University of Amsterdam
Faculty: Economics and Business Study track: Quantitative Finance Master thesis
Ruonan Wang 10714359 Supervisor: Jeroen Ligterink
The Effect of Credit Ratings on Firm Leverage
--How does credit ratings deviation affect firm leverage
Statement of Originality
This document is written by Student Ruonan Wang, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
This thesis investigates the effect of credit rating deviation, which is the difference between credit default swap (CDS) implied ratings and actual credit ratings, on leverage for U.S based firm in the period from 2004 to 2016. The empirical results show that credit rating deviation significantly influence the firm leverage on both statistical and economic.
Specifically, I find that an increase of credit rating deviation would lead to an increase of leverage, and vice versa, showing that corporations are trying to maintain target rating by adjusting their leverage to gain the benefits of using debt. Additionally, I show that CDS spreads incorporate the new information much quicker than credit ratings by comparing the changes of them during the financial crisis period, indicating that CDS spreads might be a viable substitute for credit ratings.
1. Introduction
Credit ratings are first published in the early 1900s by John Moody from JP Morgan. Since the credit ratings have been introduced, it becomes a crucial factor that corporations always concern and plays an important role to all market participants (De Haan &
Amtenbrink, 2011). According to Standard & Poor (n.d), which is one of the big three credit rating agencies (abbreviated CRAs), credit ratings express a forward-looking
opinion about the overall creditworthiness of firm itself. Conventionally, it is considered as an assessment of ability to access the capital market and conveys financial information to investors, borrowers and creditors about firm’s credit to meet the financial obligations (Sufi, 2009).
Several existing literatures have provided the evidence that credit ratings are important for managers to determine the capital structure, and they would usually set a target rating to minimize the cost of debt and adjust the leverage to maintain the target rating, such as invest with debt or postpone the investments to reduce debt load. As Graham and Harvey (2001) claimed, credit ratings are the second highest concern for CFOs in determining the capital structure. More specifically, Judge and Korzhenitskaya (2012) found that firms with higher credit rating would increase the demand for debt financing, which leads to a higher leverage. Hovakimian, Kayhan and Titman (2009) also provide evidence that firms increase the leverage to maintain the target rating by repurchasing securities once receiving a higher credit rating. On the other hand, Kisgen (2009), Huang and Shen (2015) claimed that firms would decrease debt load to retain their rating after being downgraded.
In accordance with the criterion of S&P, credit ratings are classified from rating D to rating AAA and were assessed mainly based on the firm’s financial situation and the likelihood of default. To be specific, a firm with a higher rating is usually under better financial situation, meaning that its ability to fulfil their obligations is relative higher than firms with lower credit ratings. In other words, the firm with lower default probability is usually assigned a more favorable rating to show its high creditworthiness for market participants. However, CRAs have been criticized for the late reflection of information and the inherent conflicts of interest make researchers doubting whether credit ratings can truly reflect the creditworthiness of a firm (Fons, Cantor & Mahoney, 2002). Hence, some literatures have proposed that since CDS spreads is a market-based measurement, it might be a substitute of credit ratings to measure creditworthiness because it captures
informational changes more quickly than credit ratings (Flannery, Houston & Partnoy, 2010; Kiesel & Spohnholtz, 2017).
Credit Default Swap (abbreviated CDS) was invented in 1994 by JP Morgan and experienced a dramatic growth during the past twenty years. It is a financial swap agreement between two counterparties that could shift the CDS buyer’s credit risk to the CDS sellers. Therefore, the buyers are insured that they have the right to be compensated for the loss in the event of default, and the sellers take possession of defaulted loan. To acquire this right, CDS buyers have to make periodic premium payment to the sellers, this is generally so-called CDS spreads, which is charged by sellers (Kiff, Jennifer, Elias, Joda & Carolyne, 2009).
CDS spreads are one of the most common used methods for default probability calculation (Chan-Lau, 2006). Specifically, the decrease of CDS spreads means that the firm’s financial situation has improved so that it faces lower default probability than
previous periods. In other words, it is equivalent to an upgrade of credit ratings, showing the higher creditworthiness than before (Hull, Predescu & White, 2004). With less concern about the default, a firm with lower CDS spreads would likely to demand more debt to expand its company scale such as making further investment or acquisitions, which in turn increase the leverage (Saretto & Tookes, 2013). This is considered from the demand side. On the contrary, stand on the view of capital supply side, whether the firms have ability to borrow enough money is largely depends on its creditworthiness, especially for public investors who buy securities in the secondary market (Hilscher, 2013). Because those investors are usually not inside traders, credit ratings are generally the most available public source for them. Hence, the debtors generally evaluate the firm’s creditworthiness by credit ratings to decide whether the security is worthy to be invested or not. However, CRAs have been criticized for a long time about theirs late reaction to the new information. So credit ratings are not an exactly accurate indicator to project current financial situation (Altman & Rijken, 2004). With the change of default probability of firms is not fully reflected in the credit ratings, debtors who are able to lend money might still use the same ratings as assigned in the previous period to evaluate the firm’s ability to fulfil obligations, which might limit the availability to borrow money from debtors.
So, because of the delayed incorporation of credit ratings to the new information, current firm’s financial situation cannot be projected in the both of CDS spreads and credit ratings at the same time. This problem creates a rating gap between actual credit ratings and CDS implied ratings. With higher CDS implied rating but relatively unchanged actual credit ratings when CDS spreads decrease, how could the corporations adjust the leverage have not been discovered. I thereby choose the effect of credit rating deviation on leverage
as my research direction to explore the adjustment on capital structure and the economic reasons behind that, which focus on the central question: How does credit rating deviation affect firm leverage?
To examine the central question, I firstly formulate the hypothesis to check whether there exists a rating gap between CDS implied ratings and actual credit ratings because of the quicker response of CDS spreads to the new information. And then I use the sample
constructed with 18587 U.S based firm-quarterly observations for the period of 2004 to 2016 to examine whether credit rating deviation influences firm leverage or not.
Through empirical investigation, I find that CDS spreads incorporate new information much faster than credit ratings by comparing the change of CDS spreads and credit ratings during the financial crisis period. Then I test the effect of credit rating deviation on leverage, showing that firm leverage is positively correlated to the credit rating deviation.
