How much is too much? Debt Capacity and Financial Flexibility

Transcription

1 How much is too much? Debt Capacity and Financial Flexibility Dieter Hess and Philipp Immenkötter January 2012 Abstract We analyze corporate financing decisions with focus on the firm s debt capacity and its financial flexibility. We provide debt capacity estimates that are based on target ratings and novel financial flexibility measures that account for financial constraints due to the firm s debt capacity. We show that capital issuances and reductions are driven by concerns of preserving financial flexibility. Debt is issued and equity is repurchased when the firms can afford an increase of leverage due to unused debt capacities, and in contrast, debt is reduced and equity is issued when firms need to reestablish their financial flexibility. Our results offer implications for the interpretation of the trade-off theory and pecking order behavior. Keywords: debt capacity; credit ratings; capital structure; capital budgeting; JEL classification: G31; G32 We are grateful for valuable comments from Oliver Pucker and participants of the Corporate Finance Research Seminar, Cologne. University of Cologne, Corporate Finance Seminar and CFR Cologne, hess@wiso.uni-koeln.de, Tel Corresponding author, University of Cologne, Corporate Finance Seminar, immenkoetter@wiso.unikoeln.de, Tel

2 1 Introduction A common question in corporate finance is whether a firm can take on additional debt to finance its upcoming projects. While most theories of corporate capital structure rely on modeling different forms of costs and benefits of leverage to derive financing strategies, Graham and Harvey (2001) s survey on the practice of corporate finance reveals that preserving unused debt capacities is the most common determinant of financing decisions. In order to answer the question of how much debt is too much, we estimate firm-year specific debt capacities and analyze financing decisions in the light of preserving unused debt capacities. In this paper, we introduce novel measures of firm-specific debt capacities and financial flexibility to analyze capital issues. Extending the analysis of Altman (1968) and de Jong, Verbeek, and Verwijmeren (2011), we define a firm s debt capacity as the threshold of the debt ratio for keeping a target rating. The distance between the firm s debt capacity and its debt ratio is called the debt buffer and yields information on firms financial flexibility. The debt buffer before capital issues indicates how much additional debt the firm can issue and whether this is enough to finance its projects and settle its financing deficit. The debt buffer after capital issues indicates the firm s financial situation after investments and financing decisions and its unused debt capacities. Our main finding is that capital issuances and repurchases are driven by the target of preserving financial flexibility. If firms have a large debt buffer before financing activities, they issue debt, repurchase equity, or substitute equity with debt. Leverage increasing financing activities are only realized if the unused debt capacities are large enough to maintain the target rating after the debt issue. If firms cannot afford to take on enough additional debt to settle their financing deficit, i.e. the debt buffer is too small, they choose equity or dual issues. Firms that issue debt even though their debt buffer is too small have a high probability of losing their target rating and hence decide not to do so. Debt and dual repurchases as well as debt-to-equity substitutions are common for firms with large debt buffers, because they can afford an increase in leverage without facing financial constraints. The results show 2

3 that firms financial flexibility measured as preserving unused debt capacities is a common target of financing decisions. By preserving unused debt capacities, firms have money available to fund new projects and react to changing market conditions. The financing decision is not driven by past investment opportunities because the size of the financing deficit is independent of the choice between equity and debt. Our study provides firm-year specific debt capacity estimates. The debt capacity of a firm depends on the target of the financing strategies. If firms target on maintaining an investment-grade rating, BBB rated firm face the toughest constraints while AAA rated firms are relatively free in their financing decision. In a second specification, we assume that firms target on staying in their respective rating category and find that the amount of debt firms can issue decreases with with increasing ratings. On average, BBB rated firm can hold up to 30.3% of debt while firms in the highest rating category AAA can afford on average only 8.6% of debt. Our results are robust to various definition of the debt ratio in market and in book values and controlling for endogeneity as proposed by Molina (2005) does not affect our results. Our approach of explaining the capital issues with unused debt capacities determined through target ratings is motivated by Graham and Harvey (2001) s survey on the practice of corporate finance. In their study, preserving unused debt capacity means remaining financially flexible to be able to fund future investments and acquisitions. 59% of all CFOs consider financial flexibilty to be important, outranking all other relevant factors. Additionally, 57% of the CFOs named maintaining target credit ratings to be important for their financing decision, indicating a link between target ratings and debt capacities. We use this link and define the debt capacity of a firm as a function of the firm s target rating and other observable firm characteristics. We provide estimation techniques and evidence on a broad sample of rated and publicly traded firms that support Graham and Harvey s findings. The trade-off and pecking order theory support our approach of explaining capital issues with the firm s debt capacity and financial flexibility. In the dynamic trade-off theory with 3

