Credit default swaps, financing decisions and firm bankruptcies

(1)

Credit default swaps, financing decisions and firm

bankruptcies

Jiaying Chen 10563571

University of Amsterdam Amsterdam Business School

July 6, 2014

Abstract

There is ongoing debate about the costs and benefits of Credit default swaps (CDSs), and it became more drastic in the wake of the financial crisis. Using a dataset of U.S nonfinancial public firms, this paper explores whether the use of CDSs is associated with changes in the probability of bankruptcy of firms with firm leverage and debt maturity decisions playing connective roles in between. Employing a two-step model, it fails to find a link between the onset of CDS trading and bankruptcy rate through corporate financing decisions. Nevertheless, after addressing the endogeneity problem by constructing an instrument of excess CDS exposure, significant results are observed. On average, increase in excess CDS exposure is negatively related with bankruptcy rate with firm leverage acting in between. However, the respective impact of CDSs on healthy firms and distressed firms is contrary to each other. It is found that an increase in CDSs leads to a decrease in probability of bankruptcy through firm leverage for healthy firms while it increases the bankruptcy probability of distressed firms through both firm leverage and debt maturity.

Keywords: CDS, bankruptcy, firm leverage, debt maturity, healthy firms, distressed firms.

MSc Business Economics, Finance track Master Thesis

(2)

1. Rationale and Aim of the Research

In the early 1990s, J. P. Morgan & Co. bankers devised credit default swaps (CDSs), which protects lenders in the way of transferring the default risk to a third party. The CDS market experienced a dramatic rise in the years before the 2008-2009 financial crisis but then shrank rapidly. In the wake of the financial crisis, quite a lot debates about whether CDSs contributed to the crisis arose. According to the International Swaps and Derivatives Association (ISDA), total CDS notional amounts outstanding reached a peak of $62 trillion in 2007, but as soon as the 2008-2009 crisis burst out, CDS market shrank considerably to $25 trillion by the end of 2012. It seems firms realized the negative effect of CDSs and decrease CDS trading during and after the crisis. Matta (2011) indicates that CDSs might play a prominent role in the bankruptcy of Lehman Brothers during the crisis. Does CDSs played an absolute negative role during the depression period or it also had some positive contributions?

As a contract protects a buyer against the cost of default of the reference entity, CDSs mitigate concentrations of credit risk, and thus reduce the transaction costs of insurance or risk transfer, at the same time, CDSs help increase borrowers’ debt capability, that means more positive present value projects can receive financing or can be better financed (Bolton and Oehmke, 2011). Also, CDSs promote diversification outside the banking system and enhance trading liquidity. Except for those benefits, some argue that credit derivatives are attached with disadvantages. Hu and Black (2008) argue that the presence of CDSs gives rise to empty creditor problem, that is, investors protected by CDSs have less incentive to monitor debtors, and investors that have negative ownership even have incentives to cause the firm’s value to fall. This problem will be even severe if lenders “overinsure” (Matta, 2011), that is, the privately optimal level of creditor protection excess the social optimum, this over-insurance problem gives rise to an high incidence of bankruptcy (Bolton and Oehmke, 2011). Matta (2011) states that CDS contracts may alter the dynamics of corporate financing in the sense that optimal lending decisions are affected by expected distress outcomes. Therefore, CDS protection benefits firms that are in distress and not in distress differently, particularly in terms of firm financing. Research done by Hirtle (2010) finds that only for large firms, greater use of credit derivatives is associated with banks’ improved credit supply in terms of longer loan maturity and lower spreads. For smaller term borrowers, both new loan volume and maturity decrease as credit protection rises. Consistent with Hirtle (2010), Ashcraft and Santos (2009) state that although firms do not experience improved lending terms on average, the introduction of CDSs leads to an improvement in borrowing terms for

(4)

safe and transparent firms. Different from the two articles mentioned above, Saretto and Tookes (2011) argues that, on average, firms with traded CDS contracts on their debt are able to maintain higher leverage ratios and longer debt maturities.

According to the conflicts in the previous literature, several questions may arise. How does the presence of CDS trading affect the bankruptcy rate of firms? Does the introduction of CDSs affect firm leverage and debt maturity of healthy firms and distressed firms adversely? And how do the changes in firm leverage and debt maturity subsequently influence the probability of bankruptcy of firms? This paper addresses the question that whether CDSs affect firm financing differently for firms in distress and firms not in distress.

There is a growing literature examining the effects of CDSs on credit market outcomes, and the impact of CDSs on banks’ incentives to monitor, or on the ability of CDSs to improve risk sharing, but there is no specific empirical research investigates how CDSs affect firm financing and subsequently contribute to the financial crisis, which is measured as the bankruptcy probability of firms. Thus the goal of this paper is to empirically identify the impact of CDSs on financial crisis by analyzing the role of firm financing decisions play in the relationship between CDSs and the probability of bankruptcy of firms. The research question is: Anticipating firms’ leverage and debt maturity decisions, how does CDSs affect the probability of bankruptcy of firms in distress and not in distress?

This paper extends the previous literature by empirically investigating the relationship between credit default swaps and firm bankruptcies with corporate financing decisions playing a connective role between, and examining how the relationship differ from healthy firms to distressed firms. Using a dataset of 3,905 U.S. nonfinancial public firms during the period from 2008 to 2012, a two-step model is constructed. The first step is to examine how the firm leverage and debt maturity change as a firm introduces CDS protection; the second step is to test the impact of firm leverage and debt maturity on the probability of bankruptcy of firms. The full sample is divided into healthy firms and distressed firms according to the Altman Z score.

Initially, the effect of CDSs is evaluated by using a binary indicator that identifies firms with traded CDSs on their debt. When testing the relationship between firm financing and CDS trading, no evidence is found corresponding to the hypotheses that the presence of CDS trading increases firm leverage and debt maturity of healthy firms, but decreases leverage and maturity of distress firms. Nevertheless, significant negative relationship between firm leverage and bankrutpcy probability is found for distressed firms and the full sample, besides, it is observed that a rise in short-term debt proportion significantly increases the bankrutcy

(5)

rate on average. In sum, the empirical analysis reveals no significant link between CDSs and the probaility of bankruptcy via firm leverage or debt maturity. The potential concern for the insignificant results is the the reverse causality between CDS trading and firm financing decisions, which is referred as endogeneity problem.

To address the reverse caulsality problems, a two-stage approach is developed to estimate an orthogonal instrument called excess CDS exposure. By using this instrument, strongly significant correlations of CDSs with firm leverage and debt maturity are observed. For both healthy firms and distressed firms, firm leverage increase significantly as firms obtain additional CDS protection. The likelihood of bankruptcy decreases as firm leverage increases for healthy firms, while for distressed firms the bankruptcy rate increases with the increases in leverage. When examining debt maturity, it is only found that the introduction of CDSs increases the proportion of short-term debt for distressed firms, which means it shortens the debt maturity of distressed firms. Shoter debt maturity then leads to an increase in bankrutpcy rate. For healthy firms, positive realtionship between the CDS trading and the proprotion of short-term debt is observed, however, the short-term debt propotion does not significantly affect the bankrutpcy rate of healthy firms. In sum, increases in excess CDS exposure result in decreases in probability of bankruptcy through firm leverage for healthy firms but double increases the bankruptcy probability of distressed firms through both firm leverage and debt maturity. Even though CDS trading has inverse impacts on distressed firms and healthy firms, when testing the full sample, it shows increases in CDS exposure lead to a decline in the probability of bankruptcy with firm leverage acting in between. These results are both statistically and economically significant.