Consistent with most existing literatures (Hovakimian et al., 2009; Kisgen, 2006; Sajjad & Zakaria, 2018), the results indicate that corporations are trying to maintain a target rating to obtain the optimal capital structure by adjusting leverage once receiving higher or lower ratings.
My thesis has two main contributions. First, I contribute to the existing literatures that CDS spreads incorporate the new information much quicker than credit rating, identifying the existence of credit rating deviation, which is the difference between CDS implied ratings and actual credit ratings. Second, with the existence of credit rating deviation, I provide evidence about how do corporations adjust leverage to maintain the target rating if they face potential rating upgrade or downgrade when CDS implied rating is higher or lower than actual credit rating. Although the adjustment of leverage to the credit rating change has been identified by several researches, the effect on leverage from the view of credit rating
deviation has not been discovered.
This thesis is organized as follows: the first section gives an introduction by describing the background of credit ratings, CDS and research question. Following the introduction, a relevant illustration of credit ratings, CDS spreads and capital structure are provided in section 2. Next section contains the methodology, which carefully describes the research method and specific model. The process of data collection and tested hypothesis are also presented in this section. Section 4 shows the results of tested regressions accompanied with some detailed discussion and robustness check. A summary of key points and suggestions is stated in the final section.
2. Literature review
This section discusses the relevant studies on credit ratings, CDS spreads and capital structure. The first part illustrates the importance of credit ratings in financial market and its effect on capital structure. After that, I will describe the problems of credit ratings to explain why credit ratings is not a good measure of creditworthiness and its linkage to CDS spreads. The introduction of CDS spreads and its effect on leverage are demonstrated in the following part. part 4 explains that how does the deviation could cause an inequality of demand and supply for debt in financial market and Part 5 provides the evidence of credit rating deviation, which is the difference between CDS implied ratings and actual credit ratings.
2.1 Importance of credit ratings
Credit ratings are an assessment of the creditworthiness of borrowers concerning its ability of paying back obligations. It is a crucial factor that corporations always take into account when it has been introduced (De Haan & Amtenbrink, 2011). Over the past few decades, with the dramatic growth of financial markets, the importance of credit ratings to the corporations has increased significantly as well. Claimed by Hovakimian et al. (2009), credit ratings provide valuable information about firm’s financial situation, it matters for market participants such as debtors, creditors and others financial institutions who rely on credit ratings. Furthermore, credit ratings act as a signal that convey the information about firm’s credit quality, which parts of them are not available for public (Kisgen, 2006). Because in order to assess credit ratings, CRAs may receive material corporation
information before it has been disseminated. Therefore, the implicit information attached to the credit ratings can help those uninformed investors to make appropriate investment decisions, which is considered as an efficient mechanism to address the asymmetry information and moral hazard problems (Becker & Milbourn, 2011; De Haan &
Amtenbrink, 2011; Amiram , Beaver, Landsman & Zhao, 2017). Apart from that, Strier (2008) stated that one of the main causes of 2008 financial crisis is the favorable credit ratings assigned by CRAs. He explained that most of the collateralized debt obligations were rated as the highest rating AAA before the crisis. With the AAA imprimatur, these securities were widely perceived as reliable and became very prevalent. As the crisis began, those popular securities defaulted, and then triggered as much of value.
Although credit ratings matter for lots of institutions, it is especially an important consideration in managers’ decisions-making of capital structure. A survey conducted by Graham and Harvey (2001) showed that CFOs regard credit ratings as the second highest
concern when determining their company’s capital structure. In generally, credit ratings are viewed as an indicator that reflect the firm’s ability of accessing the financial market (Hovakimian et al., 2009). To be specific, a firm with lower default risk is usually been assigned higher credit ratings, meaning that it is in a better financial situation. Hence, those firms are provided with greater ability to access the financial markets that might able to borrow more money, which would in turn increase the leverage ratio. In addition, usually, firms might set a target credit rating as they could enjoy lower costs of debt and approach the optimal capital structure with this rating. By adjusting the debt load, the default
probability is likely to change, and then firms are able to retain the target rating to increase the firm value most (Hovakimian et al., 2009; Kisgen, 2009; Sajjad & Zakaria, 2018).
Previous literatures have already investigated the relationship between credit ratings and firm leverage. Kisgen (2006) is the first one who directly examines the effects of credit ratings on capital structure by analyzing the discrete costs (benefits) of different rating class. He illustrated that the significance of credit ratings on leverage from three reasons, which are bond investment regulations, information content of ratings and direct costs incurred to the firm. Later, he extended his research and provided the evidence that firms would decrease leverage after being downgraded to retain their rating if they target specific ratings (2009). Judge and Korzhenitskaya (2012) argued that having a rating is associated with an increased demand for debt financing. In their study, they concentrated on the period of tightening loan supply to the corporate sector, providing the evidence that having a credit rating has an influence on capital structure. Apart from that, Huang and Shen (2015) claimed that firms adjust their capital structure after receiving downgraded ratings, but do not significantly respond to it when ratings are upgraded. Contrary to the view of Huang and Shen, Hovakimian et al. (2009) explored that when firms receive higher ratings than their target, firms make security and repurchase decisions to maintain their optimal rating. Furthermore, Sajjad and Zakaria (2018) discovered the relation between credit rating scales and leverage, where they showed that mid-rated corporations have a relative higher level of leverage than high-rated and low-rated corporations. According to their arguments, compared to the low-rated corporations, mid-rated firm are able to borrow money with lower cost of debt due to their lower default probability and they also enjoy tax shield benefits because of the higher leverage ratio.
As the importance of credit ratings for corporations and financial market, it is crucial to have an accurate credit rating. However, credit ratings might not be a reliable
measurement of creditworthiness.
There have two main reasons. Firstly, CRAs cannot incorporate new information immediately to the credit rating assessment so that they cannot publish the accurate ratings on time. Thus, credit ratings might not be a good measurement of firm’s creditworthiness because of the slow response of CRAs (Altman & Rijken, 2004; Fons et al., 2002). This is the reason why CDS implied ratings do not match the credit ratings, which I will discuss in later sections.
