Credit Ratings and Financial Ratio’s as Determinants of
the Corporate-Treasury Yield Spread
An analysis of credit risk through the last decade
Stefanus Nicolaas Nijssen July 2011
University of Groningen International Business & Management Specialization International Financial Management
Credit Ratings and Financial Ratio’s as Determinants of
the Corporate-Treasury Yield Spread
An analysis of credit risk through the last decade
by
Stefanus Nicolaas Nijssen1
Dr. H. Vrolijk Supervisor
University of Groningen International Business & Management Specialization International Financial Management
University of Uppsala Economics & Business
ABSTRACT
This paper explores the determination of credit risk, or more specifically, creditworthiness of corporate bond issues, and their relation to credit ratings during the last decade. In doing so, the answer is sought to the question of whether investors in the bond market have changed their focus towards Long Term Issue Credit Ratings (LTRs) and publicly available accounting information in the determination of creditworthiness of bond issues due to failing predictions of rating agencies. Using ordinary least squares (OLS) it is confirmed that credit ratings are an important factor in the determination of creditworthiness, which is not expected to be changed soon since regulations hold the position of rating agencies firm in place. However, investors do look further than credit ratings and it is found that at least for non-industrial issues ratio analysis is getting increasingly important. Furthermore, it is found that ratio analysis, opposed to the usage of credit ratings, is negatively related to the conjunctural cycle. This underlines investors’ awareness of the time danger posed by the time lag included in credit ratings, which is especially costly during economic turbulent times. It is stressed that credit ratings should not be seen as a definition of creditworthiness, rather a tool aiding the determination of such, next to other information sources.
Key words
Determination of creditworthiness, corporate-Treasury yield spread, default risk, Long Term Issue Credit Ratings (LTRs), accounting ratio’s, corporate bond market
ACKNOWLEDGEMENTS
Table of Contents
CHAPTER 1 INTRODUCTION 5
CHAPTER 2 LITERATURE REVIEW 7
2.1. Background 7
2.1.1. The Importance of Creditworthiness 7
2.1.2. The Development of Indicators of Creditworthiness 7
2.1.3. Bonds 9
2.2. The Risk Assessment of a Bond Issue 10
2.2.1. Bond Yield and Credit Spread 10
2.2.2. Creditworthiness and Credit Ratings 11
2.2.3. Creditworthiness, Credit Ratings and Publicly Available Accounting Information 13
2.3. Research Design 15
CHAPTER 3 METHOD AND DATA 17
3.1. The Model 17
3.2. Data 18
3.2.1. Corporate-Treasury Yield Spread 19
3.2.2. Credit Ratings 19
3.2.3. Accounting Ratio’s 19
3.2.4. Characteristics other than Accounting Data 21
CHAPTER 4 FINDINGS 22
4.1. Results and Discussion 22
4.1.1. Credit Ratings 22
4.1.2. Accounting Ratio’s 23
4.1.3. The Effect of Credit Ratings and Accounting Ratio’s 23
1. INTRODUCTION
In the most general sense of the word, creditworthiness is the ability of an individual or organization to meet its debt obligations. In evaluating a persons’ or an organizations’ ability to meet its debt obligations, risk assessment is a valuable skill to master, which, if understood poorly, can result in bad decisions with costly results. The process of determining creditworthiness is difficult and relevant in all loan segments of the entire financial market. Depending on the branch, evaluations will be different, with different characteristics to take into account to determine the risk factor in the particular situation. This paper will focus on the determination of creditworthiness in the bond market.
According to modern portfolio theory, investors able to choose between two assets with the same return but different risk will choose the less risky one, since they are assumed risk averse. As a result, investors will want to be compensated (demand a higher return) for increased risk (West, 1977). It follows that the costs that come with a loan are related to the loan issuer’s creditworthiness. If an issuer is considered high risk (bad creditworthiness), the issuer reasonably has to pay more for the loan compared to an organization subject to low risk. This is due to the potential costs (losses) for the investors that come with the deal. Various institutions assist, for regulative, protective or other purposes, investors or lenders in the determination of creditworthiness of an issuer. Rating agencies (Standard&Poor’s, Moody’s, Fitch) for instance, create scores for creditworthiness of governments, companies or particular debt issues. These agencies, among others, provide a manner of identification of credit quality to assist investors in their risk assessment.
There are a few different ways companies get ratings. Ratings requested by companies are so-called solicited ratings, at which the company will open the doors for a thorough investigation. When investors, or any other party, let the rating agency believe there is a need for a rating of a certain company, non-solicited, or involuntary ratings can be issued. Non-solicited ratings are often based on a less ‘in-house’ look, since the to be rated company does not cooperate (Poon, 2003). Although most companies are rated, some are not. An example of which is Heineken, who claims to be non-rated simply because they believe to cope better without, having no trouble issuing debt at competitive prices.
This study will evolve for a great deal around credit ratings and the role of rating agencies in financial markets, the usage of ratings in the determination of creditworthiness in the bond market to be precise. Long term issue credit ratings, as a measure of credit quality of a bond issue, are believed to be a basic benchmark tool on which bond investors base their pricing decisions for a great deal. In 1999 Partnoy makes this point by saying “Ratings are mnemonics for credit quality, simply put: AAA-rated companies don’t default.” And even though many investors base their opinion on multiple factors, some solely rely on rating agencies to do the job for them (House, 1995). The bond market, focussed around ratings, is supposedly structured; companies with a higher rating know they can issue their debt at lower costs compared to lower rated opponents. It can therefore be beneficial for companies to request a rating, since the rating provides them a benchmark within their market, a sign to investors as where to place the company.
For years, AAA rated companies issued the cheapest debt, with a clear line to more expensive debt as the rating declined. However, the early 2000s showed the first disruptions in this system. Accounting scandals and fraud, as was the case at Enron’s collapse, caused numerous high rated companies to fail.2 These investments, classified as being as
creditworthy as it could be, turned out to be not that safe at all. Turbulent financial times have caused the clear structure of the market to be disturbed. As the usage of ratings heavily depends on the trust of companies and investors granted to the agencies, failure in providing an accurate indication of creditworthiness can be disastrous to their image. Issuers of debt have proclaimed not to be clear about the state of the financial markets anymore, since multiple high rated (mainly financial) corporations have collapsed. Investors seem to prefer fixed-asset backed businesses these years, even though
general industrials are rated lower than their financial counterparts. It can be doubted whether companies are still (and should still be) willing to pay for rating services, since they seem to provide misjudgments at the expense of companies. Would not it be safer to provide the information to the market directly, opposed to paying an intermediary to perform a misconduct and end up with an expensive debt?