The reminder of the paper is organized as follows. The next section reviews previous literature on advantages and disadvantages of CDSs, and the relationship between CDSs and financing decisions. Section 3 develops testable hypotheses in relation to the prior literature, and presents the empirical specification used in the analysis. The dataset is described in Section 4. Section 5 presents the key results with a detailed examination of the endogeneity concerns. Section 6 concludes.

(6)

2. Literature Review

This paper is related to a growing literature examining the benefits and costs of CDSs and the literature focusing on the impact of CDSs on corporate financing.

2.1 Effect of CDSs

A credit default swap is an insurance contract against the loss of bond’s principal in the case of a credit event (Chernov, Gorbenko, Makarov, 2013). In a CDS, the CDS buyer should pay a fixed periodic premium fee to the CDS seller for a certain period of time, in exchange, the CDS seller promises to make compensation to the CDS buyer in the case of a credit event on a pre-specified reference asset. The credit event can refer to non-payment of the debtor, the bankruptcy filing of the debtor, debt restructuring or a credit-rating downgrade (Bolton and Oehmke, 2011). Generally, the default payment is determined as the difference between the face value of the loan and the market price of a specified debt obligation. A CDS contract has obvious similarities with a regular insurance, however, as stated by Stulz (2010), there are two key differences between a CDS and a regular insurance policy. First, CDSs are not as well regulated as regular insurance policies, people can buy a credit default swap without holding the bonds, in this case, the amount you insure with a CDS is called the notional amount. Second, unlike regular insurance contracts, CDSs are bilateral contracts that are usually traded over-the-counter, they can be resold to another party.

As proposed in the paper by Bolton and Oehmke (2011), CDSs can reduce the cost of corporate debt contracts arising from the possibility of firms’ strategic default on their debt. Strategic default refers to the situation that a firm defaults because managers want to divert cash to themselves (Bolton and Scharfstein, 1996). Due to the incompleteness of corporate debt contracts and the absence of default penalties, firms do not credibly commit to repay their debt, they can always choose to default and divert cash flows to themselves even though their cash flows are sufficient to repay the debt. The possibility of strategic default makes lenders reluctant to lend and thus impose extra cost on debt financing, in this case, firms’ capacity to raise debt capital is reduced. The presence of CDS improves the contacting technology and alleviates the limited commitment problem caused by incomplete debt contracts. Bolton and Oehmke (2011) show that CDSs reduce the incidence of strategic default and improve efficiency by strengthening creditors’ bargaining power in case of debt renegotiation upon strategic default and increasing the debtor’s pledgeable income. With the protection of CDSs, creditors tend to loss less when firms default, and the better bargaining

(7)

position enables creditors to require more in debt renegotiation, therefore, firms are less willing to strategically renegotiate down their debt payments to their own advantage. Also, acting as a commitment device in renegotiations, CDS protections increase firms’ debt capacity and enable more positive net present value projects to receive financing or to be better financed ex-ante.

There are some other researchers stating that CDSs play an important role in the debtor-creditor relationship, they facilitate efficient risk sharing (Morrison, 2014), improve diversification and trading liquidity (Peristiani and Savino, 2011), increase lending efficiency for high-quality borrowers (Parlour and Winton, 2012), and those benefits can further lead to increase in access to credit and decrease in funding costs (Arping, 2012). Arping (2012) argues that CDS protection insulates lenders against losses from default, it strengthens lenders position in debt renegotiations and rewards lenders for letting renegotiation fail and forcing borrowers into bankruptcy and liquidation. Faced with a more severe threat of foreclosure, borrowers may be inclined to exert higher effort and take less opportunistic actions in order to make it less worthwhile for their lenders to exercise the termination threats. In this way, CDS protection is working as an instrument that induces borrowers’ incentives to work harder.

From the perspective of creditors, the introduction of CDSs also can be beneficial. Duffee and Zhou (2001) investigate the effect of the introduction of CDSs on banks, which usually refers to creditors. They demonstrate that banks can transfer the credit risk of their loans to outsiders by using a CDS contract, and CDSs are more flexible in transferring risks than some other mechanisms, such as loan sales without resources, they make it easier to mitigate adverse selection problem in credit-risk transfer markets. Allen and Carletti (2006) are in line with this opinion indicating that depending on the nature of liquidity shocks credit risk transfer is beneficial.

Nevertheless, despite the benefits mentioned above, CDSs can also lead to inefficiencies. Risk sharing may reduce creditors’ incentive to monitor their investment project. What’s worse, while the ex-ante commitment benefits that creditors can get with the socially optimal choice of credit protection cannot offset the costs of inefficient renegotiation, creditors tend to over-insure, which gives rise to the empty creditor problem. Empty creditors (i.e. holders of debt and CDSs who have no interest in the continuation of the debtor) are unwilling to renegotiate with debtors in order to obtain compensation on their CDS positions even though renegotiation would be socially efficient. This CDS over-insurance is apparently inefficient, and in this sense, the introduction of CDSs will induce the increase of severe creditors and

(8)

lead to high incidence of bankruptcy of the reference entities (Bolton and Oehmke, 2011). Hu and Black (2008) focus on the detrimental effect of empty creditor problem on efficient debt restructuring, they argue that the empty creditor problem can negatively affect the probability of bankruptcy of borrowers in two ways. First, it may rigidify the interactions between the creditor and the borrower when negotiating or renegotiating debt structure. Second, it may decrease creditors’ incentive to assess and monitor the borrowers’ repayment ability (Subrahmanyam, Tang and Wang, 2011). Besides, some articles have argued that CDS trading could affect the debtor-creditor relationship in a negative way, and thus be potentially deleterious to borrowers (e.g., Duffee and Zhou (2001), Arping (2012) and Morrison (2014)). Morrison (2014) find that CDSs can cause a wealth destruction by disintermediation and reduction in banks’ monitoring incentives, this could incur the situation that entrepreneurs inefficiently substitute bank loans or mixed finance by junk bond financing and reduce the quality of their projects. Arping (2012) states that in extreme cases, the introduction of CDSs could lead to a breakdown of the credit market. CDS protection enables lenders to abuse their enhanced bargaining power and extract additional surplus even if the continuation of the project is more efficient. This ex-post hold up threat could eliminate borrowers’ incentive to exert effort to a large extent. Ultimately, the presence of CDS protection may tighten credit constraints and be detrimental to welfare.

In a larger scope, CDSs could have negative effect on the market stability. The recent financial crisis has revealed several shortcomings of CDS, regulators started to concern about this. CDSs are traded in over-the-counter market, where the transparency is lacking and liquidity and efficiency are low. As Terzi and Ulucay (2011) argue, the opaque linkages within the over-the-counter CDS market have led to the situation where market participants have become too big or interconnected to fail. When the CDS sellers are inappropriately financed, it could lead to contagion and even worse, systematic risk.Allen and Carletti (2006) show the same viewpoint that credit risk transfer can lead to contagion and contribute to financial crisis. Duffee and Zhou (2001) indicate that CDSs can lead to the breakdown of other risk-transferring mechanisms, which is detrimental to welfare.