Secondly, the conflicts of interest inherent in the CRAs business model also disclose the problem of credit ratings. In the past few decades, credit ratings were purchased by investors to help them in making financial investments. Later in the 1970s, CRAs changed their model to issuers-based compensation model, which gradually become the most dominant model applied in the industry (De Haan & Amtenbrink, 2011). In this model, CRAs receive payment from the issuers rather than investors for publishing the ratings. This creates a conflict of interest between CRAs and issuers that CRAs have the incentive to meet the preferences of issuers so they could build and keep a good relationship with them and might not lose those customers in the following years (White, 2010). But, this incentive contradicts the primary purpose of CRAs that they aim to provide an objective, independent and accurate credit-risk assessment. Therefore, there is a trade-off between satisfying the requirements of paying issuers and providing objective and accurate credit ratings.
Given the above two problems, researchers start to look for other reliable measures of creditworthiness. With much quicker and accurate reflection of new information in CDS market, researchers moved to evaluate creditworthiness by CDS spreads (Forte & Pena, 2009; Flannery et al., 2010; Kiesel & Spohnholtz, 2017).
2.3 CDS spreads
The CDS markets have grown rapidly over the last decade. CDS is a credit derivative contract with aiming to shift the default risk to a third party (Kiff et al., 2009). To exchange the protection that the loss incurred for CDS buyers will be compensated when credit events occur, they would pay the premium to CDS sellers periodically, which is known as CDS spreads (Weistroffer, 2009).
CDS spreads are one of the widely accepted methods that directly calculate the default probability (Chan-Lau, 2006). And similar to credit ratings, CDS spreads also evaluate the default risk or credit event associated with the underlying obligations, which provide valuable information about firm’s creditworthiness (Giglio, 2016). CDS spreads are depend on the supply and demand for the contracts and change over the time (Flannery et al., 2010). For instance, once the default probability of creditors increases, the debtors might look for some protection in case of credit events, leading to an increased demand for CDS contracts.
Research by Saretto & Tookes (2013) found that firms with CDS trading tend to maintain higher leverage ratio because of the less concern for the potential default. In addition, Hull et al. (2004) examined the relation between CDS spreads changes and credit ratings, where they found that CDS spreads changes provide helpful information in
anticipating the credit ratings changes, especially for the negative credit ratings events. Viral and Timothy (2007) further argued that CDS spreads might be a better measure of credit quality because CDS markets appear to reflect non-public information, which provide a more accurate projection of firm’s current financial situation. What’s more, Zhang and Zhang (2013) tested the information efficiency of CDS market by examining earnings surprises, providing an overall conclusion that CDS market timely reflect the credit risk of the underlying reference entities.
As mentioned before, credit ratings play a considerable role in financial market that provide valuable information about firm itself. However, CRAs who provide credit ratings have been criticized for its late reaction in updating ratings so that credit ratings may not be a reliable measurement of firm’s financial situation (Altman & Rijken, 2004; Fons et al., 2002). Although CDS spreads have similar functions as credit ratings, the key difference between them is that CDS spreads incorporate new information more quickly than credit ratings (Forte & Pena, 2009; Flannery et al., 2010). It better reflects the change of firm’s financial situation, especially it has high correlation with liquidity factors and credit risk, meaning that the higher the credit risk, the higher the CDS spreads (Diaz, Groba & Serrano, 2003).
Based on those finding, recent studies proposed that CDS spreads might be a better measurement of credit quality alternative to credit ratings. Flannery et al. (2010)
discovered that CDS spreads reflect the material information change more promptly and responsively than CRAs by comparing the change of mean CDS spreads for different bond ratings. Kiesel and Spohnholtz (2017) constructed a linear regression that uses the
logarithm CDS spreads as a viable explanatory variable to predict credit ratings. In the research, they argued that CDS market has the ability to complement the risk assessment of CRAs. So Kiesel and Spohnholtz implemented an independent measure of credit ratings based on CDS spreads, where they found this CDS spreads based measurement is able to anticipate credit ratings prior to CRAs.
2.4 Inequality in financial markets
Traditionally, banks, investors or other financial institutions that provide borrowing service will always evaluate the credibility of borrowers before lending money. Credit ratings are the most often used evaluation of credit risk, especially for the financial market (Hilscher, 2013; Kiesel, 2016). For instance, as the credit ratings become lower, the default risk of firms becomes higher. So investors who supply money are less willing to buy
low-rated corporation’s securities, limiting the debt that corporations can be borrowed. But, since new information is not fully reflected in credit ratings immediately, the credit ratings would remain relatively unchanged (Flannery et al., 2010). So the supply of debt also does not have a significant change.
Contrary to the supply side, the demand for debt is reflected in the changes of CDS spreads. As the CDS spreads decrease, the financial situation is improved, meaning that the default probability decreases. So because of the less concern about default risk, corporations are willing and have ability to make further investments. According to the pecking order theory (Myers & Majluf, 1984), companies prefer the financing source of debt rather than equity, with relatively unchanged debt supply, the decrease of CDS spreads would result in an excess demand for debt.
2.5 Credit rating deviation
As discussed in section 2.2, because of the delayed response of CRAs, new information might not be immediately reflected in the credit ratings, which provide unreliable credit quality. Since CDS spreads are also a valuable source of credit information and well anticipated the new information, more and more researches start to rely on CDS spreads to assess creditworthiness (Angelini, 2012; Kiesel & Spohnholtz, 2017). However, due to the different reaction time of credit ratings and CDS spreads, the CDS implied ratings do not exactly match the actual credit ratings. The difference between CDS implied ratings and actual credit ratings is what I called credit rating deviation.
According to the previous literatures, a firm that experiences a downgrade significantly decreases its leverage (Huang & Shen, 2015). And an upgrade would potentially increase firm leverage by making different types of financial actions such as investments, acquisitions (Hovakimian et al., 2009). Regard to the CDS spreads, generally, the lower the spreads, the lower the probability of default, which might upgrade the credit ratings and increase the firm leverage in turn (Saretto & Tookes, 2013). The effect of actual credit ratings on capital structure has been discovered widely, but the effect of credit rating deviation on capital structure is not clear yet.
As a conclusion, because of the quicker response of CDS spreads than credit ratings, the CDS implied ratings do not match the actual credit ratings so that it creates a credit rating deviation. And CDS spreads can be considered as a leading indicator to reassess the credit ratings. With aiming to maximize the benefits of using debt and maintain target rating, corporations would take actions to adjust leverage. However, how would corporations adjust its capital structure with such a positive or negative credit rating deviation has not been discovered.