These developments make this subject even more interesting to look at. This paper takes a look at the evolvement of credit ratings in the bond market throughout the last decade. Reliance on ratings might have diminished compared to the time before the last financial crisis. The question is asked whether investors have shifted their reliance on credit ratings towards other measures of displaying creditworthiness, like publicly available accounting information.
Based on the above stated, the research question this study is determined to answer is formulated as follows: Have investors changed their reliance on credit ratings in the determination of creditworthiness of corporate bonds?
The current study will aim at identifying both the importance of Long Term Issue Credit Ratings (abbreviated: LTRs, from now on referred to as credit ratings) and publicly available accounting information on the determination of creditworthiness by investors in the bond market. It is believed that the role of rating agencies has diminished in recent years, although, due to the usage of ratings in regulations, their usage will never be terminated. The effect ratings have in financial markets before and after the crisis will be analyzed.3 It will be done so relating the credit spread as a risk
factor in bond yields to measures of creditworthiness as the credit rating and various financial accounting variables. As to identify differences in characteristics of creditworthiness both on a geographical and an industry level, this will be done for both the US and Europe, covering three industries; Financials, Industrials and Utilities. Furthermore aim is at identifying which characteristics were important before the crisis and are important nowadays, in determining creditworthiness.
The results of the study will contribute to the understanding of the determination process of creditworthiness done by investors, identifying significant determinants of creditworthiness and the necessity of inclusion of credit ratings as such. Furthermore, the results will provide insights in a highly turbulent and changing bond market, comparing the significant determinants of creditworthiness of today’s market with those almost a decade ago. This will provide financial managers with knowledge of the changing environment they find themselves in and will help them in the process of new debt issues.
Outline
Chapter 2 will provide an introduction in the relevant concepts and a thorough investigation of the literature and past research available on this topic will be presented. Chapter 3 introduces the model, after which the data will be introduced. In chapter 4 the hypotheses will be tested and the data and findings will be discussed in detail. Chapter 5 evolves around summarizing the conclusions.
2. LITERATURE REVIEW
This section consists of three main parts. First, the background section will explain the conceptuality of creditworthiness (2.1.1.), the history of determination of creditworthiness in the bond market, and the main indicators used in practice to determine creditworthiness (2.1.2.), and a brief introduction into bond pricing (2.1.3.). Second, hypotheses will be formulated, backed with existing theories build in past research (2.2.). And third, the research design section (2.3.) will elaborate on the research type as well as the research outline.
2.1. Background
2.1.1. The Importance of Creditworthiness
Creditworthiness, to be likely to and able of fulfilling ones debt obligations (Sinclair, 2005), is widely used in risk assessments by practitioners in financial markets. Lending and borrowing funds is a common practice on an individual, corporate, governmental and institutional level, making risk assessments and uncertainty determination an important issue in the banking environment (McNamara and Bromiley, 1997). Strong growth in this sector has led many to question the magnitude and impact of the market’s size and their dependence on the financial market. Interconnectedness due to internationalization of industries and companies make economic downturn and financial distress of market leaders a global issue, wherefore the determination of creditworthiness as a manner of risk assessment is even more important.
What is needed for a proper determination of creditworthiness is reliable and trustworthy information. In determining creditworthiness, information will be gathered to create a picture as complete as necessary, to assess the reliability and trustworthiness of a debtor or corporation. This can be done through several ways, as will be explained. However, the reliability of information provided by companies or institutions is essential in this process. If trust is violated through false statements or fraudulent reporting, the determination of creditworthiness will become very difficult.
Therefore, due to the importance of reliable information in the process of determination of creditworthiness, laws on transparency (IFRS, Sarbanes-Oxley act, SOLVENCY II) have been set in place to assure market and investors’ safety, providing insights in and comparability between financial statements. Professional evaluators of creditworthiness, as credit rating agencies, are forced by law to comply to strict standards of integrity and transparency to ensure their trustworthiness. Despite these imposed laws and regulations, financial markets are prone to failure, and trust in the financial market is a very sensitive issue. Although several regulatory attempts have been made to increase transparency of firms and institutions, defaults have violated investors trust and governments had to intervene in the financial markets in a number of occasions during the past couple of years. The vulnerability of the world’s financial market is great, wherefore proper risk assessments are relevant. The determination of creditworthiness of debt issuers (borrowers) is important in the process of understanding ones risk exposure.
2.1.2. The Development of Indicators of Creditworthiness
After Beavers (1966) study, many failure prediction studies using ratio analysis, would follow. Altman (1968), criticizes past research (including Beaver’s) for using univariate analysis, working with individual ratio’s as predictors in stead of combinations of ratio’s. As an answer to this criticism, Altman (1968) introduced multiple discriminant analysis, able to create a prediction model based on multiple ratio’s simultaneously. This model received great attention under his colleague scholars, and became, after multiple refinements, the basis for credit evaluation and portfolio selection (Moyer, 1977). The basis concept of Altman’s 1968 model is the construction of a cut-off value indicating the turning point between failure and non-failure of firms, the Z-score, based on ratio analysis. He further refines this score into his ZETA-model in 1977, which became widely used in the decades after as an accounting based credit scoring model (Altman, 1997).
Where ratio analysis started to develop as a tool in risk assessment, default prediction and determination of creditworthiness in the early 20th century, the first signs of the credit rating industry started to appear as well. In 1909 John Moody started to interpret data on U.S. railroad bonds and assigned the railroads a rating for their ability to fulfill their financial obligations (West, 1973). During the 1920’s and 1930’s, as rating agencies (Moody’s, Poor’s and Standard4) rated a wider variety of industries and capital markets started to grow, these ratings became popular among
investors in determining creditworthiness, since this saved them the time and the effort performing the analyzation themselves. Usage of ratings grew especially after American financial authorities imposed restrictions in the banking environment after the 1929 financial crisis. In the 1936 Banking Act it was stated that banks were prohibited to invest in speculative funds, meaning banks were not able to invest in funds with a rating lower than a certain level. Later, in the 1950’s the term investment grade bonds became synonym for “bonds having a rating of Baa or higher5” (West, 1973).
These regulations caused the rating agencies to become deeply settled in the financial markets. By the 1970s, as Weinstein (1977) proclaims, practically all bonds had been rated by at least one of the major rating agencies. Through the decades that followed, ratings have grown in popularity as the expansion of financial markets continued. Today most banks and investment agencies are forced to invest in investment grade bonds only, as a manner of risk aversion.