This paper is part of the growing literature on CDSs and their effect on the probability of bankruptcy of firms. Several existing literature have already given a sight into this effect. Peristiani and Savino (2011) investigate the question whether firms with CDS positions on their debt are more likely to fall into bankruptcy in the period 2001-08. But they did not find a significant causality for the entire period. When test the single year 2008, they did find the result that firms with CDSs have a higher bankruptcy probability. The possible explanation

(9)

for this relationship can be the fire-sale spiral theories. Firms get higher pressure in the fire sales of their financial assets because of the financial crisis. Different from Peristiani and Savino (2011), Subrahmanyam, Tang and Wang (2011) find a strong relationship between probability of bankruptcy of reference firms and CDS trading. The bankruptcy probability increased more than one time after the introduction of CDSs for average firms. They also state that in financial distress, firms with CDSs on their debt are more likely to file for bankruptcy than firms without CDSs. This paper contributes to the existing literature by emphasizing the effect of CDSs on the bankruptcy risks of reference firms with the interpose of firm financing decisions. One prediction in this paper is that the CDS trading increases the bankruptcy probability of firms in distress, but decreases the bankruptcy probability of firms not in distress. Firm financing (i.e., firm leverage and debt maturity), as a role playing in between, can be one of the factors that lead to the adverse effect.

2.2 CDSs and firms’ financing decisions

There are several explanations on how CDSs affect firms’ financing. From the aspect of capital supplier, the presence of CDSs provides lenders the ability to hedge the credit risk, and hence reduce the frictions on the supply side. According to Saretto and Tookes (2011), there are mainly four channels through which CDSs can affect firms’ capital structures. First, when there is a market segmentation, lenders are separated into who are willing to keep the credit risk and who would like to lend the capital they have, lenders can achieve the opportunity to reduce regulatory capital requirements by holding single-name, which can then increase the supply of credit to firms. Second, banks and some other financial institutions need to provide loans in order to maintain their relationship with important clients, the presence of CDSs diminishes the portfolio risk that lenders are faced with. Third, the existence of CDS markets can make it more appealing for those potential investors to obtain corporate debt when treasuries are in short supply. Last but not least, it is expected that even if lenders do not purchase CDSs at the very beginning, the presence of CDS markets still could have influence on firms’ financing. The CDS markets can provide a liquid resale option and make it more attractive to hold credit risk.

In their study of the relationship between CDSs and firms’ financing, where firm leverage and debt maturity are taken as measures for firm financing, Saretto and Tookes (2011) investigate in how the ability of lenders to hedge credit risk using CDSs influence firms’ capital structure and support the viewpoint that CDSs can relax firms’ capital supply

(10)

constraints. In addition to leverage, observing the change in debt maturity allows Saretto and Tookes (2011) to capture the variation in the firms’ financial contracts with the introduction of CDSs, which enhances lenders’ ability to hedge the credit risk. If keep everything else equal, shortening debt maturity can help to reduce lender’s risk, and keep borrowers under close control, thus lengthening debt maturity without asking for higher rates is considered as a relaxation of supply. Saretto and Tookes (2011) find that CDS markets allow firms to achieve higher firm leverage and longer debt maturity even during the 2008-09 crisis. They also find that the influence of CDSs on firms’ financing is stronger when credit constraints are binding, indicating that the impact of capital supply on firm capital structure varies with time. Considering the results, the arisen question is if increases in firm leverage and debt maturity are good or bad for firms. Recent evidence from the 2008–2009 financial crisis suggests thatreduction in credit constraints and extension in debt maturities can improve real investment (e.g., Duchin, Ozbas, and Sensoy (2010), Ivashina and Scharfstein (2010), Almeida et al. (2012)). Bolton and Oehmke (2011) also show that CDS protection, from an ex-ante perspective, increases banks’ bargaining power and thus increase the efficiency of debt financing.

Many have claimed that CDSs are instruments that contribute to a reduction in the cost of debt financing to firms from two aspects, one is the introduction of CDSs brings about more hedging opportunities, and the other is CDSs’ price is a potentially important source of new information on firms. Considering those opinions, Ashcraft and Santos (2009) examine how the cost of debt for the average firm in the corporate bond and syndicated loan markets is affected by CDS trading in the US. However, they find no evidence that CDS trading lowers the cost of debt financing for the average borrower, contrast to the previous opinions they find that the debt financing of risky firms and informationally opaque firms even increased with the introduction of CDSs. But they also find that safe firms and informationally transparent firms with CDSs on their debt have experienced a small reduction in the spreads they paid to borrow from suppliers of capital. One possible explanation for this adverse effect of the CDS market on the cost of corporate debt could be that lenders that are less exposed to credit risks with CDS protection become less willing to monitor borrowers, therefore, when it comes to riskier and informationally opaque firms for which monitoring is necessary, lenders tend to require more compensation to offer or extend loans.

From the perspective of lenders, Hirtle (2009) examines the existence and size of the impact of banks’ use of credit derivatives on bank credit supply and loan spreads in US. He finds very limited evidence supporting the statement that CDS trading induces banks to

(11)

increase credit supply, especially for commitment borrowers (no matter large or small) and small term borrowers, the presence of CDSs even leads to a reduction in the volume of new loans they receive. On the contrary, large term borrowers tend to receive more supply of bank credit as banks increase CDS protection. The same results apply to loan maturity and spreads. For commitment borrowers and small term borrowers, average maturity either decreases or stay the same as CDSs increases, while for large term borrowers, spreads fall and average maturity either keeps unchanged (at banks that do not hedge actively) or increases (at active hedging banks). In sum, the increase in CDSs is more likely to have positive impact on large term borrowers but negative impact on commitment borrowers or small term borrowers. Campello and Matta (2012) investigate the link between the state of the economy and the demand for CDS and the influence of CDSs on the supply of credit over the business cycle. They propose three major implications in their research, first, safer firms with projects of positive NPV and higher continuation values can get more benefits from CDSs; second, firms’ financing still can benefit from CDS protection even in harsh time when aggregate credit supply is in shortage; third, CDS overinsurance is procyclical, which means if a firm is in distress, CDS overinsurance is more likely to occur during economic upturns, and during economic downturns, CDS overinsurance may act as an instrument to push the firm into bankruptcy .

Although Saretto and Tookes (2011) hold different opinion about the impact of CDSs on firms’ financing of smaller borrowers from the later three articles, it can be concluded that the above literature all state that CDSs can have positive effect on larger and informationally transparent firms in terms of reduction in debt cost and increase in the size of credit supply.

As an extension to those literatures, this paper is examining the relationship between CDS contracts and firm financing and how does this relationship affect firms’ probability of bankruptcy during and after the financial crisis. The relationship will be tested separately on firms in distress and firms not in distress. A more recent data sample (2008-2012) is adopted in this paper to examine the response of probability of bankruptcy of firms to the presence of CDSs. Consistent with the previous literature, one hypothesis in this paper is that firms that are not in distress will have higher leverage ratio if they have CDSs on their debt; another hypothesis is that firms that are in distress are more likely to fall into bankruptcy with CDSs on their debt.

(12)

3. Methodology and Hypotheses

3.1 Hypotheses

Hypothesis 1: CDSs decrease firm leverage and debt maturity of distressed borrowers,

which subsequently lead to an increase in the probability of bankruptcy.