3. Methodology and data
Having discussed the relevant literatures, this section describes the hypotheses and methodology that will be applied to check the statements. Data gathering and construction of sample with descriptive tables are explained in the third part.
3.1 Hypotheses
As argued by Flannery et al. (2010), Kiesel and Spohnholtz (2017), due to the characteristic of market-based, new information could be reflected faster in CDS spreads than credit ratings. Therefore, they claimed that CDS spreads are a viable substitute for credit ratings. To derive the credit rating deviation, I first formulated the following hypothesis to check whether CDS spreads incorporate the information faster or not:
1) CDS spreads response quicker to the changes than credit ratings
Based on the existing studies (Flannery et al., 2010; Viral & Timothy, 2007), CDS spreads incorporated new information faster and provide more accurate a projection of current financial situation than credit ratings. So if a firm’s financial situation has improved,
CDS spreads would decrease to reflect the decrease of default risk, which correspond to a rating upgrade, whereas its credit rating remains relatively unchanged. That results in a positive difference between CDS implied ratings and actual credit ratings, and vice versa.
With the financial situation improves or deteriorate, corporations face the potential credit rating upgrade or downgrade so that they might adjust their leverage to maintain the target rating to minimize the cost of debt (Hovakimian et al., 2009; Kisgen, 2009). Thus, I would expect an increased leverage for positive credit rating deviation, and a decrease of leverage for negative credit rating deviation. Then the following hypothesis is formulated to check the effect of credit rating deviation on firm leverage:
2) An increase (decrease) of credit rating deviation increases (decreases) the firm leverage
3.2 Research methodology
Consistent to the method used by Flannery et al. (2010), to test the first hypothesis, I compare the changes of average CDS spreads for four different credit ratings for the entire sample of firms, which are AAA, AA, A and BBB. The AAA rating represents the highest attainable rating, whereas the BBB rating the threshold of investment grade.
The sample period applied is the financial crisis period from 2006 to 2009. During this period, the dramatic increase of default risk causes the change of both CDS spreads and credit ratings. With both of them face the severe trigger event, it is much easier to observe and identify which one changes faster. Furthermore, to have a more intuitive understanding, I compare the average CDS spreads for a selected subsample of individual firms. This subsample consists of nine financial institutions, including three large investment banks, three commercial banks and three other firms in which has relative larger market
capitalization.
The methodology of testing the second hypothesis is divided into three steps. Similar to the research conducted by Kiesel and Spohnholtz (2017) and Almeida et al (2017), as a starting point, in order to classify the credit ratings, I first convert the rating grades into numerical value, which take value of 1 for rating D, up to value of 22 for rating AAA. Table 1 shows the rating grade and its corresponding converted numerical value.
Table 1
Rating Grade with Corresponding Numerical Value
The first step is to estimate the relation between CDS spreads and credit ratings by running the following regression.
!"#$%&',) = +,+ +./01 3456"7',)+ +89%(;)',) + +=!>;',)+ +? @
AB)CDE ',)+
+F1GH"567 9%(;)',) + +I1GH"567 !>;',)+ +J1GH"567 @
AB)CDE ',)+ K$5L KM +
N%7H3#5O KM + P$L6 KM + Q',) --- (1) In this regression, the dependent variable is converted credit ratings numerical value, and the explanatory variable is 5-year CDS spread, which is considered as the most liquid one and commonly used in other literatures (Longstaff et al., 2005). To avoid endogeneity, a set of control variables such as lnassets, ROA that are used to predict credit ratings is included (Kisgsen, 2006). Firm fixed effect, industry fixed effect and time fixed effect are added in the regression as well.
S&P Rating grade Numerical value
AAA 22 AA+ 21 AA 20 AA- 19 A+ 18 A 17 A- BBB+ BBB BBB- BB+ BB BB- B+ B B- CCC+ CCC CCC- CC C D 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
After obtaining the estimated coefficients from the above regression, the next step is to apply those coefficients to calculate the predicted credit ratings for each sample firm.
In the last step, the second regression is regressed to test the effect of credit rating deviation on firm leverage.
R0!',) = +,+ +.(S567$T#67 5"#$%& − ;T#H"V 5"#$%&)',)+ +8K$5L 3$W6 + +=S5XY$#"Z$V$#O',)+ +?L"5[6# − #X − ZXX[ 5"#$X',)+
+FK$\67 "336#3 5"#$X',)+ +I06456T$"#$X% 5"#$X',) + K$5LKM + N%7H3#5O KM + + P$L6 KM + Q',) --- (2)
Following the method used by Flannery & Rangan (2006), the dependent variable is the market-to-debt ratio, which is defined as total liabilities divided by total liabilities plus the market value of equity. The key explanatory variable is the difference between predicted rating and actual rating so-called credit rating deviation. According to the existing literatures (Hovakimian et al., 2001; Fama & French, 2002; Flannery & Rangan, 2006; Almeida et al., 2017), there have some other factors that determine the capital structure. Hence, some control variables that regularly appeared in the existing literatures are chosen to be included in the regression. The implication of control variables with respect to firm leverage and their measurement are explained below.
1. Firm size: lnassets is defined as the natural logarithm of assets that is used to proxy firm size. Larger firms often manage their company with more debt because they face the lower probability of default, and have less uncertainty about cash flow (Flannery & Rangan, 2006). Therefore, a positive relation between lnassets and firm leverage is expected.
2. Profitability: EBIT is an indicator of profitability that usually applied, which is defined as earnings before interest and taxes divided by total assets. According to Flannery and Rangan (2006), corporations with higher earnings could either have lower or higher leverage. Higher earnings would reduce firm leverage. On the other hand, higher leverage might be the reflection of the firm’s ability to fulfil their debt obligations. The expectation of coefficient sign is thus hard to define.
3. Market-to-book ratio: Market-to-book ratio is related to future growth opportunities. It is measured as market value divided by net book value.
Usually, higher MTB ratio is seen as a sign of potential growth options, so firms tend to limit the leverage because they want to protect those opportunities (Flannery & Rangan, 2006). From the opposite perspective, firms might
increase their demand for debt to make investments. Hence, it could be either positive or negative relationship among the market-to-book-ratio and leverage. 4. Fixed assets ratio: Fixed assets ratio is measured as a proportion of tangible
assets such as property, plant over total assets. Generally speaking, the greater tangible assets, the higher debt capacity because firms could use those assets as collateral (Flannery & Rangan, 2006). So, a positive coefficient is expected. 5. Depreciation ratio: Depreciation ratio is the percentage of depreciation
expenses to total assets. The higher the depreciation, the lower the EBIT, which could make the firms to have less need for the interest deductions provided by debt financing, which might decrease the leverage (Flannery & Rangan, 2006).