The exact process credit rating agencies use to determine their ratings is not publicly known, though many have tried to estimate this.6 The rating agencies themselves state that the rating is partly based on accounting information (as well
disclosed and publicly available), and partly on in-depth interviews with the company’s managers. The rating agency will therefore both relate the rating to an objective and a subjective assessment.7 Formally, a credit rating is the rating
agency’s opinion on the creditworthiness of the rated company itself or in relation to a particular debt issue.8
In all, the market practitioners are the ones in need of creditworthiness assessments. In order to assess the risk of investing in a certain bond, they can, concluding from the previously stated, do their own research or base their evaluation (partly) on the risk assessment done by a professional agency. In doing their own research, investors are most likely to solely have access to publicly available data, whereas rating agencies have potential access to inside information besides the publicly available data. Logically, for investors to base their assessment on the opinion of rating agencies, trust in and the image of these agencies is very important. Wrong assessments can be disastrous for the agency’s image and reliability.
The last decade has proven disastrous for this trust. Although regulations are tied to, and therefore markets dependent on, credit ratings, the oligopoly of rating agencies has been under heavy attack as multiple assessments have not turned out what they should have been. Federal interventions have tried to restore trust in the rating market, where lack of
4 Moody’s, S&P’s and Fitch are the so called “big three” currently dominating the major part of the credit analysis market.
5 Following Moody’s rating system. This divided the scale of ratings in two distinct groups, investment grade and junk bonds/speculative grade. For a complete picture of all ratings, see Appendix 2.
6 Hwang et al., 2009.
7 Poon, 2003; Hwang et al., 2009.
competition and conflicts of interest keep fueling criticism towards rating agencies. The developments in the rating market have led us to put our focus toward the influence of the crisis in the rating market, asking whether damaged trust in the rating industry has let investors to change their way of determining credit risk.
2.1.3. Bonds
As to clarify some of the concepts used in the following sections, bonds and their pricing process will be briefly summarized.
Bond Price
Through a bond issue a corporation or government is able to attract cash from the market in the form of debt (opposed to equity issues). A bond is issued at a certain par value or face value, the underlying value, on which coupons are paid, indicated by a fixed annual percentage of the par value. The price of the bond is calculated using the required rate of return to obtain the sum of the present value of each coupon payment and the par value at maturity. Since the price is related to the required rate of return, a bond will trade at a discount (current price is lower than the face value) whenever the required rate of return is higher than the coupon payment or at a premium whenever the required rate of return is lower than the coupon payment. However, most bonds will be quoted in yields in stead of prices. The yield is the return an investor will obtain holding the bond until maturity.
See figure 1 for a visualization of the price calculation of bond X. The par value of bond X is € 1000, with a coupon rate of 5% and a maturity of 4 years. Each of the first three years the investor receives a coupon payment of € 50 and at the end of year 4 he will receive both the coupon and the par value of his investment, € 1050. He has a required rate of return of 6%, which is used to calculate his time value of money. The present value of all cash flows will be € 965, the current price of the bond. Bond X is trading at a discount (€ 965 < € 1000, the required 6% is higher than the coupon of 5%. wherefore this investor will value the bond lower than its par value).9
Yield to Maturity
In the example 6% has been used as the investors required rate of return. In practice this is not known and will be different per individual. The yield quoted in the market is the yield the investor will obtain holding the bond until maturity, given the price the bond is trading for, taking into account the time value of money. Thus, the calculation is turned around. In stead of using the required return to calculate the price, the price is used to calculate the yield. This is mathematically done using trial and error, which is rather time consuming. Programs can be used for this. In the case of bond X in
example 1 a yield to maturity of 6% will be found. An investor will decide for himself whether or not to acquire a bond for the price quoted, according to his required return. Therefore, since the yield quoted in the market is the result of supply and demand from many investors, the yield will quote the markets required rate of return of a certain bond.10
Bond Types
The bond market trades a few different type of bonds. Corporations as well as governments, can issue bonds. Most relevant to this research will be the corporate and government bonds. Government bonds, referred to as Treasury bonds,
9 These figures are examples. Real data (par values, coupon payments, payment dates or cycles) may defer. 10 Assuming prefect markets. In Illiquid markets this might not be the case.
are bonds issued by national governments in their local currency. As shall be seen in section 2.2. these bonds are often seen as the least risky bonds, since governments rarely default.11Corporate bonds however, are said to be more risky
and will generally trade at higher yields. In isolating the risk factor of corporate bonds, the corporate-Treasury yield spread will be calculated.12
(a) Sp = Yc - YT
The corporate-Treasury yield spread (Sp) of a certain bond is simply the difference between the yields of a corporate bond (Yc) and that of a Treasury bond (YT).13 The corporate-Treasury yield spread will from now on be referred to as
spread. Shortly, the spread of bond X is the excess yield of bond X over Treasury bond Y.
Now that the main concepts are made clear, the next section will identify the hypotheses relevant to this research, building on past literature.
2.2. The Risk Assessment of a Bond Issue
This section will explore existing literature relevant to the topic, helping to define the hypotheses. Since most studies on the creditworthiness of companies start from an investors point of view, we will do so as well. This will make the research question approachable and easier to investigate. First links between the creditworthiness and the risk factor of the bond spread will be covered (2.2.1.), followed by the relation between credit ratings and their effect on yields and spreads (2.2.2.). The last part will deal with the additional information content investors might consider in determining the credit spread; publicly available accounting information (2.2.3.).
2.2.1. Bond Yield and Credit Spread
Multiple articles have been published determining the composition of bond yields.14 Generalizing the results of these
studies, four main components can be identified, visualized in figure 215: (1) the level of interest rates, (2) bond specific
characteristics (time to maturity, coupon payments, etc.), (3) a capital gains component (asymmetrical tax conditions) and (4) the default risk. Using these four components, results will defer quite significantly in distinguishing between corporate bonds and Treasury bonds. As can be found in Elton et al. (2001), the coupon payments of Treasuries are not taxed. The required rate of return16 is the aftertax return investors will obtain. Hence, in order to obtain the required rate
of return, a tax premium on corporate bonds is required by investors, creating factor (3) for corporate bonds. In Treasury yields such a component will be absent. Furthermore, the default risk (4) of Treasuries is significantly lower than that of a corporate bond.17
Since Treasury bonds, especially those in the U.S. and Europe18, are seen to be the
most secure bond investments available, investors price the risk premium in terms of basis points in excess of Treasury bonds,
11 Later it will be noted that Treasury yields are often referred to as riskless, since the Treasuries referred to are countries that seldom default on their payments. However, it must be said that no bond is riskless. Treasury bills do include a risk factor, think of for instance developing or unstable countries.
12 More on the isolation of the risk factor can be found in section 2.2. For now however, focus is on the calculation of the corporate-Treasury yield spread.