4 adjustment costs (Fischer, Heinkel, and Zechner (1989)), managers balance the costs and benefits of debt financing to maximize the firm value. The debt capacity is the upper boundary of the resulting range of optimal capital structures. Firms within the range of optimal capital structures, but not close to the boundaries, are financially flexible, while those close to their debt capacity are constrained in their capital issuance decision. In Myers (1984) s pecking order theory, firms prefer internal over external financing and debt over equity due to adverse selection costs. The pecking order behavior offers an explicit measure of the debt capacity. The threshold when firms switch from debt to equity financing defines an intuitive measure for the debt capacity. Strict interpretations of both theories do not offer firms to be financially flexible because balancing of costs and benefits determines the issuance decision uniquely. A less strict interpretation offers firms choices for capital issues and introduces financial flexibility into the models. Even though both theories provide evidence for a theoretical foundation of the debt capacity and importance of financial flexibility, they do not offer an empirical strategy to test implications on the debt capacity. Therefore, we use the framework described above to measure debt capacities. Capital structure research has long been silent about firms debt capacities. Since the early work of Turnbull (1979), debt capacity remained only a small aspect in empirical and theoretical studies (Leland (1994), Morellec (2001)). Recently, three studies evolved which discuss empirical issues of measuring debt capacities. Lemmon and Zender (2011) measure firms debt capacities through a sample split that depends on the firms bond market access. Their measure does not provide firm-specific estimates and accounts only for the access to the bond market, but not for financial constraints due to the firm s capital structure. Leary and Roberts (2010) propose a novel empirical model to test the relevance of the pecking order and provide industry-specific debt capacity estimates. Under strict pecking order assumptions, their approach explains less than half of the observed capital issues. If they account for factors that are usually attributed to other theories, they can explain up to 80% of capital issues. In their study on financing decisions when the static 4

5 trade-off and the pecking order theory disagree, de Jong, Verbeek, and Verwijmeren (2011) provide firm-specific debt capacity estimates. We extend their methodology by introducing further debt capacity specifications and the debt buffer to explain issuance decisions. None of the three papers provides measures for firm s financial flexibility, nor do they combine preserving financial flexibility and capital issuing or repurchasing decisions. We extend the present state of the literature by providing further insights into firms debt capacities and financing motives. Our paper is structured as follows. In the following section, we explain the relation between capital structure theories, firm s financial flexibility and their debt capacity. In section three, we derive firm-specific measurements for debt capacities and analyze capital issuances with regard to the firms financial flexibility and their debt capacity and finally section four concludes. 2 Debt capacity and theories of capital structure In order to answer the question how much is too much, we need to find measurements of the debt capacity and financial flexibility. We define the debt capacity of a firm as the maximal amount of debt that is allowed in the firm s financing strategy. Knowing the firm s debt capacity, we say that the firm is financially flexible if its debt ratio is smaller than the debt capacity so that the firm has enough internal funds available to pursue new projects when they come along (Graham and Harvey (2001)). To build our study on solid theoretical grounds, we derive implications for the debt capacity and financial flexibility from the tradeoff and pecking order theory. The trade-off theory of capital structure assumes that there is an optimal capital structure for each firm that is the result of balancing costs and benefits of leverage and thereby maximizing the levered firm value. Costs of debt financing include default and agency cost, whereas benefits mainly correspond to tax savings. In a dynamic version of the trade-off 5

6 theory, the optimal capital structure varies over time and the firm can readjust its capital structure but faces issuance costs when doing so. Fischer, Heinkel, and Zechner (1989) show that the optimal capital structure can be extended to a range of optimal debt ratios because for small deviations from the optimum, the capital issuance costs might exceed the benefits of readjusting. In a rational setting without market frictions, the debt capacity equals the optimal debt ratio (Hong and Rappaport (1979)). Since capital structure adjustments are not costly, the firm will always stay in the optimum because deviations are not rational. If we take issuance costs into account, then the debt capacity is the upper boundary of the range of optimal capital structures. A firm never exceeds the debt capacity because the benefits of readjusting to the optimum outweigh the restructuring costs (Leary and Roberts (2005)). As long as the firm does not get close to the upper end of the range, it remains financially flexible. In a less strict setting where firms do not maximize the levered firm value, but focus on keeping the levered firm value larger than the unlevered asset value, the debt capacity corresponds to the debt ratio where benefits to leverage are fully offset by the costs. This interpretation of a less strict trade-off theory results in a debt capacity that is larger than in the fully rational setting. A mixture of both approaches, namely the firm trying to maintain a certain minimum level of benefits to leverage, leads to a debt capacity that lies between these two extreme cases. In Myers and Majluf (1984) s pecking order theory, firms have a hierarchical order of preferences for financial sources that is driven by adverse selection. Internal capital is always chosen over external capital because it has no adverse selection costs. If internal capital is insufficient, then debt is the next preference. Equity is chosen as last resort because equity issues have the largest information asymmetries. Pecking order behavior implies a threshold that defines the maximal amount of debt that a firm can issue. If the firm is in need of more external capital than admissible as indicated by the threshold, then the firm has to issue equity. 6