3.3 Data
To analyze the effect of credit ratings deviation on firm leverage, I use empirical data. The sample used in this thesis is constructed with US firms for the period from 2004 to 2016. To construct a proper sample, firms are selected based on the following criterion:
a) Firm has been rated and has available credit ratings datas
b) CDS has been introduced in that firm, and the traded currency is U.S dollar c) The firm is the public company and was founded in the U.S
d) There exists available data for other control variables
Company’s credit ratings are retrieved from Wharton Research Data Services
(WRDS)--Capital IQ--North American--Ratings, using the S&P domestic long term issuer credit rating. The 5-year CDS spreads mid are obtained from Datastream. Additionally, data for other control variables of regression one and two are collected from WRDS
Compustat--North America--Fundamentals Quarterly.
The entire sample consists of 18587 U.S based firm-quarterly observations. Table 2 presents the frequency of credit ratings over the period of 2004 to 2016. In this table, it shows that most of the firms are graded range from rating A+ to rating BB-. Given those observational data, it can be seen that the most frequently assigned credit rating is BBB
that account for over 17 percentage of total observations, while no firm has been graded as rating C in this sample.
Table 2
Frequency Table of Credit Ratings
This table presents the frequency of credit ratings over the 2004 to 2016 period with quarterly frequency. Based on the criterion of S&P, credit ratings are classified from rating D to rating AAA.
Table 3 reports the descriptive statistics of other variables in regression one and two. To mitigate the influence of extreme values and get more desirable results, each variable has been winsorized at level 0.1 and 99.5 percentile including CDS spreads. From the data displayed in this table, the mean of CDS spreads is 217.17 basis points with a standard deviation of 702. Apart from that, the average firm size and squared of that are 8.17 and 69.74. Regard to the ROA, the mean of ROA and squared ROA is 0.07 and 0.18 with standard deviation of 0.5 and 2.2 respectively. Both of debt to total capitalization ratio and squared ratio have the mean that is around 0.4 and similar standard deviation.
Numerical Value Rating grade Frequency Percent (%)
22 AAA 247 1.33 21 AA+ 92 0.49 20 AA 343 1.85 19 AA- 414 2.26 18 A+ 1029 5.54 17 A 2038 10.96 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 A- BBB+ BBB BBB- BB+ BB BB- B+ B B- CCC+ CCC CCC- CC C D Observations 1630 2430 3163 2325 1000 1013 911 669 585 424 147 39 6 24 0 58 18587 8.77 13.07 17.02 12.51 5.38 5.45 4.90 3.60 3.15 2.28 0.79 0.21 0.03 0.13 0.00 0.31 100.00
In terms of the variables related to the leverage, the mean of MDR is 0.48 with a standard deviation of 0.2 as described in Panel B of Table 3. The variables of profitability, depreciation and R&D have mean of 2, 1 and 1 percentage respectively, which both of them have relative smaller standard deviation. In addition, the average market to book ratio is 2.4, and the percentage of fixed assets is around 30.
Table 3
Summary Statistics of Variables
This table presents the summary statistics of variables. Panel A reports the statistics of variables in regression 1, and Panel B reports regression 2. The sample consists of U.S firms over the period of 2004 to 2016 with quarterly data. Standard deviations of means are reported in parentheses. Panel A N Mean CDS spreads 18587 217.173 (702.588) Firm Size 18344 8.169 (1.732) ROA 18325 0.066 (0.508) Debt/Total Cap 16712 0.472 (1.451) Squared Ln(assets) 18344 69.736 (28.573) Squared ROA
Squared Debt/Total Cap
18325 16712 0.187 (2.182) 0.406 (1.743) Panel B N Mean MDR Profitability Market/Book Fixed assets Depreciation 17948 18389 17948 16831 16360 0.484 (0.212) 0.015 (0.034) 2.411 (46.027) 0.308 (0.249) 0.009 (0.006)
4. Results
The first part presents the empirical results with discussions. The second part shows the robustness check.
4.1 Results
Figure 1 shows the changes of mean CDS spreads for four different rating grades in the period of 2006 to 2009. Due to the severely financial crisis occurred during this period, plenty of corporations have suffered financial distress, resulting in a remarkable increase of default probability. Consistent with Flannery et al. (2010), the figure shows that CDS spreads are relatively stable prior to the crisis for all rating grades, and then start to rise immediately at the mid of 2007, which is the beginning of financial crisis. Later, with the government intervention, economy starts to recover so that CDS spreads tend to decrease. What is more, it is clear that firms with higher creditworthiness have relatively lower CDS spreads than other firms, meaning that the default probability is lower. For instance, firms who have been graded as rating BBB have higher CDS spreads compared with AAA firms, and the spreads increase significantly because of the financial shock. Although the mean CDS spreads might not be a reliable and precise estimator due to the limit data included in the subsample or survivorship bias, the overall trend of CDS spreads for different credit rating grades can be clearly identified.
Then to have a more intuitive understanding of quicker reflection of CDS spreads, I examine the changes of CDS spreads and credit ratings for a subsample of individual firms. As displayed in Table 4, CDS spreads immediately respond to the available information about the increased riskiness since the crisis started in mid-2007. On the contrary, credit ratings remained relatively constant throughout the sample period, especially at the
beginning of crisis. Take Merrill Lunch as an example to illustrate. Prior to the crisis, both the CDS spreads and credit ratings remain relatively stable. However, with the significant effect on credit risk, CDS spreads increased dramatically and reached the peak at the end of 2008, whereas credit ratings only have minor downgrades from AA- to A.
Flannery et al. (2010) explained that credit ratings did not capture informational change as quickly as CDS spreads because spreads are a market-based measurement of information. They further explained that different from the credit ratings, CDS spreads that based on five-year agreements reflect both the change of risk in systematic and firm
specific. So, even though the firm itself does not have individual risk exposure, if the whole market expects that the default probability might increase over a five-year period, a
tiny change would still be reflected in the CDS spreads. Angelini (2012) also concluded that CDS markets act as a leading price indicator as CDS spreads provide valuable credit conditions.