13 For example (see equation a): if the yield of corporate bond X is 7%, the yield of Treasury bond Y is 4%, the spread of bond X is (7-4) 3%. 14 Ingram et al., 1983; Barrett et al., 1986; Lui et al., 1999.
15 It should be stressed that this is the general result in the search for the decomposition of yields, these models are not perfect and not collectively exhaustive. Furthermore, the relative size of the factors in the figure is random.
16 see Bond Price in section 2.1.3.
17 Huang and Huang (2002) even note that Treasuries are non-defaultable. This is however just an assumption, made in most credit spread studies, where Treasury yields are taken as a riskless rate.
18 The financial crisis of countries like Greece, Ireland, Iceland and Spain undermine this statement. A portion of risk will most certainly be included in their Treasury yields.
leading to the credit spread. Calculating the spread, both the corporate bond and the Treasury bond are necessary to be similar in characteristics (2). Furthermore, the level of interest rates (1) will be included in the investors’ calculation of their required rate of return, and hence in the respective corporate and Treasury yields. Therefore, we can say, following past research on the decomposition of bond yields, the main determinants of the spread will be a tax component and a default premium.19
2.2.2. Creditworthiness and Credit Ratings
In the estimation of the determinants of creditworthiness of bonds, the main focus will be towards the risk factor incorporated in the yield. The risk perceived by investors closely relates to what is known about the company (issuer of the bond). The more an investor knows, the lower the risk premium he is to demand. As Duffie and Lando (2001) point out, in a hypothetical world with complete information, the risk factor in yields will be zero for surviving firms (if an investor knows everything about the future of a company, there will be no risk, just certainty). However, since we do not live in a world with complete information, a risk premium will be demanded to compensate for unexpected losses. Keeping this in mind, issuing corporations will be eager to be as transparent as possible, lowering their cost of debt (the default factor, see figure 2, will be smaller). However, at the same time, this transparency can be beneficial to competitors, wherefore companies would be reluctant to reveal information to the public.
As explained in section 2.1.2., rating agencies have positioned themselves in the market between issuing corporations and investors, providing information on the creditworthiness of the companies. Rating agencies generally have access to deeper data sources on a company, giving them the advantage of providing a rating on the company based on data not available to the public.20 This data, encrypted in the rating, communicates the creditworthiness of the
company to the market, without revealing sensitive information to competitors. Hence, credit ratings resolve companies’ dilemma on transparency, providing a way of being open yet closed at the same time.
It might be understood that reliability of ratings and trust in rating agencies is essential in the use of ratings to determine creditworthiness of an investment. Since even regulatory institutions base laws on these ratings, rating agencies are pressured to provide qualitative assessments. Several times rating agencies have however overestimated the creditworthiness of some companies, after which companies with a high rating defaulted. The crucial role of rating agencies is since then under heavy debate.21 As Partnoy (1999) summarizes the discussion, the financial markets have
increased their dependence on the ratings (mainly due to regulations), whereas the ratings have decreased in informative power. Credit ratings are said to be adjusted too slow to reflect new information properly, and have become reactive instead of proactive.
Failing to fulfill in their obligations, rating agencies have lost image and trustworthiness. Partnoy proclaims, already in 1999, that the institutionalization of regulations in the financial markets should be stopped, since rating agencies do not provide proper reflections of credit risk. This has been confirmed by the collapse of several high rated companies during the last crisis. As a substitute of credit ratings, Partnoy (1999) claims credit spreads to be reflecting credit risk in a more accurate way. In example, the German TUI Group, rated B- by S&P’s, is able to place debt at a credit spread of 3%, a level the average BBB rated company would pay. The average B rated issue traded at 6% at that time.22 The other
way around, the Fortis NV debacle is a typical example of a high rated company facing default. Fortis NV was rated Aa by Moody’s, months before it collapsed.
19 A factor often overlooked is the liquidity premium in the yield of bonds. Huang and Huang (2002) state that this is a factor to consider, however, since most of the research evolves around bonds traded on highly liquid markets, it might not be visible. Whenever a bondholder wants to sell his investment in an illiquid market, he might experience difficulties trading for the desired price. Therefore, he might add a premium to his price in order to offset this risk. Since we however will not use illiquid markets in this research, we will ignore this factor.
20 Pettit et al., 2004; Hwang et al., 2009.
21 Partnoy, 1999; Pettit et al., 2004; Altman and Rijken, 2006.
Even though examples of failure can be found, regulations still let rating agencies be as powerful in financial markets as never before. As to illustrate the importance of rating agencies in today’s financial markets, Greece proves to be an interesting example. At the moment of writing, financial institutions as well as governments throughout Europe try to find a solution for the Greek deficit problems. Restructuring Greek debt, relieving Greece from part of its short term debt burden in exchanging the debt for long term issues, has been posed as a possible solution. However, even thought this is intended to aid Greece, governments fear a downgrade of the Greek debt to D (default) by Moody’s. Since over a quarter of Greek debt is in hands of international banks, this would impose large revaluations of their debt holdings, heavily stimulating sales of Greek paper, since Greek debt is not investment worthy after the downgrade. Where the plan is intended to avoid default in the first place, it would involve risking a rating downgrade, bringing Greece even closer to bankruptcy,. This paradox reveals the impact rating agencies have in the market.
Therefore, the importance of ratings is yet not to be underestimated. Whenever a bond upgrades from speculative grade to investment grade, it will be under attention of a significantly broader range of investors than before. The higher the rating the more creditworthy the bonds should be. Furthermore, as Lui et al. (1999) claim, investors use the ratings in their determination of issues’ creditworthiness and bond valuation since this saves them precious time and money in doing this themselves. Many scholars have tried to estimate the relation between the rating and a bond’s price. Empirical evidence has not been found until West’s (1973) study, who claim the relation between bond yields and credit ratings exists due to the authorities using ratings in regulations. Later studies supported this finding.23 Lui et al. (1999)
find ratings to influence prices by comparing prices before and after rating refinements by Moody’s. Ederington et al. (1984) find ratings to be significantly correlated to bond prices in combination with a set of publicly available data. Huang and Huang (2002) find that credit risk does not account for the same portion of credit spread in all rating classes. They find that the higher the rating the less credit risk is accounted for. Supporting this, Elton et al (2001) argue that since investment grade bonds rarely default, expected default will only account for a small part of the credit risk. A visualization of the theoretical relation is given in figure 3.