7 The pecking order theory offers an explicit measure of firms debt capacities. The firm s debt capacity can be defined as the implied threshold when firms switch from debt to equity financing (Leary and Roberts (2010)). The distance from the actual debt ratio to the debt capacity is a firm s debt buffer. In a strict interpretation of the pecking order, the issuance decision is uniquely determined by the relation of adverse selection costs and hence firms are not financially flexible. Since the explicit measurement of adverse selection costs is empirically difficult, a liberal interpretation of the pecking order that allows firms to have cash reserves offers firms to be financially flexible. If the financial deficit does not exceed their debt buffer, firms can choose between equity, debt, and a dual issue depending on which one is more adequate for them. Equity and dual issues preserve financial flexibility, while debt issues reduce it. The implications for debt capacities and financial flexibility depend on the validity of these theories. Neither theory offers an empirically testable measurement but both theories highlight the importance of the debt capacity and financial flexibility. In the following sections, we will derive a testing methodology that relies on the financial flexibility as a function of the debt capacity. 3 Estimating the debt capacity and financial flexibility The empirical measurement of the debt capacity and financial flexibility is of special relevance to our study. We introduce two measures of the debt capacity and define financial flexibility as the additional amount of debt a firm can issue without exceeding its debt capacity. Using these measurements, we give answers to the question How much is too much?. The survey of Graham and Harvey (2001) revealed that financial flexibility and maintaining target credit ratings are the two most relevant determinants of capital issuance decisions. Without using restrictive assumptions, one can view the debt capacity of a firm as an amount 7

8 of debt that a firm should not exceed (Turnbull (1979)). The precise realization of this concept depends on the targets of the firm s financing strategy. Kisgen (2006) documents a close relation between credit ratings and capital structure targets, indicating that the debt capacity is a function of a firm s target credit rating. We use this relation and define a firm s debt capacity as the debt ratio where the firm is about to lose its target rating. Since target ratings are not observable, we will discuss different measures of the debt capacity that are based on different target rating definitions. Ratings are important to firms because they influence their cost of capital significantly. A downgrade results in increasing cost of capital and upgrades usually reduce interest rates. Especially, the change of yields around the investment-grade boundary is of importance due to the reduced liquidity of securities in the speculative-grade category (Kisgen and Strahan (2010)). Consequently, the smallest investment-grade credit rating (BBB) depicts an intuitive target rating to preserve financial flexibility. In our first specification of the debt capacity, we will use the BBB rating as target rating and define the debt capacity as the debt ratio where the firm would fall below a BBB rating. In a second specification of the debt capacity, we follow Kisgen (2009) and argue that firms do not intend to maximize their rating, but rather target an individual minimum credit rating. Investment-grade firms target to keep an investment-grade rating, but not to lose more than one rating class. Firms rated BB or worse target on maintaining the current rating. Building on the different specifications of the debt capacity as a debt ratio that causes losing a target credit rating, we define the debt buffer as the distance from the debt capacity to the current debt ratio. This debt buffer measures a firm s financial flexibility because it indicates how much additional debt the firm can take on. A large distance from the observed debt ratio to the debt capacity gives the firm an ample scope in its financing decision, while a small debt buffer indicates that debt issuances significantly increase the cost of capital and reduce flexibility in financing decisions. 8

9 3.1 The Data Set To implement our debt capacity measures and test hypotheses about financing decisions, we need a sample of firms that are unconstrained in their access to financial markets. If firms are constraint in their access to finance, then their financing decisions might be rather determined by market entrance constraints instead of debt capacity concerns. Publicly traded firms with credit ratings have access to both the stock and the bond market and are therefore not constrained in their access to finance. Nevertheless, these firms are constrained in their financing decision if they have target credit ratings (Kisgen (2009)). Our sample is a panel data set of US-American firms with Standard&Poor s (S&P) Long Term Credit Issuer ratings during the period of 1985 up to 2011 listed in the COMPUSTAT annual file. To each firm-year observation we add the monthly S&P credit rating from one month after the report date of the annual financial statements. We add this additional month so that rating agencies have time to incorporate new information and adjust the rating. Report dates are taken from the COMPUSTAT quarterly file and missing report dates are replaced with the median time between the end of the fiscal year and the report date in the sample. We exclude financial firms and utilities (Standard Industrial Classification Code (SIC) and ) from the sample because their capital structure is subject to regulation. We drop observations with missing information and firms with negative sales or assets. Variables with extreme outliers are trimmed at the upper 0.1% level. We transform credit ratings into a discrete variable on an ordinary scale from 1 to 10, where 10 refers to a AAA rating, 9 to AA, 8 to A and so on, and do not distinguish between microratings (i.e. AA+ or AA-). We use two definitions of the debt ratio. The first follows Drobetz and Wanzenried (2006) and defines the debt ratio in market values as short-term debt plus long-term debt divided by capital. Capital is defined as the sum of long-term debt, short-term debt and the share price at the end of the fiscal year times the number of shares outstanding. This definition of the debt ratio captures the primary sources of external capital and does not include other use of debt, such as account payable and convertibles. The second 9