To conclude, according to the data presented in Figure 1 and Table 4, there are two conclusions could be retrieved. Firstly, the default probability for different credit rating grades is explicitly reflected by the changes of CDS spreads. Secondly, the evidence suggests that credit ratings remained largely unchanged during the entire sample period while CDS spreads displayed significant change along with the crisis period. Hence, it is reasonable to conclude that CDS spreads incorporate the information quicker than credit ratings when responding to credit events, which supports the first hypothesis. In addition to that, consistent with the first hypothesis implicit confirms the existence of credit rating deviation, which will be used later to identify the adjustment of leverage.
Table 4
CDS Spreads and Credit Ratings Changes in Period 2006-2009
This table presents the CDS spreads and credit ratings changes for subsample firms, including nine selected large public firms. Bear Stearns Spread Rating Merrill Lynch Spread Rating Morgan Stanley Spread Rating Bank of American Spread Rating Citigroup Spread Rating JPMorgan Chase Spread Rating AIG Spread Rating Fannie Mae Spread Rating Washington Mutual Spread Rating
31/03/2006 A 20.8 A+ 21.9 A+ 14 AA- 13.5 AA- 18.2 A+ 20.4 AA 11 AAA 31.5 A-
31/06/2006 25 A 23 A+ 25 A+ 12 AA- 12 AA- 18 A+ 17 AA 6.5 AAA 32.5 A-
30/09/2006 23 A 20 A+ 21 A+ 9.8 AA- 9 AA- 15 A+ 11.3 AA 7.8 AAA 25.2 A-
31/12/2006 21.7 A+ 16.5 AA- 23 A+ 8.6 AA- 7.4 AA- 16 A+ 10.7 AA 7 AAA 22.2 A-
31/03/2007 38 A+ 34.7 AA- 32.4 A+ 13 AA 13 AA 18.2 AA- 13.9 AA 10 AAA 53.5 A-
31/06/2007 51.7 A+ 37.7 AA- 37.3 A+ 14 AA 13.8 AA 20 AA- 13.5 AA 10 AAA 41.7 A-
30/09/2007 84 A+ 63.2 AA- 61.2 AA- 31.9 AA 32.1 AA 35.5 AA- 32 AA 18.5 AAA 91.4 A-
31/12/2007 176 A 125.3 A+ 99.5 AA- 49.2 AA 72.5 AA 49.7 AA- 68.9 AA 39 AAA 412.5 A-
31/03/2008 170 AA- 267.2 A+ 186.5 AA- 97.1 AA 171 AA- 97.5 AA- 181.9 AA 50.9 AAA 488.8 BBB
31/06/2008 102.8 AA- 250.2 A 201 A+ 113.5 AA 140.1 AA- 104.2 AA- 217.5 AA- 66.6 AAA
30/09/2008 412.5 A 972.1 A+ 147.5 AA- 292.9 AA- 141.7 AA- 1442.3 A- 37.5 AAA
31/12/2008 455 A 400 A 117.3 A+ 189.1 A 119.3 A+ 535 A- 37.5 AAA
31/03/2009 363.5 A 390 A 395.9 A 629.6 A 203.4 A+ 2083.1 A- 37.5 AAA
31/06/2009 247.9 A 200.5 A 217.9 A 416.9 A 105 A+ 1386.2 A- 37.5 AAA
30/09/2009 163.7 A 139.8 A 119.6 A 197.1 A 69.7 A+ 772.5 A- 37.5 AAA
As can be seen in Table 5, the coefficients of CDS spreads displayed in both columns are significantly and negatively correlated with credit ratings.
After obtaining the coefficients of each variable from this regression, I calculated the CDS implied ratings by applying those estimated coefficients for each sample, and then take the difference between CDS implied ratings and its actual credit ratings named as credit rating deviation.
Table 5
Effect of CDS Spreads on Credit Ratings
This table presents the results of regression 1. The dependent variable is credit ratings and the key explanatory variable are CDS spreads. This sample consists of 18120 observations with quarterly frequency. Robust-heteroscedasticity standard errors are reported in parentheses. ***, **, and * indicate the significant level of 1%, 5%, and 10% respectively.
Variables (1) (2) CDS spreads Ln(assets) -0.00058*** (0.00005) -0.00044*** (0.00005) 0.17638* (0.09512) ROA 0.14129*** (0.02889) Debt/Total Cap 0.00872 (0.01370) Squared Ln(assets) 0.01019* (0.00603) Squared ROA 0.02637*** (0.00610) Squared Debt/Total Cap -0.09956***
(0.01921) Firm FE Industry FE Time FE YES YES YES YES YES YES F-squared 111.79 51.03 R-squared 0.8628 0.8877
Table 6 reports the second regression result with robust-heteroscedasticity standard error and multi fixed effects. In that regression, I test the effect of credit rating deviation on leverage to explore the actions of corporations when facing potential upgrade or downgrade.
From this table, the coefficient displayed in column 1 indicated a positive relationship between credit rating deviation and firm leverage with an estimated coefficient of 0.00064,
which is statistically significant at 1%. To mitigate the effects of other variables that have a potential influence on both of leverage and credit rating deviation, control variables are added in the regression and results are reported in column 2. Compared to the previous one, the coefficient of credit rating deviation remains significant and positive, meaning that an increase of 1 unit of credit rating deviation would increase the leverage ratio by 0.076%. This is consistent with the existing literatures that firms would maintain their target rating to approach the optimal capital structure, which can bring more benefits to the firm most. Therefore, firm’s leverage would change significantly to target the threshold set by CRAs in order to maintain this certain rating (Begley, 2015; Hovakimian et al., 2009; Kisgen, 2006; Sajjad & Zakaria, 2018). The firms who receive higher credit ratings might finance with debt to make further investments. Contrary to that, for those firms with relatively lower credit rating compared to theirs target rating, they are likely to postpone the investments in order to reduce the debt load or finance with equity so that the default probability could decrease, which could help them to upgrade the credit ratings. On the other hands, based on pecking order theory, debt is a priori financing source than issuing equity. So if the firm’s financial situation has improved, the leverage is expected to increase as the investment is largely financed with debt (Myers & Majluf, 1984).