Previous attempts to assess the effect of ratings on yields or bond prices have been done by Weinstein (1977). Like some of his colleagues (Wansley and Clauretie, 1985), Weinstein (1977) used rating changes to determine the impact of ratings on the respective yield. He compares prices before and after the event of a rating change and found weak support for their statement. Ederington et al. (1984) find credit ratings to be significant in the determination of the risk factor of bond yields, though they found publicly available information to increase the predictive ability of the model. Their model uses not only rating changes but all ratings (new, old and revised). The Markov model, used by Jarrow and Turnbull (1995, 1997), determines the risk
spread in credit, in which credit ratings are used to determine default risk. Elton et al. (2001) find credit ratings to be influencing credit spreads (as indicators of default risk), however only for a minor 20%. Huang and Huang (2002) find similar results. The theory is summarized in table 1 below, and for a complete summary table on the literature, see table 1 in the appendix.
It is widely believed that ratings provide a tool communicating public and non public information, facilitated by specialized companies
23 Ederington et al., 1984; Das and Tuffano, 1995; Lui et al., 1999.
which are able to perform such analyses in a more efficient way than individual investors can (Griffin and Sanvicente, 1982). For these reasons, rating agencies have settled deep in today’s financial markets and ratings are widely accepted. Providing access to sensitive information, encrypted in the rating, and saving investors’ time and costs analyzing publicly available corporate information, one might expect credit ratings to influence the price of a bond. The higher the credit rating, the higher the creditworthiness of the bond, the lower investor’s risk and therefore the lower the price of the bond. Following this train of thought, the following is hypothesized:
H1: As an indicator of creditworthiness, the level of credit ratings negatively influences the corporate-Treasury yield spread.
Table 1: Literature Summary Table 1: Literature Summary Table 1: Literature Summary Table 1: Literature Summary
Author Dependent Independent Result
Das & Tuffano (1995) Corporate
Spreads Default rate (credit rating), recovery rate Provide a model that proves spreads vary even if ratings do not.
Ederington et al. (1984) Bond Yield Credit rating, accounting information
and bond characteristics Both ratings and financial information have an independent impact on yields.
Elton et al. (2001) Corporate
Spreads Systematic risk, unsystematic risk (bond ratings), taxes By adapting factors that explain risk premia of common stock they create a model able to explain 66-85% of corporate spreads.
Huang and Huang (2002)
Corporate Spreads
Credit rating, leverage, equity premium, default probability, asset volatility, asset risk premium
They find a small fraction (20%) of spreads to be caused by ratings for investment grade, a larger fraction for junk bonds.
Lui et al. (1999) Corporate
Spreads Bond characteristics Through comparison of means, they find downward changes in ratings to have a bigger independent impact on spreads than upward changes
Weinstein (1977) Bond Yield Credit ratings Through comparison of means before and after rating change
he finds small support for an independent effect of ratings on yields. It seems however that the market adjusts faster than the rating to change.
West (1973) Bond Yield Credit ratings Through a theoretical discussion he finds that credit ratings
must have an independent impact on yields. 2.2.3. Creditworthiness, Credit Ratings and Publicly Available Accounting Information
Economic crises have had a severe impact on corporate performance, reflected in increased spreads during the last years. The state of the economy seems to impact the yield of bonds. In line with this statement, Duffee (1995) finds support that yield spreads fall as economic growth increases, regardless of the rating category. Wakeman (1981) finds security price changes to be related to rating changes, where the relation seems to be stronger in economic turmoil opposed to times of economic growth, though the causal link is unclear. Liu and Thakor (1984) see that economic conditions affect corporate performance and hence credit ratings. Therefore, the relation between credit ratings and bond yields, if accounted for economic growth, might be biased.
Since credit ratings are said to be long term, they will not, and should not, capture the full magnitude of t h e c o n j u c t u r a l c y c l e . Ratings are not continually revised, as Weinstein (1977) discusses, wherefore ratings will always lag the market. Spreads are more volatile than ratings, hence, ratings do not capture the complete m o v e m e n t o f s p r e a d s , wherefore more factors to be incorporated in the risk
factor of corporate bond yields should be considered. This can be clearly seen in the first graph presented. The spread difference between both lines seems a constant markup for the difference in rating class. However, the spread is volatile through time, indicating other factors influencing the spread. As rating agencies obtain their data from companies, so do investors. Accounting data available to investors might be faster and more accurate in reflecting the current state of creditworthiness than a credit rating does, wherefore it seems not surprising that investors will look at more than just a rating.
Using publicly available financial data to estimate the creditworthiness, read riskiness, of a bond, has often been done creating default prediction models, as explained in section 2.1.2. Altman et al. (1977) use financial statement data and market data to create a value of bankruptcy risk, called the ZETA score. This model is a refinement of their 1968 model, which created a Z-score. In both cases he used multiple discriminant analysis (MDA) to define a set of significant financial ratio’s to determine default risk. As Kaplan and Urwitz (1979) point out, through past studies as well as their own, it is fairly easy to imitate the bond rating process using financial statement data, and to build a bond rating prediction model. Peavy et al. (1983) use ratio analysis to predict credit ratings. Through MDA they are able to classify 5 out of 18 variables to correctly predict 90% of the data into the correct rating group. Lui and Thakor (1984) agree that if a bond rating
affects the yield of a bond, a thorough understanding of the determination of ratings is necessary for proper f i n a n c i a l m a n a g e m e n t . Ederington et al. (1984) test bond yields for determining the information content of the ratings by comparing the relation of the risk factor of yields to that of credit ratings and publicly available data respectively. They test all r a t i n g s ( n e w, o l d a n d
Figure 5: Bond Spread vs. Bond Yield
0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0%
Dec-01 Dec-03 Dec-05 Dec-07 Dec-09
BBB (Spread) BBB (Yield)
Source: Bank of America Merrill Lynch Global Bond Index via Bloomberg’s Database, US issues Figure 4: Bond Spreads
0.0% 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 7.0% 8.0% 9.0%
Dec-01 Dec-03 Dec-05 Dec-07 Dec-09
BBB (Spread) AAA (Spread)
revised) and find both ratings as well as publicly available information to be significant in the determination of credit yields.
It must be recognize that in investors’ determination of creditworthiness more than just historical financial data is involved. A qualitative look at a company, providing a “through the cycle” view towards the future, combined with market and positioning or competitive factors can be incorporated in a proper and complete determination of creditworthiness (Pettit, 2004). It will therefore be reasonable to include both ratings and financial ratio’s in this determination. However, since rating agencies claim that ratio analysis is an important part of their assessment, as well as the fact that ratio analysis provides manageable targets to work toward, in the process of identifying which factors matter most in creditworthiness determination, ratio analysis seems an interesting target to aim at. Furthermore, since ratings lag in responsiveness, investors expectedly look further than credit ratings to base their determination upon.
Therefore, the following is hypothesized:
H2: Investors take publicly available accounting information into account next to credit ratings to determine creditworthiness.