Ashcraft and Santos (2009) find that risky and informationally opaque firms have been negatively affected by CDSs in terms of the cost of debt financing. Hirtle (2009) shows that the presence of CDSs has a negative effect for small term borrowers, the volume of new loans decrease and average maturity either decreases or is unchanged.

Bolton and Oehmke (2011) indicate that lenders tend to over-insure, which give rise to a high incidence of bankruptcy. In their model, banks’ bargaining power in debt renegotiations increases with the introduction of CDS protection, and at the same time, borrower strategic defaults decrease. Therefore, creditors are more willing to lend. Whereas, for those tough creditors, some of whom maybe empty creditors, they may have incentives to force distressed borrowers into bankruptcy if they are over-insured by holding CDS protection with an amount larger than the maximum amount they can receive in restructuring. Campello and Matta (2012) build a model to demonstrate that CDS overinsurance is procyclical, CDS overinsurance is more likely to occur during economic upturns, but during economic downturns, CDS overinsurance can act as an instrument to push the firm into bankruptcy. Che and Sethi (2012) argue that CDSs can lead to higher bankruptcy probability from another perspective. With the existence of CDSs, creditors can choose to sell CDS contract instead of offering loans to borrowers, this leads to a decline in credit supply and make it more difficult for distressed firms to roll over their debt. Hence, for financially distressed firms, both empty creditors and rollover risk concerns would deteriorate their credit risk, and push them into bankruptcy.

Hypothesis 2: CDSs increase firm leverage and debt maturity of borrowers that are not in

distress and subsequently decrease the probability of bankruptcy.

Hirtle (2009) finds the supply of bank loans to large corporate borrowers increases with CDS trading, thus the benefits of CDS protection could be captured by relatively large firms. Aschcraft and Santos (2009) document the differential impact of CDS trading on the cost of debt across firms’ riskiness and show that safe and transparent firms experience a small decrease in their cost of debt. Matta (2011) indicates that when a firm’s distress degree is sufficiently low, the positive effect of CDSs (i.e., increase the borrower’s debt capacity and

(13)

decrease debt cost) may exceed the detrimental effect of CDSs (i.e., tough creditors have incentives to force firms into bankruptcy), thus for healthy firms the presence of CDSs brings about surplus.

3.2 Methodology

According to Hirtle (2009) and Aschcraft and Santos (2009), the benefits of a CDS contract are narrow, in the sense that they accrue mainly to large and safe companies rather than to small and risky companies. The sample of the firms in this paper is divided into two groups, one includes firms in distress – distressed firms, while the other one refers to firms not in distress, that is, healthy firms. To determine if a firm is in distress or not, the Z score model built by Altman (1968) is used. Altman Z score model is an easy-to-calculate method wildly used in academic studies to predict the probabilitythat a firm will go into bankruptcy within two years. The model uses five business ratios weighted by discriminant coefficients to access a firm’s bankruptcy potential, which is shown as follows:

z-score = 1.2X1 + 1.4X2 + 3.3X3 + 0.6X4 + 0.999X5 (1)

Where, X1 is Working Capital/Total Assets ratio, which is a measure of the net liquid assets of the firm relative to the total capitalization. X2 represents Retained Earnings/Total Assets, which measures the cumulative profitability of a firm over time. X3 is Earnings Before Interest and Taxes/Total Assets that reveals the true productivity of a firm’s assets. X4 is Market Value of Equity/Book Value of Total Debt, it shows the amount a firm’s assets can decline before the firm becomes insolvent. X5 stands for Sale/Total Assets that measures a firm’s management’s capability in competitive conditions. Altman (1968) indicates that firms with a Z score below 1.81 are all bankrupt, thus should be considered as distressed firms while firms with a Z greater than 2.99 are considered as firms that will not fall into bankruptcy. The area between 1.81 and 2.99 is defined as “grey area”, and it’s hard to classify firms in the “grey area”. However, Altman (1968) still mentions that firms that have Z scores between 1.81 and 2.67 are classified as bankrupts even though they may not bankrupt for sure. Thus, in this paper, firms with Z scores lower than 2.67 are classified as

distressed firms and firms with Z scores higher than 2.67 are classified as healthy firms.

To investigate how a CDS contract affects a firms’ probability of bankruptcy through its relationship with firm financing (i.e., firm leverage and debt maturity), a two-step model is built. The first step is to find the impact of CDSs on firm leverage and debt maturity, two

(14)

OLS models are built as below (regression (2) &(3)). The design of dependent variables and major independent variables of the two regressions are from model of Saretto and Tookes (2011) while the control variables are based on the model of Johnson (2003), the predicted signs for control variables are also based on Johnson (2003). The second step is to test the influence of firm leverage and debt maturity on the probability of firms that go bankrupt with the presence of CDSs. An OLS regression is built to test the relationship of bankruptcy with CDS trading and firm financing. Control variables, as well as the expected findings of the control variables, of this regression are derived from Peristiani and Savino (2011). More formally, the models are

FLit = α0 + α1CDSQit + α2CDSDi + α3DMit + α4Xit (2)

DMit = β0 + β1CDSQit + β2CDSDi + β3FLit + β4Yit (3)

Bankruptcyit= γ0 + γ1FLit + γ2DMit + γ3CDSQit + γ4CDSDi + γ5Zit (4) The three regressions are test on distressed firms and healthy firms separately. The reason

for not building a interaction term between distress firms and CDS trading in regression (2) and regression (3) is that it is expected in the hypothesis that CDS trading has adverse effect on firm financing, the interaction term only can reflect if the existence of CDS trading has different effect on firm financing, but can not reflect adverse effect, thus testing the two groups separately can be a better method.

Two credit default swaps variables are included in the empirical specifications. CDS

Trading (CDSQ) is equal to one if the firm has quoted CDS contracts on its debt during year t,

and zero otherwise. CDS Traded (CDSD) is equal to one if the firm has a traded CDS contract on its debt at any time during the 2008–2012 sample period, and zero otherwise. The most important coefficient in both the leverage and maturity equations is that on CDS

Trading, which captures the impact of CDSs on leverage and maturity in all years following

CDS introduction. For the last regression, the main coefficients of interest are that of CDS

Trading, Firm Leverage, Debt Maturity. CDSD captures time-invariant unobservable

differences between CDS and non-CDS firms. In the three equations, FL stands for Firm

Leverage and DM stands for Debt Maturity, X, Y, Z stands for the firm specific control

variables that used to control for omitted variable bias in each equation.