10 definition follows Baker and Wurgler (2002) and equals book liabilities over market assets. Book liabilities are calculated as total assets minus book equity and book equity equals total assets less total liabilities and preferred stock plus deferred taxes and convertible debt. This definition classifies a larger fraction of the assets as debt and displays a more conservative view on the capital structure. All results of our study remain the same if we define the debt ratio in book values and hence we do not report the results. Table 1 shows the summary statistics of our sample. Our final sample consists of 19,389 observation of 2,573 firms, on average seven to eight observations per firm. The median of firm-year observations per firm is 5 while only 9% of the firms stay more than 20 years in the sample. 95% of the firms have a rating between AA and B, while BBB and BB ratings are the most common ones. In comparison to prior studies, we have a larger data set over a longer period of time. The average firm has a debt ratio of 35.9% in the debt-to-capital definition and the debt ratio is increasing with declining ratings. AAA rated firms have 8.8% debt while AA firms carry 15.6% debt. With decreasing ratings, the debt ratio increases up to 80.5% for firms close to default. The order of the mean values indicates that target ratings go along with target debt ratios, namely an increase in rating is correlated with a decrease in the debt ratio. If the debt ratio is defined as liabilities to assets, then 46.4% of the assets are classified as debt. Again, the debt ratio is increasing with declining ratings. The average negative financing deficit in the top three rating categories indicates that highly rated firms have a financing surplus even though the mean value is close to zero. In the speculative-grade category, firms have a larger need for external capital indicated by the somewhat larger positive financing deficit. Net issues are defined as capital issues less repurchases denote the amount of external capital that the firm issues or repurchases. Due to the financing surplus of highly rated firms, we find that these firms repurchase capital on average (negative net issues). 10

11 Table 1: Summary Statistics. We report the mean values and standard deviation in parentheses of our variables by the different rating categories. We use annual COMPUSTAT data from 1985 to 2010 of 2,573 different firms with a S&P long-term credit issuer rating excluding financial firms and utilities. The debt-to-capital ratio is defined as long-term plus short-term debt over long-term and short-term debt plus share price times number of shares outstanding. The liabilities-to-assets ratio equals book debt over market assets. Book debt is calculated as total assets minus book equity and book equity equals total assets less total liabilities and preferred stock plus deferred taxes and convertible debt. Size is measured as log(sales) and profitability is EBITDA over total assets. Liquidity ratio 1 is retained earnings scaled by total assets and liquidity ratio 2 is working capital scaled by total assets. Tangibility is defined as property, plant and equipment over total assets. The financing deficit is defined in equation (11) and scaled by total assets. Net issues are issuance of long term debt less reductions plus sale of common stock less repurchases scaled by total assets. N denotes the number of observations. Ratings Variables All AAA AA A BBB BB B CCC CC-D debt/capital (0.240) (0.074) (0.126) (0.133) (0.163) (0.208) (0.244) (0.233) (0.259) liabilities/assets (0.227) (0.103) (0.136) (0.147) (0.169) (0.210) (0.248) (0.214) (0.087) size (1.521) (1.209) (1.400) (1.241) (1.176) (1.162) (1.284) (1.428) (1.522) profitability (0.078) (0.063) (0.052) (0.061) (0.060) (0.070) (0.083) (0.106) (0.101) liquidity ratio (0.469) (0.159) (0.205) (0.211) (0.207) (0.320) (0.626) (0.943) (0.659) liquidity ratio (0.176) (0.120) (0.152) (0.142) (0.148) (0.162) (0.189) (0.247) (0.482) tangibility (0.233) (0.212) (0.209) (0.209) (0.239) (0.243) (0.231) (0.250) (0.226) financing deficit (0.125) (0.043) (0.053) (0.072) (0.091) (0.130) (0.172) (0.129) (0.114) net issues (0.125) (0.046) (0.053) (0.072) (0.092) (0.130) (0.174) (0.129) (0.112) N 19, ,021 3,408 4,624 5,086 4,

12 3.2 The credit score regression We use a credit score regression (Altman (1968), Kaplan and Urwitz (1979), and de Jong, Verbeek, and Verwijmeren (2011)) and estimate firm-specific credit ratings as a function of the firm s debt ratio and other characteristics in an ordered logit regression. The results of the regression will later be used to define the debt capacity of a firm. The ordered logit regression to estimate credit ratings as a function of the debt ratio and other firm characteristics reads credit score it = αdr it + β 1 x it + β 2 z it + ε it rating it = j, if µ j 1 < credit score it < µ j, j = 1,..., 10. (1) The left hand variable credit score it is the unobserved latent variable and µ j, j = 1,..., 9, denote the estimated thresholds that separate the credit scores into the rating categories. We set µ 0 = and µ 10 = +. The debt ratio is denoted by dr it and x it is a vector of observable firm characteristics. Besides the debt ratio, Standard & Poor s (2008) name firm size, profitability, liquidity, age, characteristics (tangibility) of assets, and industry specific effects as the most important rating determinants. We measure these factors using the following proxy variables: firm size it = log(sales it ) (2) profitability it = ebitda it / assets it (3) liquidity ratio 1,it = working capital it / assets it (4) liquidity ratio 2,it = retained earnings it / assets it (5) tangibility it = property, plant & equipment it / assets it (6) The measurement of firm size, profitability, and tangibility correspond to the commonly used determinants in capital structure research (Rajan and Zingales (1995)) and the liquidity measures to the standard proxies in the rating literature Altman and Rijken (2004). Table 12