Following the reasoning presented before, where rating agencies have failed to deliver in expectations and face a declining image due to heavy criticism during the last decade, it hypothesized that:
H3: Publicly available accounting information (credit ratings) has increased (decreased) in importance in the determination of creditworthiness of corporate bonds.
Answering the hypotheses, we hope to obtain insights into investors’ determination of creditworthiness. Again, see table 2 for a summary of the theory used in this section, or table 1 in the appendix for a complete summary on the literature. Finalizing hypothesis construction, we will now continue introducing the research design.
Table 2: Literature Summary Table 2: Literature Summary Table 2: Literature Summary Table 2: Literature Summary
Author Dependent Independent Result
Altman et al.
(1968/1977) Z-score 7 Financial accounting ratio’s Use accounting variables to create a failure prediction model (using MDA)
Ederington et al. (1984) Bond Yield Credit rating, accounting information
and bond characteristics Both ratings and financial information have an independent impact on yields.
Kaplan & Urwitz (1979) Credit Ratings 3 Financial accounting ratio’s and bond
characteristics Create a model that predicts 2/3rds of credit ratings correctly.
Peavy et al. (1983) Credit Ratings 5 Financial variables Create a model that correctly assess 90% of the sample
ratings. 2.3. Research Design
Central to this research is creditworthiness in the bond market. The assessment of credit risk and the determination of creditworthiness have been debated through a review of existing literature in the preceding subsection. It is found that creditworthiness determination relies largely on the analyzation of available information, be it directly from financial statements, or indirectly through a pre-cooked rating. From the available information on the company, rating agencies and investors create their opinion on the credit risk involved in investing in the company. The credit risk, directly or through a credit rating, will affect the price of a bond. Figure 6 provides an overview of the theoretical relationships relevant to this research.
available accounting data as determinants of the corporate-T r e a s u r y y i e l d s p r e a d throughout the last decade will be tested. Through qualitative analysis of pricing data on bonds and the underlying corporate
financial figures, the credit spread will be explained. Furthermore this will be done in threefold, using datasets covering data as of 2001, 2006 and 2010 as to capture the possible effects the financial crisis might have had on the determination of the credit spread.
The result of this study will be valuable for financial managers in understanding the process investors, as buyers of initial offering bonds, use in evaluating and pricing the bond to be issued. The financial markets of both Europa and the US will be considered, as well as three different industry types, to generate results as useful and detailed as possible.
Concluding
The research thus far evolves around exploring creditworthiness assessment in the bond market. Reflecting upon recent developments in the financial markets the following question is asked:
Have investors changed their reliance on credit ratings in the determination of creditworthiness of corporate bonds?
And, in sum, after explaining the main concepts of the bond market and credit ratings, and through an extensive review of the existing literature on the topic, the following hypotheses are formulated:
H1: As an indicator of creditworthiness, the level of credit ratings negatively influences the corporate-Treasury yield spread.
H2: Investors take publicly available accounting information into account next to credit ratings to determine creditworthiness.
H3: Publicly available accounting information (credit ratings) has increased (decreased) in importance in the determination of creditworthiness of corporate bonds.
The proceeding section, the methodology section, will deal with operationalization of the hypotheses and the model in which they will be tested.
3. METHOD AND DATA
The methodology section will elaborate on the operationalization of the hypotheses, aiming at answering the research question. At first, the model and the equations that will help answering the hypotheses will be explained (3.1). The data used for testing will be introduced in section 3.2., with the spread (3.2.1.), credit ratings (3.2.2.), accounting ratio’s (3.2.3.) and characteristics other than accounting data (3.2.4.) explained.
3.1. The Model
As has been explained in section 2, and confirmed by previous research, the risk factor in relation to a bond investment, as an indicator of creditworthiness, can be found in the spread. Therefore, bond spreads will be regressed to both Moody’s long term issue bond ratings, in dummy format, and to a set of publicly available accounting variables. The following relationship is estimated, based on the research method of Ederington et al. (1984)24:
Equation (1): Si = f ( Ci , Ri , Xi , ei )
Where:
Si is the corporate-Treasury interest rate spread of issue i;
Ci is a set of publicly available accounting variables pertaining to the issuing firm of issue i;
Ri is Moody’s rating of issue i;
Xi is a set of characteristics of issue i other than its creditworthiness;
εi is the random error term of issue i.
The first hypothesis, that credit ratings provide investors with an information content extensive enough for investors to rely their determination of creditworthiness upon, implies that the rating of bonds negatively relates to the yield spread. In simple formula format this gives us a function where Ci can be suppressed, as Ri provides the accounting
information to investors:
Equation (2): Si = f ( Ri , Xi , εi )
In order to test the independent impact of publicly available accounting information on credit spreads, the rating (Ri)
in the estimated equation will be suppressed, leaving: Equation (3): Si = f ( Ci , Xi , εi )
Assuming that investors base their evaluation of creditworthiness on more than credit ratings or publicly available accounting information alone, both credit ratings and accounting information will have to be included. Both credit ratings and publicly available accounting information will be incorporated and the complete function will be used to test for hypothesis 2:
Equation (1): Si = f ( Ci , Ri, Xi , εi )
3.2. Data
As to test the model, we need information on issues (corporate-Treasury yield spreads (3.2.1.) as well as the ratings on these issues (3.2.2.)), issuers (corporate information (publicly available accounting data (3.2.3.)) and other factors influencing yields (3.2.4.).
Data on bond prices and information will be gathered from both Bloomberg’s Corporate Bonds Database. Data on corporate financial data will be extracted from Datastream. Poon (2003) uses Datastream to extract accounting information as to create a model able to predict rating scores. Bloomberg’s Corporate Bonds Database is a widely used information tool among banks and investors, quoting real time and historical bond prices, providing analyzation tools and news feeds. Furthermore, it is used as an academic data source at multiple universities.25 Bloomberg’s Database,
available at the Corus Pension Fund will provide pricing data on fixed income debt securities, corporate and Treasury yields to our particular interest. Through Datastream the underlying corporate accounting data will be obtained. Three datasets will be created, for the years (1) 2001, (2) 2006 and (3) 2010.
The initial dataset consisted of 2480 bond issues available through Merrill Lynch at the end of 2010, 1810 and 2060 for the year 2001 and 2006 respectively. Merrill Lynch provides an index of bonds, available through Bloomberg’s database. Securities with a minimum outstanding amount of USD100 million, a fixed coupon schedule and at least 1 year to maturity qualify for inclusion, as stated by Merrill Lynch. The index makes a clear distinction between industries, region and rating category and is easily requested for different points in time.
For each time period, a few selections have been made. Only three sectors (industrials, financials and utilities) from the regions EMU, UK and US have been extracted. No more than two issues for each company have been included.26
Furthermore, only bonds up until category B are included with a minimum maturity of five years.27 Summary statistics
can be found in the appendix, table 3 and 4.