In firm leverage regression (regression (2)), the dependent variable Firm Leverage is measured as total debt (long-term debt plus debt in current liabilities) divided by the firm

(15)

value, which is calculated as book value of assets, minus the book value of common equity, plus the market value of equity and deferred taxes. Debt Maturity is included as one of the explanatory variables in order to prevent the potential rollover risk. The rollover risk is, for example, firms with less leverage are expected to choose debt of shorter maturity. Since Debt

Maturity is measured as the proportion of short term debt maturing within three years, it is

expected to be negatively related with Firm Leverage. In addition to the major independent variables, several control variables, presented as X in the regression, are included according to Johnson (2003). Market-to-Book ratio is defined as the ratio of market value of assets to book value of assets. As a proxy for growth opportunities, Myers (1977) states that firms with high market-to-book ratio will use lower leverage to avoid underinvestment problems. The

Fixed Assets Ratio is the proxy for the tangibility of assets. It is calculated as net property,

plant, and equipment divided by the book value of assets. More tangible assets indicate higher collateral values, leading to higher leverage. Profitability is defined as earnings before interest, taxes, depreciation, and amortization (EBITDA) over book value of assets. It is indicated that more profitable firms are more likely to have lower leverage (Johnson, 2003). As a proxy for firm Size, net sales is included as a control variable. Larger firms usually have lower expected bankruptcy costs and greater debt capacity, since they are more diversified then smaller firms. However, since larger firms have greater access to equity market, from this perspective, their debt may decrease. Firms with higher earnings volatility are expected to have lower leverage, thus Volatility is included in the control variable. It is calculated as the standard deviation of first differences in EBITDA over the four years preceding the observation year, scaled by average assets for that period. Since firms with alternative tax shields care less about high leverage (DeAngelo and Masulis, 1980), two proxies for alternative tax shields are added, and they are expected to be negatively related with firm leverage. One is a dummy variable equal to one if firms have net operating Loss Carry

Forward and zero otherwise. The other one is a dummy variable equal to one for firms with Investment Tax Credits and zero otherwise. Abnormal Earnings is an unusual measure for

idiosyncratic risk, taking into account of the signaling effects in leverage choices, Abnormal

Earnings is positively related with firm leverage. Abnormal Earnings in year t are equal to

change in the operating earnings per share from year t-1 to year t divided by the share price in year t-1. At last, a dummy variable called Regulated Firm is included. It equal to one if the firm’s SIC code is between 4900 and 4939, which refer to electronic and gas service industry. Because managers of those companies may have less discretion over investment decisions, this reduces debt agency costs and thus increases leverage (Smith, 1986).

(16)

In debt maturity regression (regression (3)), the dependent variable Debt Maturity is focus on short-term debt, is defined as proportion of debt maturing in three years or less. Firm

Leverage is included as an independent variable to prevent rollover risk. When firms roll over

short-term debt, they face liquidity risk, therefore firms with high leverage tend to choose long maturity debt to avoid excessive liquidation (Saretto and Tookes, 2011). Since the Debt Maturity in this paper is focus on short-term debt, it suppose to have a negative relationship with firm leverage. As a simplified version of Johnson’s model, this equation has exactly the same control variables with the firm leverage equation. Opposite to the firm leverage equation, Market-to-Book ratio should have a positive relationship with the short maturity measure, since Myers (1977) argue that there is a negative association between growth opportunity and debt maturity, firms with high growth opportunities would like to choose short maturity debt to mitigate underinvestment problems. Fixed Assets Ratio tends to be positively related with loan maturity, thus is negatively related with short-term debt.

Profitability is included in this equation because Graham, Li, and Qiu (2008) find that

profitability is positively related with debt maturity. Diamond (1991) predicts that debt maturity have a first increasing then decreasing relationship with credit quality or liquidity risk, hence short debt maturity should be negatively related with firm Size. Volatility is a measure of credit risk, firms with higher volatility are more likely to default, and thus decrease their debt maturity. The two tax shield dummies are expected to have postive relation with short-term debt maturity, since alternative tax shields can reduce the value of long-term debt. The predict sign of Abnormal Earnings is negative in the debt maturity equation. Regulated Firm is also included and is expected have positive relationship with debt maturity.

In bankruptcy regression (regression (4)), Bankruptcyit is a binary variable equal to one for firms that go bankrupt in the year they file for bankruptcy and 0 otherwise. Z includes control variables that have significant effects on bankruptcy based on former literature, they areSize, Volatility, Abnormal Earnings and Market-to-Book. Size, different from the previous two

regressions, is embodied as market capitalization in this regression, it can be a proxy for the complexity of a firm’s debt structure and is expected to have a concave relationship with bankruptcy, but when firms become larger and safer, the size would be negatively related to bankruptcy (Gilson, John, and Lang, 1990). Volatility and Abnormal Earnings are considered as important determinants of corporate distress as indicated by Shumway (2001). Volatility is expected to be positively related to probability of bankruptcy while the relationship between

(17)

of a company’s performance, as a simplified version for q ratio, Market-to-Book ratio is included as a control variable. It is supposed to be negatively related with bankruptcy rate.

The main coefficients of interest in the firm leverage regression and debt maturity regression are α1 and β1. α1 measures the difference in firm leverage between firms with and without CDSs. β1 measures the difference in proportion of short-term debt between firms with and without CDSs. The most important observations of the bankruptcy regression are the coefficients on Firm Leverage, Debt Maturity, that is, γ1 and γ2. γ1 measures how firm leverage affects bankruptcy rate and γ2 measures how debt maturity affects bankruptcy rate. Besides, the coefficient on CDS Trading (γ3) is also mentionable, since it measures the difference in bankruptcy probability between firms with and without CDSs. If the hypotheses hold, for healthy firms, α1 is expected to be positive and β1 is expected to be negative, γ1 are expected to be negative and γ2 are expected to be positive. Distressed firms are expected to have negative α1, positive β1, negative γ1 and positive γ2.

The advantage of this test and hypotheses is that it corrects the viewpoint that CDSs, as a instrument to transfer risk, can benefit risker firms more, instead, this paper tries to prove that save and large healthy firms can benefit more from the CDSs in terms of increase in firm leverage and debt maturity.

4. Data and Descriptive Statistics

In this study, an unbalanced panel dataset is generated with annual data. This study uses several sources of information to identify firms with CDSs on their debt and examine the distress rate and bankruptcy probability of publicly traded firms. The entire panel contains all U.S. nonfinancial public firms from 2008 to 2012, as reported by Compustat. The sample begins in 2008 because this paper is focus on the question that whether the CDSs lead to an increase in the firms’ bankruptcy probability during the 2008-9 financial crisis and thereafter. The financial firms are not included because financial firms have totally different capital structures and are applied with different government regulation. It also because that the model developed by Saretto and Tookes (2011) to test the relationship between CDSs and firm financing excludes financial institutions, as stated in the methodology part, part of the method in this research is based on Saretto and Tookes (2011) paper.

The panel data are collected from three main resources: Datastream, the UCLA-LoPucki Bankruptcy Research Database and Compustat. The primary source for all firm-specific

(18)

financial information is the Compustat database. It is required that all firms have non-missing data for all explanatory variables of interest. CDS firms in the sample is identified by using Datastream and the source from Matta, R.. Datastream has two main sources of data on CDSs, they are CMA Datavision and Thomson Reuters, the CMA offer market information from active buy-side credit market investors from 2004, and the Thomson Reuters has CDSs data over 1500 entities with histories going back to 2007. Datastream helps to identify those firms with CDSs on their debt that trade actively, also it offers sufficient information for creditors to find a counterparty to hedge their risk. The source used to formally identify companies that are filed for bankruptcy protection (Chapter 11 and 7) during the period 2008-2012 is the UCLA-LoPucki Bankruptcy Database (BRD). The BRD contains an extensive list of large, public company bankruptcy cases filed in the United States Bankruptcy Courts dating back to 1979. It is important to emphasize that, in this paper, terminal default events are limited to firms that filed for bankruptcies, restructuring events are not determined as bankruptcy events.