13 1 implicitly indicates the correlation of ratings and the explanatory variables. Highly rated firms are on average larger, more profitable, and have a higher portion of liquid assets. The fraction of tangible assets depends on the company s business fields and hence is not correlated with the rating categories. Since we are interested in a ceteris-paribus-analysis of debt ratios and credit ratings, but all five variables are correlated with the debt ratio, we need to substitute all five variables with their orthogonal values to the debt ratio. Orthogonal values of the variables correspond to the residuals of an univariate OLS-regression of the variable on the debt ratio and are by construction of the OLS-estimator uncorrelated with debt ratio. Mählmann (2011) finds the age of the rating to influence the rating agency s decision independent of other observable firm characteristics because of the companies ability to control their information flow to the rating agency. To capture this rating-age-specific effect, we include dummy variables for the age of the firm s rating, limited to a maximum of 10 and using the first year as base category. We use the Fama-French 38 industry classification to account for differences across industries that are not captured in the explanatory variables above. We include dummy variables for each of the 38 Fama-French industries, which reduce to 34 dummy variables due to the exclusion of financial firms and utilities and using one category as base case. Moreover, we include dummy variables for each observation year to capture time-specific effects such as the change of the rating agency s standards over time (Blume, Lim, and MacKinlay (1998)). All dummy variables are contained in the explanatory variable z it. Panel (a) of table 2 shows the results of the credit score regression for the two different specifications of the debt ratio using maximum likelihood estimation to determine all coefficients and parameters. Both specifications yield similar results and all explanatory variables are significant at the 1% level. A negative sign of a coefficient indicates that the variable decreases the latent credit score and hence increases the probability of being downgraded. Positive signs work in the opposite direction. The negative coefficient of the debt ratio shows 13

14 Table 2: Credit Score Regression. Panel (a) displays the results of the ordered logit regression (1). We use annual COMPUSTAT data from 1985 to 2010 of 2,573 different firms with a S&P long-term credit issuer rating excluding financial firms and utilities. The debt-to-capital ratio is defined as long-term plus short-term debt over the sum of long-term debt, short-term debt, and share price times numbers of shares outstanding. The liabilities-to-assets ratio equals book debt over market assets. Book debt is calculated as total assets minus book equity and book equity equals total assets less total liabilities and preferred stock plus deferred taxes and convertible debt. Size is measured as log(sales) and profitability is EBITDA over total assets. Liquidity ratio 1 is retained earnings scaled by total assets and liquidity ratio 2 is working capital scaled by total assets. Age dummies indicate the age of the rating with a maximum of 10 years. Year dummies are included for every year. Industry dummies are based on the Fama-French 38 industries which reduce to 33 categories due to the exclusion of financial firms and utilities. Robust standard errors are reported in parentheses and indicates significance at the 1% level. For each dummy variable category, we show the number of significant dummy variables at the 1% level. Panel (b) reports the fraction of the correctly specified ratings in percent. The left side denotes the difference between the observed ratings and the estimated ratings. For example 39.5% of the observed ratings are one category smaller or higher than the estimated rating. The last line reports the fraction of investment-grade firms that are estimated to have an investment-grade rating. Panel (a): Estimation results explanatory variables debt/capital liabilities/assets debt ratio (0.102) (0.098) size (0.016) (0.017) profitability (0.247) (0.255) liquidity ratio (0.104) (0.110) liquidity ratio (0.144) (0.160) tangibility (0.096) (0.100) 9 age dummies year dummies industry dummies observations 19,375 17,945 pseudo R Panel (b): Correctly specified ratings (in %) rating estimation error debt/capital liabilities/assets ± > ± Fraction of investment-grade firms with credit score > µ