Due to connexion problems concerning the ISIN codes of the bond identifiers, it was not possible to create the proper link between Bloomberg and Datastream to retrieve the financial accounting data from the respective parent companies automatically. Alternative datasources proved to be unaccessible, wherefore manual matching seemed to be the best option. However, since matching all bonds manually would be rather time consuming, it has been decided to reduce the dataset at random to a manageable number of bonds.28 Looking at multiple papers from comparable research, a dataset
consisting of approximately 150 issues is reasonable to work with.29 Since the reduced dataset should ideally be able to
represent a basket of bonds conform the true state of the market, the relative number of bonds in each rating category, region and industry have not been changed. A reduced dataset of 250 bond issues is finally chosen for the each dataset,
25 Corporate bond prices and ratings in Bloomberg’s Database are found by Altman and Rishore (1998) and Amato and Remolona (2003). Altman (1989) and Driessen (2005) use Bank of America Merrill Lynch Global Bond Index, through Bloomberg’s Database. Universities like University of
Washington, University of Pittsburg, Erasmus University of Rotterdam provide access to academics.
26 Inclusion of no more than two issues of the same issuer prevents certain companies to be overweighted in the dataset (Ederington et al., 1984). By assigning a random number to each issue using excel’s RAND() function, a selection at random could be made. Sorting first by issues and second by random value the largest two have been included. This might as well have been the lowest or any combination of two.
27The data sample has been restricted to issues with at least five years to maturity (Ederington et al, 1984). As for industry choice, even though most authors do not specify their choice of industries openly, they do recognize the importance of industry differences. Peavy (1983) as well as Weinstein (1977) use a twofold of sectors, Utilities and Industrials, whereas Campbell and Taksler (2003) also include the Financial sector. As the financial sector has been heavily attacked during the financial crises, some interesting findings might result the inclusion of this sector, wherefore the choice is to include these three sectors in our sample. Furthermore, only B and BB rated speculative bonds were included, due to unavailability of data, ratings lower than B have been excluded.
28 Using the summary statistics tables as presented in the appendix, table 3, percentages of total have been taken for each subgroup (in example: for the 2010 dataset, BBB US Industrial were 599 from 2480 total, representing 24.2% of the set). The desired dataset being 250 issues, 24.2% of which should be BBB US Industrial, 60.3 issues, rounded to 61. Each issue was given a random number using the Excel function “=RAND()”, which assigns a random number to each cell. Sorting for these random numbers the highest 61 have been included in the reduced dataset (using the highest selection was chosen arbitrarily, it might as well have been the lowest or the middle). Table 6, 7 and 8 in the appendix present the summary of spreads of the reduced dataset.
for which the bond ISIN codes are matched to the (equity)share-ISIN code of the company manually. Several missing values and non existing ISIN codes in Datastream led to a further reduction from the desired 250. The final sample consists of 143 issues in 2010, 163 and 140 in 2001 and 2006 respectively. Specifics can be found in table 7, 8 and 9 in the appendix.
3.2.1. Corporate-Treasury Yield Spread
Bloomberg will provide market data on corporate bonds. The yield (average weekly for the year 2010), the rating as of end 2010, industry (industrial, financial or utility), region (EU, UK or US), ISIN-code and bond characteristics will be extracted. The same will be done to create a separate dataset as of 2001 and 2006. As to calculate the corporate-Treasury spread, the option adjusted spread (OAS) will be used30, correcting for spread differences due to inclusion of
options in bonds.31
3.2.2. Credit Ratings
The summary tables (tables 3, 4 and 5 in the appendix) give an indication of the relation of credit ratings on yield spreads. The expectation of higher yields for issues with a lower ratings seems to be true, since, generally, the yield spread increases when the ratings worsen. This seems consistent across industries and regions, as well as through time. The overall level of spreads is highest in the 2001 data, lowest in the 2006 and mediocre in the 2010 data. The level of spreads, representing a certain level of risk, seems to follow a conjunctural wave. In economic turbulent times, the spread is expected to be larger, since the risk (of default) is higher, opposed to economic sound times, where risk, and corresponding the credit spread, is lower. The data from 2006, presenting low spreads, seems to reflect the best economic condition relative to the other two time periods. Both graphs in section 2 also indicate this. Furthermore, where in 2001 industrials paid a larger premium over treasuries compared to financials, in 2010 this gap had apparently gotten smaller. This does not hold clearly for the utility sector.
The credit ratings in factor Ri, as provided by Moody’s, in equation (1) will be presented by dummy variables for
each rating category. Five dummy variables will be included: AA, A, BBB, BB and B. AAA will be omitted.32 The
equation for Ri will be as follows;
Ri = α1 AAj + α2 Aj + α3 BBBj + α4 BBj + α5 Bj
Where AAj equals 1 if issue j is rated AA and zero if otherwise.
3.2.3. Accounting Ratio’s
Ideally, one should include all information available an investor can use to determine creditworthiness, taken aside credit ratings. However, since this can only be produced at high cost, a selection will have to be made. Seeking to find whether or not credit ratings have been losing importance in investors’ determination of corporate creditworthiness, a variable set alternating the credit score has to be found. The set of variables considered in this research will be chosen
30 The dependent has been tested for normality. Using the stata function qnorm (oas) plots the residuals of this variable to the theoretical normal distribution. For the datasets 2010 and 2006 no disturbing deviations of normality can be detected. For 2001 a slightly upward deviation is present. Using the log of the dependent variable would have solved this problem. However, since no theory found does so, and the disturbance is only minor in the 2001 dataset, even absent in the 2006 and 2010 dataset, it has been chosen not to use the dependent variable in its original form. Further variables have not been tested for normality.
31 The option adjusted spread (OAS), also used by Amato and Remolona (2003), is provided by Merill Lynch’s bond index. Where the regular corporate-Treasury yield spread would simply reflect the difference between both yields, the OAS corrects for the spread due to optionality. An investor holding a callable bond faces the risk of being called pre-maturely. In order to compensate for this risk he or she will raise the required yield on the investment. The OAS is the spread corrected for these options, reflecting the corporate-Treasury yield spread free of options.
due to: (1) popularity in theory33, (2) potential
relevance to this study and (3) availability and accessibility. Past theory generally uses financial ratios which can be assigned to one of the following groups: (1) liquidity, (2) profitability, and (3) solvency34. No clear consistency through literature
can be found on a complete list of ratios, although most research operates in the same manner. First a list of variables is introduced, after which the number of variables is reduced to a manageable and statistically sound set. The variables available trough Datastream are presented in table 3.