This paper is endeavor to find the influence of CDSs on bankruptcies of firms through firm financing, that is, to find the effect of CDSs on both firm leverage and debt maturity, and the impact of changes in firm leverage and debt maturity on bankruptcy probability. However, precise information on firm leverage and debt maturity is not available in Compustat, hence the data on capital structure detail in Compustat is used to calculate firm leverage and debt maturity. Non-missing information on leverage and maturity is required for inclusion in the final sample. And firms with zero debt maturity or zero leverage are also excluded, since including those firms can lead to a biased result with an enlarged estimated effect of CDSs.

The final sample of the merged dataset contains 3,905 U.S. nonfinancial firms during the period 2008–12, of which 2,280 firms are in distress. Over the entire sample period, there are 236 firms have CDSs written on their debt at some point during the period, firms that have traded CDSs at any point during the year are identified as firms with CDSs. There are 21 firms went bankrupt at some point during the sample period, but none of them are firms with CDSs on debt. Table 1 presents a year-wise summary of bankruptcies for both distressed and healthy U.S. nonfinancial public firms with and without CDSs. It shows the number of firms and the number of bankruptcies per group. It can be seen that for healthy firms, there is still an increase in the number of firms that trade CDSs from 2008 to 2010, this could be that healthy firms were not affected by the negative part of CDSs even in the crisis. On the opposite, except for some small fluctuations, the distressed firms experienced a gradual decrease in the number of CDSs trading from 2008 to 2012, indicating that distressed firms

(19)

might be affected by CDSs in a negative way during and after the crisis. There is only one firm with CDSs on their debt go bankrupt during the sample period, and this happened after the financial crisis. All the other 20 bankruptcy events go to firms without CDSs, of which 19 firms are in distress and only 1 is healthy firms. Most of those bankruptcies occurred in 2008 and 2009, which is the crisis period.

Table 2 presents summary statistics of the regression data set for the full sample of U.S. nonfinancial firms, as well as separate summary statistics for healthy and distressed firm with and without CDSs. Some variables have extreme values, to reduce the influence of outliers, all control variables are winsorized at the 1st and 99th percentile values by each year. Looking at the total statistics for distressed firms and healthy firms, there are some vital implications. First, the firm leverage is substantial for both types of firms, with respective mean of 0.12 and 0.30. But for distressed firms, the mean of firm leverage is especially high, which is more than twice as that of healthy firms no matter with or without CDSs. The respective standard deviation for healthy firms (0.11) and distressed firms (0.21) imply that firm leverage varies widely across firms, and distressed firms have larger cross-sectional variation. Second, the proportion of debt maturing within three years of healthy firms (0.58)

Table 1. Bankruptcy rate for firms

CDS Firms Non-CDS Firms Year Number of

Firms

Number of Bankruptcies

Number of Firms Number of Bankruptcies Healthy Firms 2008 2009 2010 2011 2012 103 123 138 139 106 0 0 0 0 0 1,125 0 1,156 1 1,251 0 1,167 0 1,007 0 Distressed Firms 2008 2009 2010 2011 2012 82 0 1,244 3 81 0 1,077 6 72 0 911 3 78 0 916 2 70 1 860 1

This table provides a year-wise summary on the number of bankruptcies filed by publicly traded nonfinancial U.S. companies included in Compustat during 2008-2012. The full sample is divided into healthy firms and distressed firms that with and without traded CDSs. It shows the number of firms and the number of bankruptcies per group. Data on bankruptcies are from UCLA-LoPucki Bankruptcy Database (BRD), and data on CDSs are retrieved from Matta, R. and Datastream.

(20)

and distressed firms (0.64) are quite close, the figures imply that the debts within three years take up more than half of the total debt. In addition, there are some differences between healthy firms and distressed firms with respect to the control variables. On average, healthy firms have larger size than distressed firms in terms of net sales and market capitalization, since both sales and market capitalization have large values, the natural log of firm size is used in the regressions. Besides, compared to distressed firms, healthy firms have higher profitability, lower volatility, less net operating loss carry forwards and less negative abnormal earnings.

Comparing the CDS and non-CDS firms gives useful insights. Obviously, CDS firms have slightly higher firm leverage and much more long-term debt with maturity over three years relative to non-CDS firms. Since firms with CDSs traded on their debt are usually large firms (Peristiani and Savino, 2011), CDS firms are larger than non-CDS firms in terms of sales and market capitalization. At the same time, CDS firms maintain higher profitability and lower volatility.

5. Results and Robustness

This part consists of the results of the benchmark specifications and the results after robustness. First, the effects of CDSs on firm leverage and debt maturity are displayed in Table 3. Second, further results about how firm leverage and debt maturity affect the bankruptcy probabilities of firms are shown in Table 4, combining the results from Table 3 and Table 4 gives us a sight into how CDSs influence the bankruptcy probability indirectly through firm leverage and debt maturity. Last but not least, the endogeneity problem is taken into account and an orthogonal instrument of excess CDS exposure is generated to solve this problem.

5.1 Firm leverage, debt maturity and CDSs

Table 3 reports the estimated coefficients of the leverage and maturity regressions. Firm

Leverage is the dependent variable in the results shown in Panel A and Debt Maturity is the

dependent variable in the results in Panel B. In both panels, Columns 1 and 2 present the results of the full sample, Columns 3 and 4 present the results of healthy firms and Columns5 and 6 present the results of distressed firms. Columns 1, 3 and 5 are the benchmark

(21)

Table 2. Summary statistics

Total CDS Firms Non-CDS Firms

Mean Stdev Mean Stdev Mean Stdev

Healthy Firms Distressed Firms Healthy Firms Distressed Firms Healthy Firms Distressed Firms Healthy Firms Distressed Firms Healthy Firms Distressed Firms Healthy Firms Distressed Firms Firm Leverage 0.12 0.30 0.11 0.21 0.16 0.39 0.08 0.13 0.12 0.30 0.11 0.21 Debt Maturity 0.58 0.64 0.37 0.39 0.29 0.27 0.18 0.19 0.62 0.66 0.37 0.39 Market-to-Book 1.47 1.54 3.58 5.67 1.67 1.18 0.65 0.21 1.46 1.76 3.77 5.88 Fixed Assets 0.25 0.32 0.23 0.30 0.34 0.39 0.23 0.26 0.25 0.32 0.22 0.30 Size (Sales) 6.38 4.22 2.44 3.29 9.63 8.68 1.17 1.14 6.02 3.91 2.28 3.16 Profitability 0.01 -1.30 1.15 4.47 0.16 0.08 0.06 0.08 0.00 -1.40 1.21 4.62 Volatility 0.08 0.38 0.34 1.09 0.03 0.04 0.04 0.07 0.09 0.14 0.36 1.13 Investment Tax Credit 0.00 0.01 0.05 0.09 0.02 0.01 0.02 0.10 0.00 0.01 0.04 0.09 Loss Carry Forward 0.15 0.61 0.36 0.49 0.00 0.22 0.08 0.42 0.17 0.63 0.37 0.48 Abnormal Earnings -0.01 -0.11 2.17 2.99 -0.01 -0.14 1.52 2.56 -0.01 -0.11 2.24 3.03 Regulated Firm 0.01 0.07 0.11 0.25 0.01 0.28 0.12 0.45 0.01 0.05 0.12 0.22 Market Capitalization 6.33 4.18 2.28 2.82 9.53 8.39 1.36 1.38 5.99 3.86 2.09 2.63 The summary statistics are for a sample of 11727 Compustat nonfinancial firm-year observations from 2008 to 2012. Firm Leverage is measured as total debt (long-term debt plus debt in current liabilities), divided by the book value of assets minus the book value of common equity plus the market value of equity and deferred taxes. Debt Maturity is focus on short term debt and thus is defined as proportion of debt maturing in three years or less. Market-to-Book ratio is the ratio of book value of assets minus the book value of equity plus the market value of equity, divided by the book value of assets. Fixed Assets is calculated as net property, plant, and equipment divided by the book value of assets. Size is the natural logarithm of total sales, in millions of U.S. Dollars. Profitability is defined as earnings before interest and taxes, divided by the book value of assets. Volatility is calculated as the standard deviation of first differences in EBITDA over the four years preceding the observation year, scaled by average assets for that period. Investment Tax Credit is defined as the investment tax credit, divided by the book value of assets. Loss Carry Forward is calculated as the net operating loss carry forwards, divided by the book value of assets. Abnormal