15 that with an increasing amount of debt, the firm becomes more likely to lose its current rating. Similarly, large and profitable firms tend to have higher ratings. The liquidity ratio 1, defined as retained earnings scaled by total assets, increases the credit score, but in contrast, liquidity ratio 2, defined as working capital over total assets, decreases the credit score which corresponds to the findings of de Jong, Verbeek, and Verwijmeren (2011). The combined hypothesis that the sum of both coefficients is smaller zero can be rejected at the 1% level (not reported). All together, we find that liquidity has a positive effect on the firm s credit rating as predicted by theory. In both specifications, 8 out of 9 age dummy variables are significant and indicate that the age of the rating yields information about the rating agency s decision. 23 (22 in the liabilities-to-assets regression) out of the 25 year dummies capture time-specific effects, such as the change of the rating agency s standards. The significance of 30 (29) out of 33 industry dummies shows that there are different industry characteristics that rating agencies take into account and that are not captured in the other explanatory variables. In panel (b) of table 2, we report the precision of the estimation results. We use the debt ratio, firm characteristics, and dummy variables to estimate the firm s credit rating and then calculate the difference between the actual and the estimated rating. Using the debt-to-capital definition of the debt ratio, our results show that 55.5% of the firms are accurately classified into the rating categories depending on the definiton of the debt ratio. Another 39.5% are just one category off and only 5.1% deviate more than one category. We are able to classify 87.7% of the investment-grade firms correctly as firms with an estimated investment-grade rating. We find similar results using the liabilities-to-assets definition. Molina (2005) documents an endogenous relationship between the debt ratio and credit ratings because a sudden reduction in operating risk leads to an increase in creditworthiness but at the same time encourages the firm to take on more debt. Molina proposed instrumenting the debt ratio with the history of firms past market valuations (Baker and Wurgler (2002)) and the firms marginal tax rates (Graham and Mills (2008)). In our study, 15

16 the tenor of the results remains the same after introducing these instruments to control for endogeneity. As well, the sample size reduces drastically due to the need of various lags of the market-to-book ratio or missing marginal tax rates to construct the instruments. Hence, we continue our analysis without using instrumental variable regressions. As well, using an order probit regression instead of an ordered logit regression does not change the results. Our evidence show that the cross-sectional variation of credit ratings can partly be explained with the given variables. In particular, the debt ratio is of special importance because of its large significant coefficient. This finding motivates a definition of the debt capacity using the credit score regression. 3.3 Debt capacity estimates Using the functional form of the credit score regression (1), we will derive two measures of the debt capacity to answer the question of how much is too much. For each credit score it larger than µ j, the probability to be currently downgraded to a rating of j or smaller is given by the logit distribution: P(rating it j) = exp( µ j + αdr it + β 1 x it + β 2 z it ), j = 2,..., 9. (7) The first measure of debt capacity follows de Jong, Verbeek, and Verwijmeren (2011) and defines the debt capacity as the critical debt ratio where the probability of losing the investment-grade rating equals a given constant p. Solving equation (7) for dr it and setting j = 6 (BB (rating = 6) is the highest non-investment-grade rating) yields DC it = log(1/p 1) + µ 6 β 1 x it β 2 z it α. (8) The resulting value DC it is the firm s debt capacity. For every debt ratio larger than the debt capacity DC it the firm s probability of being downgraded to a non-investment-grade rating is larger than p. If the firm wants to avoid a high downgrading probability, it has to 16

17 keep its debt ratio below its debt capacity. The results of this debt capacity specification hold by definition for investment-grade firms only and all firms have the same target rating. We call this specification the investment-grade debt capacity. In our second specification of the debt capacity, we follow Kisgen (2009) and assume that all firms in the same rating class have a given minimum target rating. Investment-grade firms want to lose not more than one rating class but target to stay in the investment-grade category. In detail, AAA and AA target to maintain a rating of AA or A, respectively, while A and BBB rated firms target an rating of BBB. All other firms (ratings BB through C) want to maintain their current ratings. In this manner all investment-grade firms want to preserve an investment-grade rating and avoid a significant increase in cost of capital. All other firms want to avoid being downgraded further and to stay close to the investment-grade region. For each rating category j we denote the target rating as k j. From equation (7) we receive DC it = log(1/p 1) + µ k j β 1 x it β 2 z it α, j = 2,..., 10. (9) This specification is more general than (8), because the firm s target rating is always close to the current rating. It holds for all firms that have a credit rating while the upper specification holds for investment-grade firms only. The investment-grade specification becomes a special case of this one if one chooses BBB as target rating for all firms. We call this specification the target rating debt capacity. Since both specifications of the debt capacity are given in terms of the debt ratio, we need to limit the value for DC to the interval [0, 1]. Estimates of DC that exceed the boundaries are replaced with 1, or 0 respectively. All debt capacity estimates are functions of the exogenous probability p. A variation in p changes the size of the debt capacity, but it never changes the cross sectional order of debt capacities. For example, two firms have the same debt ratio, but firm A has a higher credit score than firm B, then independent of the choice of p, firm A will have a higher debt capacity. In our results, we use p = 0.30 which 17