From the variables presented in table 3 ratio’s usable for inclusion in our analysis are generated, see table 4. Summary statistics on the ratio’s used can be found in table 9 in the appendix. Following S&P’s35
methodology, three year averages are used. Hence, spreads as of december 2010 are explained using financial data from 2008, 2009 and 2010.
Liquidity ratios measure the firms ability to cover its short term obligations, without having to borrow additional capital. The more cash is readily available for operations, or upcoming liabilities, the more liquid the firm. The current ratio measures the short term availability of funds, over the firm’s short term obligations. Excess current assets result in a positive working capital, cash available for growth of operations. Generally, a positive working capital is a positive sign, since there exists the ability to grow. However, a positive working capital could also be seen as declining operations, where negative working capital indicates that the company is in fast growth.
Solvency ratios measure the ability of the firm to cover its long term obligations, through which bankruptcy can be avoided and can growth be continued. In essence solvency focusses on continuation of the firm, whereas liquidity focusses on continuation of operations. Apart from the long term view of these ratios, the financial structure of the firm is also considered important in evaluating solvency. Whether the company relies heavily on debt or whether the firm is able to cover its fixed debt and interest payments is essential for its continuation.
The functional form of the 11 financial variables
33 Note that the financial ratios upon which investors place much reliance will be the ones management will tend to manipulate most, called window-dressing.
34 Altman et al, 2000.
35 Poon, 2003, using S&P’s 2000b, Corporate Ratings Criteria. It seems more obvious to choose for Moody’s methodology since this research is based on Moody’s ratings. However, since these companies are not openly discussing their procedures in creating ratings, these methodologies are difficult to find. Since the ratings of different agencies are equal in most cases, and rating events follow one another in a matter of months, it is assumed that the methodology of either agency is approximately equal. Since the part of the rating process referred to in this particular case is the time span used for averaging the ratio’s, no major differences are assumed to exist.
Table 3: List of financial variables used to create financial ratio's Table 3: List of financial variables used to create financial ratio's Table 3: List of financial variables used to create financial ratio's
1 INTCOV Interest charge coverage
2 TOTASS Total assets
3 TOTDEB Total debt
4 TOTLIAB Total liabilities
5 TANASS Net tangible assets
6 TOTCAP Total capital or equity
7 EBIT Earnings before interest and taxes
8 EBITDA Earnings before interest and taxes, depreciation
and amortization
9 SALES Sales (revenue)
10 NETDEB Net debt
11 LTDEB Long term debt
12 FCASHFL Free cash flow 13 WORKCAP Working capital
14 MARGIN Net profit (margin)
15 CAPEX Capital expenditures
16 OPEREX Operating expenses
17 CURLIAB Current liabilities
18 CURASS Current assets
19 DtoE Total debt as percentage of common equity
Table 4: Publicly available financial ratios. Table 4: Publicly available financial ratios. Table 4: Publicly available financial ratios. Table 4: Publicly available financial ratios.
Liquidity
1 CUR Current Ratio Current assets / current liabilities
2 WOR Working Capital
(per 10.000.000) Current assets - current liabilities
Profitability
3 MAR Margin ( / 100)
4 OPER Operating Profit EBIT / sales
5 ROA ROA EBIT / total assets
6 ROE ROE EBIT / total equity
Solvency
7 EQU Equity/Liab Total capital / total liabilities
8 QRAT Quick ratio Total assets / total liabilities
9 DA D/A Debt / assets
10 DE D/E Debt / equity
can significantly influence the results of the model. Although the direction of the relation can be detected quite easily, its functional form proves more difficult. Each variable has been performed into four different power functions [0.5, 1, 2, 3].36 From the resulting 44 variables, a selection best representing the credit rating process, or the best alternative for
investors to turn to next to credit ratings, has to be made. It will be most preferable to include a single ratio from each of the three categories (liquidity, solvency, profitability), minimizing possibility of multicollinearity, resulting in three ratio’s in the model. Since presumably rating agencies do predict creditworthiness at best in the 2001 dataset, this dataset has been used to find the best set of accounting ratios replicating credit ratings.37 Taking the credit rating as the
dependent, the current ratio, margin and debt to assets, all three in linear form, are found to best fit the credit rating.38
An ordered probit regression has been used to test the inclusion of these variables, from which Ci in equation (1) has
been derived:
Ci = β0 + β1 (CUR)j + β2 (MAR)j + β4 (DA)j
Since the set of accounting ratio’s has been reduced using data from year 2001, estimating a best fit in this year specifically, it will expectedly follow that the years 2006 and 2010 will not fit equally well. One would expect model fit to decrease for upcoming years. Testing for intercorrelation between the independent variables in Ci no disturbing
correlations have been found, see table 11 in the appendix. 3.2.4. Characteristics other than Accounting Data
Ever since ratio analysis existed, industry effects have been recognized in the literature (Beaver, 1966). Different industries might face different risks and different measures of risk. A certain margin in the financial sector might prove devastating in the utilities sector, whereas an average level of fixed assets in the industrial sector is extremely large in the financial sector (Longstaff and Schwartz, 1995). Furthermore, as the summary statistics indicate, spreads differ across sectors as well. As for the 2010 dataset, spreads on BBB rated financials are 92 points higher compared to industrial issues within the same rating class. We therefore correct for industry differences by including an industry dummy. For the very same reason, a country dummy will be included.
Representing the characteristics other than creditworthiness, Xi will include the dummy variables for each region
(EMU, and US, where the UK is omitted) and for each industry (Industrial and Financial, where Utility is omitted); Xi = σ1 (EMU)j + σ2 (UK)j + σ4 (IND)j + σ5 (FIN)j
The overall (equation (1)) model will be the following;
Si = β0 + β1 (MAR)j + β2 (CUR)j + β4 (DA)j + α1 AAj + α2 Aj + α3 BBBj + α4 BBj + α5 Bj + σ1
(EMU)j + σ2 (UK)j + σ4 (IND)j + σ5 (FIN)j + εi
This model will be testable using ordinary least squares (OLS), which will be dealt with in the following chapter.
36 Ederington et al. (1984) perform a power transformation to their variables. This is a statistical method to estimate the power of each single variable. The power found for each variable minimizes the specification error. Therefore, the best fit can be chosen. However, since this is a rather complex method it has been advised to avoid this test and estimate using trial and error. The special cases square root (power 0.5), linear (power 1), square (power 2) and cubic (power 3) are used to find the best fit.
37 2001 has been chosen as a benchmark since for theoretical purposes this year is supposedly the year (from our three datasets) in which rating agencies were most credible. Taking a set of ratios representing the ratings, we are able to best compare the difference before and after the financial crisis.