Earnings in year t are equal to change in the operating earnings per share from year t-1 to year t divided by the share price in year t-1. Regulated Firm is equal to one if the firm’s SIC

(22)

specifications. For robustness, results in Columns 2, 4, and 6 are estimated with time fixed effects and firm fixed effects, controlling for firm fixed effects is to eliminate the possibility that unobserved (time-invariant) heterogeneity between firms may lead to bias in the results on CDS effect. The variables CDS Traded and Regulated Firm are omitted in the fixed effects regressions because that these two variables are not vary over time.

The key variables of interest are those that measure CDS Trading. The most important observation from the leverage regression results is that the coefficient on CDS Trading in the benchmark regressions of healthy firms (Column 3) is statistically significant. Keep other variables constant, the coefficient in Column 3 implies that the introduction of CDS trading leads to a decrease of 0.0232 in firm leverage of healthy firms, which is about 19.3% of the mean leverage of 0.12 (as presented in Table 2). This result is just opposite to what Saretto and Tookes (2011) get, they find that CDS markets allow firms to achieve higher firm leverage and longer debt maturity even during the 2008-09 crisis, but according to the results in Table 3, the presence of CDS market do not have significant effect on distressed firms, but have a negative effect on the healthy firms. One potential concern is that due to the special research period— financial crisis, the negative side of CDS trading has been revealed to a large extent, such as empty creditor problem. Large CDS firms with high leverage are afraid to be forced into liquidation, as they have greater access to equity market, they may decrease debt. As shown in benchmark specification in Table 3, on average, firms with traded CDS have not benefited from an increase in the firm leverage (Columns 1). Except for that, negative effect of CDSs on firm leverage of healthy firms is an economically abnormal result, the possible reason is the reverse-causality between CDS trading and firm financing decisions.

The results of fixed effects regressions (Column 2, 4 and 6) in Panel A show that CDS trading does not have a statistically significant effect on leverage for any types of sample. In the firm fixed effects specification, the estimated coefficient on CDS Trading is interpreted as the average difference of firm leverage and debt maturity before and after the onset of CDS trading for the firms that introduce CDSs during the sample period. The point estimates are, to some extent, lower than the benchmark specifications since controlling for fixed effects limits the degrees of freedom.

The coefficients on CDS Traded in Panel A of Table 3 are statistically insignificant for all types of firms. CDS Traded are used in order to enable the CDS Trading to capture the effect of the onset of CDS contacts, however, they do not show an average difference in firm leverage between CDS and non-CDS firms after controlling for other characteristics. Debt

(23)

Carry Forward are consistently statistically significant in the benchmark regression in Panel

A. Except for Loss Carry Forward, all other variables have estimated coefficients that are in line with the predicted signs in the methodology part for all types of samples.

Table 3. Firm leverage, debt maturity and credit default swaps

Full Sample Healthy Firms Distressed Firms

(1) (2) (3) (4) (5) (6)

Panel A: Firm leverage regression

CDS Trading -0.0116 -0.0143 -0.0232** -0.0146 0.0099 -0.0126 (-0.72) (-1.05) (-1.73) (-1.33) (0.26) (-0.44) CDS Traded -0.068 0.0143 -0.0291 (0.43) (1.07) (-0.76) Debt Maturity -0.1023*** -0.0230*** -0.0625*** -0.0191*** -0.0339*** -0.0044 (-16.90) (-5.57) (-13.22) (-5.32) (-3.39) (-0.50) Market-to-Book -0.0055*** -0.0038*** -0.0059*** -0.0057*** -0.0030*** -0.0030*** (-8.93) (-12.03) (-2.85) (-9.88) (-7.01) (-7.01) Fixed Assets 0.1644*** 0.1629*** 0.0472*** 0.1012*** 0.1356*** 0.1428*** (19.11) (8.99) (7.04) (4.01) (11.73) (5.50) Profitability -0.0194*** -0.0117*** -0.0449*** -0.1876*** 0.0103*** -0.0081*** (-6.77) (-6.89) (-2.78) (-10.26) (-4.66) (-3.55) Log(Sales) -0.0018** 0.0143*** 0.0028*** 0.0194*** 0.0102*** 0.0139*** (-1.71) (5.52) (2.84) (5.36) (6.70) (3.53) Volatility 0.0087 0.0121*** -0.0218 -0.0190** 0.0093 0.0126*** (1.23) (2.81) (-1.23) (-1.95) (1.42) (2.33) Investment Tax Credit -0.0570*** 0.0100 0.0142 0.0365 -0.1260*** -0.0150

(-6.15) (0.29) (1.27) (0.83) (-11.76) (-0.28) Loss Carry Forward 0.0163*** 0.0389*** -0.0397*** -0.0081* -0.1094*** 0.0089 (2.94) (9.91) (-6.08) (-1.54) (-14.62) (1.22) Abnormal Earnings -0.0015** -0.0008*** -0.0009 -0.0011*** 0.0003 0.0001 (-1.69) (-2.18) (-1.22) (-2.52) (0.29) (0.10) Regulated Firm 0.0269*** -0.0077 -0.0668*** (2.75) (-0.61) (-6.13) Constant 0.2400*** 0.0926*** 0.1606*** 0.0247 0.3179*** 0.2088*** (25.07) (5.42) (16.35) (0.98) (24.33) (8.67) Time Fixed Effects No Yes No Yes No Yes Firm Fixed Effects No Yes No Yes No Yes Number of

Observations

8843 8843 4991 4991 3852 3852 R2 0.1372 0.0942 0.1369 0.0380 0.2519 0.1422

(24)

Table 3. Continued

Full Sample Healthy Firms Distressed Firms (1) (2) (3) (4) (5) (6) Panel B: Debt maturity regression