18 sets the debt capacity to the debt ratio where the probability of losing the target rating equals 30%. Table 3 reports the results of the debt capacity estimation. We calculate the debt capacity in four different settings. We have two specifications of the debt capacity and for each specification we use both definitions of the debt ratio. The results on the investment-grade debt capacity are shown in panel (a) and panel (b) reports the target rating debt capacity. We find the average debt capacity to be while the average debt ratio is only defined as debt-to-capital. Firms operate on average below their debt capacity and could take on additional debt. The difference between the debt capacity and the debt ratio is statistically and economically significant. The debt capacity of AAA rated firms is the largest (0.592) and decreases in ratings. BBB rated firms are just a little above their debt capacity and can only afford to take on another 2% debt. Statistically, this difference is significant, but its economical size is only marginal. The standard deviation of the debt capacity in all subgroups is similar indicating that only the mean changes but not the dispersion of the estimates. The results are similar using the assets-to-liabilities definition of the debt ratio. The debt capacity is a little higher (0.522) because more capital is defined as debt as indicated by the average debt ratio. If the financial target of the firm is to maintain a BBB rating, then firms in the upper rating categories can take on more debt than firms close to the BBB threshold. Firm s liquidity, profitability, size, and asset characteristics allow them to carry more debt than firms with lower ratings. In the target rating specification of the debt capacity, the average debt capacity in the debt-to-capital definition is while the average debt ratio is On average, firms can take up another 13% of debt. Both figures are higher than in the first specification, because the sample includes highly levered firms in the speculative-grade segment. Taking a look a the debt capacity across different ratings, we observe that AAA firms face the hardest financing restrictions. On average only 8.6% of their assets can be financed with debt if they want their probability of losing the rating to stay below p = 30%. Eventhough 18

19 Table 3: Debt capacity estimates. The table shows the results for the two different debt capacity specifications and for each specification two definitions of the debt ratio. We report mean and standard deviation of the debt capacity and the mean of the debt ratio for each subgroup. The investment-grade debt capacity corresponds to equation (8) and the target rating debt capacity to equation (9). p is set to 0.3. We use COMPUSTAT data from 1985 to 2010 of 2,573 different non-financial and non-utility firms with S&P long-term credit issuer ratings. The ratio dr defined as debt-to-capital corresponds to long- and short-term debt scaled by the numerator plus equity. The liabilities-to-assets definition of the debt ratio equals book debt over market assets. Book debt is calculated as total assets minus book equity and book equity equals total assets less total liabilities and preferred stock plus deferred taxes and convertible debt. The significance indicates a t-test of mean(dc dr) < 0.,, correspond to a significances at the 1%, 5%, or 10% level, respectively. debt/capital Panel (a): Investment-grade debt capacity assets/liabilities rating DC std(dc) dr DC std(dc) dr AAA BBB AAA AA A BBB debt/capital Panel (b): Target rating debt capacity assets/liabilities rating DC std(dc) dr DC std(dc) dr AAA C AAA AA A BBB BB B CCC CC C

20 AAA firms have the largest fraction of liquid assets, are bigger and more profitable than the other firms, they cannot take on more debt on average. Having the second best firm characteristics, AA firms can afford a debt ratio of before exceeding their debt capacity. Intuitively, speculative-grade firms can use more debt than investment-grade rated firms. BB firms have a debt capacity of and firms with high default risk can finance more than 80% of their assets with debt. The standard deviation of the debt capacity across different ratings increases with declining ratings. This finding supports that financing constraints are stronger for highly rated firms. Defining the debt ratio as liabilities-to-assets results in similar findings. Again, the debt capacity is on average higher due to the definition of the debt ratio, and the debt capacity and its standard deviation increase with declining ratings. The cross-sectional variation of the debt capacity within each rating category can be inferred from the regression results in table 2. Size and profitability increase the amount of debt a firm can take on, as well as the tangibility of assets and firms liquidity. The significance of the year and industry dummies indicate that debt capacities change over time and across industries. All these explanatory factors are incorporated into the debt capacity estimates in equation (9). Our results emphasize the importance of the debt capacity for financing decisions and its heterogeneity across different rating categories. If firms target maintaining an investmentgrade rating, then highly rated firms have more choices for financing decisions, but if firms target a minimum rating close to their current rating, then highly rated firms face the tightest financing constraints. 3.4 Measuring financial flexibility A central question in our study is if financing decisions are driven by concerns of financial flexibility. We will measure financial flexibility as the difference between the firm s debt capacity and its debt ratio before or after financing, called debt buffer. The debt buffer indicates how much additional debt the firm can issue. If the debt buffer is large, then a 20