Furthermore, controls variables are almost significant except market-to-book ratio. The significant and positive coefficient of firm size is in line with explanations that larger firms tend to manage the company with debt (Flannery & Rangan, 2006). In addition, as discussed before, although there is no clear expectation about the effect of profitability and
market-to-book ratio on leverage, both of them show the negative sign, which indicates a negative relationship. Also, the coefficient of fixed assets and depreciation are positive and significant, implying that firms would increase their leverage with more fixed assets. Contrary to the expectation, the coefficient of depreciation is positive. This could be the reason that holding more fixed assets leads to higher depreciation.
In sum, given the empirical evidence, it is reasonable to draw a conclusion that credit rating deviation is positively correlated with firm leverage, supporting the second hypothesis that an increase (decrease) of credit rating deviation increases (decreases) the firm leverage.
Table 6
Effect of Credit Rating Deviation on Leverage
This table presents the results of regression 2. The dependent variable is market to debt ratio and the key explanatory variable is the credit rating deviation. This sample consists of 16038 observations with quarterly frequency. Robust-heteroscedasticity standard errors are reported in parentheses. ***, **, and * indicate the significant level of 1%, 5%, and 10% respectively.
Variables (1) (2)
Credit rating deviation
Firm size 0.00064*** (0.00024) 0.00076*** (0.00025) 0.02185*** (0.00212) Profitability -0.71670*** (0.07327) Market/Book -0.00001 (0.00000) Fixed assets 0.16897*** (0.02637) Depreciation 1.39825*** (0.45514) Firm FE Industry FE Time FE YES YES YES YES YES YES F-squared 7.01 64.47 R-squared 0.8593 0.8587 4.2 Robustness check
To check the quality of estimated coefficients from regression two, two robustness checks with different restriction are further regressed.
The first robustness check restricts the industry by excluding financial firms and utilities (SIC 6000-6999, 4900-4999) from the sample. Because for those type of firms, they might consider some special factors when making capital structure decisions, which is consistent with the research conducted by Flannery & Rangan (2006), and Kisgen (2009).
As shown in table 7, the coefficients of credit rating deviation remain relatively unchanged, which are still positively related to the leverage and significant at 1%. And other variables also do not show any conflict with the previous conclusion. Therefore, without the influence of financial firms and utilities, it still provides the evidence for supporting the hypothesis.
Table 7
Effect of Credit Rating Deviation on Leverage
This table presents the results of first robustness check. The dependent variable is market to debt ratio and the key explanatory variable is the credit rating deviation. This sample consists of 12199 observations with quarterly frequency. Robust-heteroscedasticity standard errors are reported in parentheses. ***, **, and * indicate the significant level of 1%, 5%, and 10% respectively.
Variables (1) (2)
Credit rating deviation
Firm size 0.00057** (0.00028) 0.00075*** (0.00028) 0.02431*** (0.00220) Profitability -0.66825*** (0.07285) Market/Book -0.00001 (0.00000) Fixed assets 0.28294*** (0.02970) Depreciation 1.21751** (0.52510) Firm FE Industry FE Time FE YES YES YES YES YES YES F-squared 4.07 68.04 R-squared 0.8126 0.8310
The second robustness check expands the dataset by including R&D as control variables. R&D is measured as a proportion of total assets. Firms with more intangible assets in the form of R&D will have a lower leverage ratio (Flannery & Rangan, 2006).
Compared with previous results, the sign of credit rating deviation coefficient shown in Table 8 is still positive, but it is not significant anymore. One possible explanation could be the relatively smaller sample size. Over 50 percent of firms do not report R&D expense or do not report quarterly, which limits the observations can be used.
Other control variables do not change significantly, and the coefficient of R&D is consistent with the expectation that higher R&D lead to lower leverage.
In sum, although the statistic results of two robustness checks with different
specifications are a little bit different with previous one, the effect of credit rating deviation is still in line with the hypothesis from the economics perspective.
Table 8
Effect of Credit Rating Deviation on Leverage
This table presents the results of second robustness check. The dependent variable is market to debt ratio and the key explanatory variable is the credit rating deviation. This sample consists of 5737 observations with quarterly frequency. Robust-heteroscedasticity standard errors are reported in parentheses. ***, **, and * indicate the significant level of 1%, 5%, and 10% respectively.
Variables (1) (2)
Credit rating deviation
Firm size 0.00064** (0.00024) 0.00017 (0.00037) 0.01929*** (0.00329) Profitability -0.89301*** (0.10175) Market/Book -0.00003 (0.00006) Fixed assets 0.16598*** (0.04200) Depreciation
Research & Development
1.12016 (0.24079) -0.87348*** (0.24079) Firm FE Industry FE Time FE YES YES YES YES YES YES F-squared 7.01 28.51 R-squared 0.8593 0.8502 5. Conclusion
The purpose of this thesis is to estimate the effect of credit rating deviation, which is the difference between CDS implied rating and actual credit rating, on firm leverage. To test this relationship, a sample that contains 18587 U.S based firm-quarterly observations for the period of 2004-2016 is constructed.
Similar with the researches by Flannery et al. (2010), Viral and Timothy (2007), I first found that CDS spreads incorporate the new information much quicker than credit rating because of its characteristic of market-based, showing that CDS implied rating does not match the actual rating of the firm. Hence there exists a credit rating deviation.
With the credit rating deviation, then a regression analysis is performed with the key explanatory variable—credit rating deviation and a set of control variables. And the dependent variable is firm leverage. The results indicate that credit rating deviation is significantly and positively correlated with firm leverage, meaning that corporations would likely to increase debt load when credit rating deviation increases. This is consistent with Hovakimian et al. (2009), Kisgen (2009), Huang and Shen (2015) that firms increase or decrease leverage to maintain the target rating if their rating has upgraded or downgraded. Additionally, the results are also in line with the pecking order theory that debt financing is preferable than equity financing, causing a potential increase on leverage ratio (Myers & Majluf, 1984).
There also have some limitations of this thesis. Firstly, I only use the credit ratings from S&P in the thesis, so it cannot ensure that Moody and Fitch would also face the same problem as S&P that change of credit ratings are slowly than CDS spreads. For further research, it might helpful to check whether the results would still be consistent or not by using credit ratings from others CRAs. Secondly, other control variables that are not
included in this thesis might also influence the leverage ratio. For example, industry median debt ratio is used to control the characteristics of industry that not captured by the fixed effect. However, the ratios appeared in other literatures were calculated almost twenty years ago. Hence, it is not appropriate to use them and it is hard to find reliable ratios to control the characteristic.