CDS Trading -0.0378 -0.0426 -0.0228 -0.0408 -0.0526 -0.0742 (-1.17) (-1.00) (-0.51) (-0.75) (-1.22) (-1.06) CDS Traded -0.0028 -0.0117 0.0358 (-0.09) (-0.26) (0.84) Firm Leverage -0.3440*** -0.2278*** -0.5605*** -0.4670*** -0.1040*** -0.0262 (-16.44) (-5.57) (-11.95) (-5.32) (-3.39) (-0.50) Market-to-Book -0.022*** -0.0006 -0.0043*** 0.0011 -0.0016*** -0.0006 (-5.86) (-0.56) (-2.64) (0.36) (-4.03) (-0.55) Fixed Assets -0.1982*** -0.0509 -0.2512*** -0.0265 -0.1445*** -0.1128** (-14.02) (-0.89) (-1.85) (-0.21) (-7.38) (-1.76) Profitability 0.0002 0.0019 0.0069 0.0533 -0.0003 0.0008 (0.08) (0.35) (0.43) (0.58) (-0.16) (0.15) Log(Sales) -0.0556*** -0.0369*** -0.0629*** -0.0623*** -0.0529*** -0.0311*** (-33.64) (-4.53) (-25.71) (-3.47) (-23.06) (-3.23) Volatility 0.0042 -0.0003 0.0149 0.0076 0.0078 -0.0026 (0.75) (-0.02) (0.92) (0.16) (1.39) (-0.20) Investment Tax Credit -0.0897*** 0.0963 0.0719 0.0846 -0.0857*** -0.0440

(-4.11) (-0.02) (1.19) (0.39) (-3.96) (-0.33) Loss Carry Forward 0.0826*** 0.0335*** 0.0568*** 0.0429** 0.1613*** 0.0113 (8.47) (2.70) (3.50) (1.65) (11.33) (0.63) Abnormal Earnings -0.0027** -0.0026*** -0.0047*** -0.0021 -0.0015 -0.0032*** (-1.93) (-2.18) (-2.38) (-1.03) (-0.84) (-2.11) Regulated Firm 0.0217 -0.0993* 0.0535*** (1.02) (-1.56) (2.51) Constant 0.9979*** 0.8176*** 1.1066*** 1.0163*** 0.8234*** 0.7550 (80.07) (15.46) (60.37) (8.21) (41.78) (13.06) Time Fixed Effects No Yes No Yes No Yes Firm Fixed Effects No Yes No Yes No Yes Number of Observations 8843 8843 4991 4991 3852 3852 R2 0.3284 0.2985 0.2744 0.2476 0.4231 0.3857 Panel A presents the OLS regression results of annual firm leverage on the two CDS variables and other explanatory variables from Johnson (2003). CDS Trading (CDSQ) is equal to one if the firm has quoted CDS contracts on its debt during year t, and zero otherwise. CDS Traded (CDSD) is equal to one if the firm has a traded CDS contract on its debt at any time during the 2008–2012 sample period, and zero otherwise. Other explanatory variables are all already defined in Table 2. Models are estimated with robust standard errors. Columns 1, 3 and 5 show the results of benchmark regressions without fixed effects. Columns 2, 4 and 6 are regression specifications with time fixed effects and firm fixed effects. *, ** and *** indicate significance levels of 15%, 10% and 5%, respectively. The sample is composed of U.S. nonfinancial firms during the 2008-2012 period with a total of 3,905 firm-year observations.

(25)

Table 3, Panel B reports the results for the debt maturity regression. However, all the coefficients on CDS Trading in the benchmark regressions as well as the fixed effects regressions are statistically insignificant for all types of firms, suggesting that although CDSs provide creditors the ability to hedge risks, this ability does not affect firms’ capital structure in terms of debt maturity. Despite the insignificance, coefficients on CDS Trading are negative for both healthy and distressed firms, which indicate that the presence of CDS trading may decrease in the proportion of short-term debt, in other words, firms with traded CDSs can obtain longer debt maturity. These results do accord with hypothesis 2 but conflict with hypothesis 1. The coefficients on CDS Traded are not statistically significant either.

The other variables that have consistently statistically significant coefficients across all types of firms (i.e. healthy firms, distressed firms and full sample) in the benchmark regression specifications (Columns 1, 3 and 5) are Firm Leverage, Market-to-Book, Fixed

Assets, Size (i.e. net sales) and Loss Carry Forward. With the exception of Market-to-Book,

the signs of the estimated coefficients on these variables are in line with the predictions in the methodology part. One possible reason for why estimate of Market-to-book contradicts with the previous prediction may go to the liquidity risk argument, issuing long-term debt can prevent the inefficient liquidation of firms’ risky growth opportunities. In the fixed effects regressions (Columns 2,4 and 6), all results are quite similar to the benchmark specifications except that some coefficients become insignificant.

In summary, only the firm leverage of healthy firms is significantly affected by the introduction of CDS trading in a negative way, firm leverage of distressed firms and the full sample and debt maturity of all types of samples are not significantly influenced by CDSs. There are two possible concerns that could lead to the quite insignificant results on CDS Trading. First, the models built above assume that the introduction of CDS trading are exogenous to firms’ leverage and debt maturity decisions, however, the endogeineity problem that firms may choose to enter the CDS market based on their leverage and debt maturity decisions may exist, and this problem can be severe during the sample period, which includes the 2008-09 financial crisis. Second, Saretto and Tookes (2011) use the method of two-stage least squares to estimate the two regressions simultaneously while in this research single-equation estimates are used.

(26)

5.2 Determinants of bankruptcy

The results from the bankruptcy regression are shown in Table 4. The benchmark specification is given in Columns 1, 3, and 5, while the results in Columns 2, 4 and 6 are that of regression specifications estimated with time fixed effects and firm fixed effects.

The main coefficients of interest are that on Firm Leverage, Debt maturity and CDS

Trading. The coefficients on Firm Leverage and Debt Maturity from the bankruptcy

regression in Table 4 combined with the coefficients on CDS Trading from the firm leverage regression and debt maturity regression in Table 3 are used to test hypothesis 1 and 2. If hypothesis 1 holds, the coefficients on CDS Trading in the leverage regression should be positive for healthy firms and negative for distressed firms, and the coefficients on CDS

Trading in the maturity regression should be negative for healthy firms and positive for

distressed firms since the Debt Maturity is measured as the proportion of short-term debt maturing within three years. In the bankruptcy regression the coefficients on Firm Leverage and Debt Maturity are supposed to be negative for healthy firms and positive for distressed firms.

As displayed in Table 4, coefficients on Firm Leverage are statistically insignificant in the benchmark regressions (Columns 1, 3 and 5). What is interesting is that after controlling for time fixed effects and firm fixed effects, the coefficients on Firm Leverage of the distressed firms and the full sample become statistically significant (Columns 2 and 6), and imply a negative effect on firms bankruptcy probability. It suggests that unobservable firm-specific heterogeneity is driving the main findings in the benchmark specifications. Keep other variables fixed, one unit increase in firm leverage leads to 0.2961 unit decrease in bankruptcy probability of distressed firms, and 0.1312 unit decrease in bankruptcy probability of the full sample. These results are consistent with the hypothesis 1. Whereas, together with the results obtained from Table 3 that CDS trading do not have a significant impact on firm leverage of distressed firms, it can be concluded that the introduction of CDS trading do not significantly affect the bankruptcy rate of distressed firms with firm leverage acting in between. This result does not correspond to hypothesis 1. The results in the fixed effects specifications are insignificant (Columns 2, 4 and 6). Generally, the standard errors in the fixed effects specifications are higher than in the benchmark specification since introducing firm-level effects can decline the power to identify cross-sectional differences (Roberts and Whited 2011).

Credit default swaps, financing decisions and firm bankruptcies