21 firm can use debt to cover external financial needs, while a firm with a small debt buffer has to use equity to maintain its target rating. If the debt buffer is negative, then the debt ratio already exceeds the debt capacity and the probability for the firm to lose the target rating exceeds the tolerated probability. If the firm reduces its debt ratio by issuing equity or repurchasing debt it can restore its financial flexibility. For the ease of notation, we suppress the argument i in the following equations (x t = x it ). To gain insight into firms motives of financing decisions, we examine the firm s debt buffer in three steps. First, we take a look at debt buffer after the investment decision but before the financing activity. This debt buffer DB before,t is defined as the debt capacity less debt in t plus the net debt issues less net equity issues on the same period scaled by assets in t. Let debt t denote the amount of debt that corresponds to the debt capacity and debt t the amount of debt in the debt ratio dr t. We can write the debt buffer before financing as DB before,t = debt t debt t equity t + debt t assets t = DC t dr t equity t debt t assets t. (10) This debt buffer indicates how much debt the firm can issue to finance the planned investments. The size and characteristics of the new investment are included in the debt capacity and in the assets because they are measured in t. The amount of debt at the end of fiscal year t is adjusted for capital issues to show firm s debt excluding the changes in external capital. A second informative figure is the debt buffer that results if the firm has financed all its capital needs exclusively with debt. To measure the amount of capital that firms have to issue, we follow Frank and Goyal (2003) and calculate the financing deficit def t as the amount of capital that a firm needs to invest I t, pay out in dividends DIV t, and in- or 21

22 decrease working capital W t less the amount of cash that is available C t. def t = DIV t + I t + W t C t, (11) If def t is positive, then the firm needs to raise external capital: either debt, equity, or a combination of both. If the deficit is negative, then the firm has a financial surplus and can repurchase debt and/or equity 1. The debt buffer that results from using debt to settle the deficit is defined as DB debt,t = DB before,t def t assets t. (12) This ratio indicates whether the firm has enough unused debt capacities to settle the need for external capital. If DB debt,t is close to zero, then the firm would face financial constraints after using debt because is will not be able to fund future investments with debt. If DB debt,t is negative, then the firm would exceed its debt capacity and might lose the target rating. Hence, the firm should rely solely on debt if DB debt,t is positive. In a third analysis of the debt buffer, we examine the debt buffer after investments and after financing activity. This ratio DB after,t corresponds to the debt buffer before financing activity less the change in debt plus the change in equity. DB after,t = DB before,t debt t + equity t assets t = DC t dr t (13) Adding the change in capital to the debt buffer before financing activity results in the difference between the debt capacity and the firm s debt ratio. This debt buffer incorporates all operating and financing activity of the firm in year t. All three versions of the debt buffer 1 I t, DIV t, W t, and C t are calculated using data from firms statements of cash flows. Summary statistics of the financial deficit def are reported in table 1. For a detailed description and definition of all variables see Frank and Goyal (2003). 22

23 serve as a management information criteria that can be used to make decision on capital issues and repurchases. In order to analyze empirical values of the debt buffer, we need to distinguish between capital issues, reductions, and substitutions, and differentiate each of these categories further into debt, equity, or dual financing activity. We use the firms statements of cash flows to classify the financing activities. Net debt issues are defined as issues less reduction of longterm debt and net equity issues as sale less repurchase of common or preferred stock. Net capital issues are the sum of net debt issues and net equity issues. All values are scaled by total assets. Since firms usually issue or repurchase at least a small amount of both securities each fiscal year, we need to identify debt financing activities, equity financing activities, and dual financing activities more precisely. Similar to Korajczyk and Levy (2003) and Hovakimian (2006), debt financing activities are observations where the absolute net debt issues are larger than 5% of the asset value, the absolute net equity issues are smaller than 5% and the absolute sum of both is larger than 5%. Equity financing activities are defined analogously. Firms that use dual issues, issue more than 5% of net debt and of net equity in absolute terms and the sum is larger or smaller than 5% or -5%, respectively. As well, dual issues correspond to cases when firms issue less than 5% of each security but in sum more than 5%. Negative financing activities correspond to capital repurchases while positive financing activities are capital issues. We call an observation a debt-to-equity substitution if net debt issues are smaller than -5% but net equity issues exceed +5% and the absolute value of the sum is smaller than 5%. The opposing case is considered a equity-to-debt substitution. Capital issues always correspond to a financing deficit while capital repurchases correspond to a financial surplus (negative deficit). We can identify 8,652 firm-year observations that either are classified as debt, equity, or dual financing or capital substitutions. 3,423 of them stem from investment-grade firms. A large fraction of the data set is excluded because their capital issues and repurchases are smaller than 5% of their asset value. We do not report our results for each single rating category because the number of observations is 23

24 too small in several categories. Since we employ rating specific targets and control for firm characteristics, we pool all observations into one sample and split the sample up according to the issuance and repurchase type. If we would use the change in the debt ratio instead of the capital issues from cash flow statement to identify financing activities, then we would not be able to separate debt issues from equity repurchases and capital substitutions. If firms target to preserve financial flexibility, then we expect firms to issue debt if their debt buffer before financing DB before,t is high and to issue equity if it is low. As a results the debt buffer after financing should not be non-negative for all issuance types. If dual or equity issuing firms would use debt only, then we expect them to have a debt buffer that is negative or close to zero. Hence, these firms choose to issue equity. Firms that have a financial surplus are expected to repurchase debt if their debt buffer DB before,t is low and to repurchase equity if it is high. Firms that substitute debt with equity are expected to have a lower debt buffer DB before,t than firms that substitute equity with debt. 24