Reference
Angelini, E. (2012). Credit Default Swaps (CDS) and their Role in the Credit Risk Market.
International Journal of Academic Research in Business and Social Sciences, 2(1),
584-593.
Almeida, H., Gunha, I., Ferreira, M. A. & Restrepo, F. (2017). The real effects of credit ratings: The sovereign ceiling channel. The Journal of Finance, 72(1), 249-290. Altman, E. I., & Rijken, H. A. (2004). How rating agencies achieve rating stability.
Journal of Banking and Finance, 28(11), 2679-2714.
Amiram, D., Beaver, H. W., Landsman, W. R. & Zhao, J. (2017). The effects of credit default swap trading on information asymmetry in syndicated loans. Journal of
Financial Economics, 126(2), 364-382.
Becker, B. & Milbourn, T. (2011). How did increased competition affect credit ratings?
Journal of Financial Economics, 101(3), 493-514.
Begley, T. A. (2015). The Real Costs of Corporate Credit Ratings. Paris December 2014 Finance Meeting EUROFIDAI-AFFI Paper. Available at SSRN:
https://ssrn.com/abstract=2404290
Chan-Lau, J. A. (2006). Market-Based Estimation of Default Probabilities and Its
Application to Financial Market Surveillance. International Monetary Fund, Working
Paper (104).
De Haan, J. & Amtenbrink, F. (2011). Credit rating agencies. De Nederlandsche Bank
(DNB), Working Paper (278).
Diaz, A. Groba, J. & Serrano, P. (2003). What drives corporate default risk premia? Evidence from the CDS market. Journal of International Money and Finance, 37, 529-563.
Flannery, M. J. & Rangan, K. P. (2006). Partial adjustment toward target capital structures.
Journal of Financial Economics, 79(3), 469-506.
Flannery, M.J., Houston, J. F. & Partnoy, F. (2010). Credit default swap spreads as viable substitutes for credit ratings. University of Pennsylvania Law Review, 158(7),
2085-2123.
Fons, J., Cantor, R., & Mahoney, C. (2002). Understanding Moody’s corporate bond ratings and rating process. Special Comment. Moody’s Investors Services.
Forte, S. & Pena, J.I. (2009). Credit spreads: an empirical analysis on the informational content of stocks, bonds, and CDS. Journal of Banking and Finance, 33(11),
2013-2025.
Fuller, K.P., Yildiz, S. & Uymaz, Y. (2018). Credit default swaps and firms’ financing policies. Journal of Corporate Finance, 48, 34-48.
Giglio, S. (2016). Credit default swap spreads and systemic financial risk. European
Systematic Risk Board, Working Paper (15).
Graham, J. R. & Harvey, C. R. (2001). The theory and practice of corporate finance: Evidence from the field. Journal of Financial Economics, 60(2), 187-243.
Hilscher, J. (2013). Credit ratings and credit risk: Is one enough? Said Business School,
Working Paper.
Hovakimian, A., Kayhan, A. & Titman, S. (2009). Credit rating targets. Available at SSRN:
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=109835.
Hovakimian, A., Opler, T., & Titman, S. (2001). The debt-equity choice. Journal of
Financial and Quantitative Analysis, 36(1), 1-24.
Huang, Y. & Shen, C. (2015). Cross-country variations in capital structure adjustment-The role of credit ratings. International Review of Economics and Finance, 39, 277-294. Hull, J, C., Predescu, M. & White, Al. (2004). The Relationship Between Credit Default
Swap Spreads, Bond Yields, and Credit Rating Announcements. Journal of Banking &
Finance, 28(11), 2789-2811.
Judeg, A. & Korzhenitskaya, A. (2012). Credit market conditions and the impact of access to the public debt market on corporate leverage. International Review of Financial
Analysis, 25, 28-63.
Kiesel, F. (2016). Do investors still rely on credit rating agencies? Evidence from the financial crisis. The Journal of Fixed Income, 25(4), 20-31.
Kiesel, F. & Spohnholtz, J. (2017). CDS spreads as an independent measure of credit risk.
The Journal of Risk Finance, 18(2), 122-144.
Kiff, J., Jennifer, E., Elias, K., Joda, S. & Carolyne, S. (2009). Credit Derivatives: Systemic Risks and Policy Options. International Monetary Fund, Working Paper (254).
Kisgen, D. J. (2006). Credit ratings and capital structure. The Journal of Finance, 61(3), 1035–1072.
Kisgen, D. J. (2009). Do firms target credit ratings or leverage levels? Journal of Financial
Longstaff, F.A., Mithal, S. & Neis, E. (2005). Corporate yield spreads: default risk or liquidity? New evidence from the credit default swap market. The Journal of Finance, 60(5), 2213-2253.
Myers, S. C. & Majluf, N. S. (1984). Corporate financing and investment decisions when firms have information that investor do not have. Journal of Financial Economics, 13(2), 187-221.
Sajjad, F. & Zakaria, M. (2018). Credit Rating as a Mechanism for Capital Structure Optimization: Empirical Evidence from Panel Data Analysis. International Journal of
Financial Studies, 6(1), 13.
Saretto, A. & Tookes, H. E. (2013). Corporate leverage, debt maturity, and credit supply: the role of credit default swaps. The Review of Financial Studies, 26(5), 1190-1247. Standard & Poor. (n.d). Understanding Ratings. Retrieved from:
https://www.spratings.com/en_US/understanding-ratings
Strier, F. (2008). Rating the Raters: Conflicts of Interest in the Credit Rating Firms.
Business and Society Review, 113(4), 533-553.
Sufi, A. (2009). The real effects of debt certification: Evidence from the introduction of bank loan ratings. Review of Financial Studies, 22(4), 1659-1691.
Viral, V. A. & Timothy, C. J. (2007). Insider Trading in Credit Derivatives. Journal of
Financial Economics, 84(1), 110-141.
Weistroffer, C. (2009). Credit default swaps: Heading towards a more stable system.
Deutsche Bank Research.
White, L. (2010). Markets: The credit rating agencies. The Journal of Economic
Perspectives, 24(2), 211–226.
Zhang, G. & Zhang, S. (2013). Information efficiency of the U.S credit swap default