Group3less Group 1 a Group2 Group 3 Group 1 Group 1 Group 2 Fraction with DIV Covenants Fraction with DEBT Covenants Fraction with Either COVENANT Term to MATURITY (25.00) (10.00) (9.90) RATING (11.00) (9.00) (9.00) NEWDEBT To Total Debt (-1) (0.301) (0.202) (0.187) NEWDEBT To Total Debt (+1) (0.236) (0.188) (0.153) Number Obs c Group 1 is drawn from the set of unsecured, senior debt issues reported in Directory of Corporate Financing , while Groups 2 and 3 are drawn from the Securities Data Corporation(SDC) new issues data base. Debt contract information is hand-collected from indentures or prospectuses, and only one debt issue is drawn per firm. DIV refers to instances where a debt contract restricts dividend payments; DEBT refers to those that restrict a firms ability to issue additional debt; and COVENANT is set to 1 if either covenant exists. Term to MATURITY is calculated as the number of days between the issue date and the maturity date, scaled by days and rounded to the nearest 10th. RATING translates debt ratings gathered from the SDC database to a scale from 1 to 12 where 12 indicates a higher rating, and therefore a lower interest rate. NEWDEBT is the face value of the debt issued. The table reports this relative to total debt in at the nearest annual report date preceding the debt issue (year -1) and at the nearest annual report date following the debt issue (year +1). a These columns report the mean and, in the row directly below the mean, the (median), of debt contract characteristics by groups. b These columns report the differences in means or medians reported in the first three columns., and indicate 1%, 5% and 10% significance for a test that the mean or median is the same across the two groups. c Due to lack of Compustat data, the number of observations drops to 90, 282, and 247 when we scale NEWDEBT by total debt at the first year-end following a debt issue. 23

25 TABLE 2 Pre-issue Characteristics of Selected Firms That Floated Senior Public Debt in (Group 1), (Group 2) and (Group 3) Mean and (Median) Characteristics Comparisons Across Groups b Group 2 less Group 3 less Group 3 less Group 1 a Group2 Group 3 Group 1 Group 1 Group 2 BVLev MVLev Ln (Market Value Assets) Total Assets-GNP Adjusted 5,879 7,934 11,295 2,055 5,416 3,361 2,779 2,825 4, ,021 1,975 EBITDA Collat Volatility EBITDA / Interest Exp R&D Exp SalesGrwth MV Assets/ Total Assets MVE/BVE (Winsorized) ActgCons ActgAggr Num Debt Issues per Firm Reputation Number of Observations c

26 TABLE 2 continued Pre-issue Characteristics of Selected Firms That Floated Senior Public Debt in (Group 1), (Group 2) and (Group 3) This table reports the pre-issue characteristics of firms that issued debt over three time periods. With the exception of Number of Debt Issues/Firm, Reputation and Volatility we measure firm specific characteristics as the mean of a given characteristic over the period that ends at the annual report date immediately preceding the debt issue, and begins up to four years before that. Therefore, the mean value of BVLev for Group 1 of.267 is the average of these five year individual firm means for Group 1 and the median of.250 is the median of these individual firm means. Group 1 is drawn from the set of unsecured, senior debt issues reported in Directory of Corporate Financing , while Groups 2 and 3 are drawn from the Securities Data Corporation(SDC) new issues data base. Data items are defined as follows: BVLev = Total Debt/Total Assets Total Debt = Long Term Debt plus Debt in Current Liabilities MVLev = Total Debt / (Market Value of Assets Market Value of Assets = Market Value of Equity plus Book Value of Liabilities and Preferred Stock EBITDA = operating income, before depreciation divided by total assets Volatility = The standard deviation of operating income, before depreciation measured for up to 5 years before the debt issue divided by mean assets computed over the same period. R&Dexp = Research and development expense divided by total assets. If research and development expense is missing it is set equal to zero. SalesGrwth = ((Sales t Sales t 1 )/Sales t 1 MBE/BVE (winsorized) = Market value of equity to book value of equity, winsorized at the top 2% of the distribution when negative values are eliminated. Firm-years with negative MBE/BVE are set equal to the top 2% of this distribution. ActgCons = this is equal to the log(mve/bve winsorized)if MVE/BVE, winsorized, is greater than 1; this variable is zero otherwise. ActgAggr = this is equal to the log(mve/bve winsorized)if MVE/BVE, winsorized, is less than 1; this variable is zero otherwise. Note that agressiveness increases as this variable declines in value. Num Debt Issues per Firm = this is the number of debt issues a firm released during the period covered in its grouping. Reputation = 1 if the issuing firm made more than one debt issue during the sample period, 0 otherwise. a These columns report the mean and, in the row directly below the mean, the (median), of debt contract characteristics by groups. b These columns report the differences in means or medians reported in the first three columns., and indicate 1%, 5% and 10% significance for a test that the mean or median is the same across the two groups. c The number of observations is approximate. For example, The Market Value of Equity is availble for only 260 firms in Group 2 and for 247 firms in Group 3; Volatility is available for 89 firms in Group 1, for 274 firms in Group 2 and for 247 firms in Group 3. 25

27 TABLE 3 Spearman Rank Correlations of Pre-issue Firm Characteristics and Debt Contracts Panel A: Group 1 Number Obs is Approximately 90 Matur- Debt Ln (MVE) BVLev EBITDA Volatility Collat R&D SaleGrth MV Assets ActgCons ActgAggr Num ity Rating Assets BV of Assets Issues Covenant Maturity Rating Ln(MVE) BVLev EBITDA Volatility Collat R&D SalesGrwth MV Assets/BV ActgCons ActgAggr