____________________________________________________
The Relationship Between Corporate Governance and Firms’ Credit Ratings
An Investigation Through Time
Amsterdam Business School University of Amsterdam
Master’s Thesis submitted by: Charles Bennett charlie.bennett@student.uva.nl
Student ID: 13330306
Degree Programme: Master of Science in Finance Track/ Specialisation: Corporate Finance
Supervisor: Dr. Tolga Caskurlu Amsterdam, January 2022
Statement of Originality
This document is written by student Charles Bennett who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are 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
In this paper, we investigate the impact of corporate governance on firms’ credit ratings and shed light on whether the magnitude of the impact has changed over time. We use a new corporate governance summary measure and apply it to this area of study for the first time. Ordered logistic regression analysis is applied to analyse a panel data set of 1,283 companies within a sample spanning 23 years of time. Using a sample of firms in North America, we strengthen previous claims by finding that corporate governance has a positive and significant relationship with firms’ credit ratings. However, contrary to theory and the expectations of Standard & Poor’s (2004), we show that this relationship has no significant change over time. Further, there is also no evidence found of a temporary adjustment in the relationship due to the Great Recession.
Contents
1 Introduction ... 5
2 Literature Review and Hypotheses ... 7
2.1 Determinants of Corporate Credit Ratings ... 9
2.2 The Change of the Effect of Corporate Governance ... 11
2.3 Hypotheses Development ... 12
3 Data Description ... 14
4 Methodology ... 18
5 Results ... 24
5.1 Discussion of Results ... 28
6 Robustness Analysis ... 29
7 Conclusion ... 36
References ... 39
Appendix ... 43
1 Introduction
Corporate governance mechanisms have been convincingly shown in the past to impact credit ratings and bond yields. According to Standard & Poor’s (S&P) Governance Services (2004), the governance of a firm can be a direct or indirect cause of weakness in a company’s financial ability. The assessment of a firm’s strength (or weakness) in this area by S&P can be used to appropriately impact the awarded credit rating. Although it is one of the newer elements introduced into the process of evaluating a firm’s credit ratings, having generated much attention following the Asian Economic Crisis in 1997 (Mohamad, 2004), it has come under increasing importance since. S&P (2004, p.9) since expected “the links between corporate governance and credit ratings to evolve as more research and case studies bring new issues to light”. This study investigates the effect of corporate governance on corporate credit ratings in the US and is the first to investigate the expectation of S&P for the link between the two to evolve over time. Within our analysis through time, we additionally show whether the Great Recession impacted this link, as well as credit ratings themselves directly.
A majorly utilised strategy of acquiring long-term capital in the United States is via raising debt and investing it appropriately (Bhojrar and Sengupta, 2003). Credit ratings are used by firms to assess the risk related to issuing debt to a specific firm, and firms with worser credit ratings have more costly debt – for example due to higher interest expenses – than an equivalent firm with relatively better credit ratings (Kisgen, 2006).
Therefore, factors that influence the credit ratings of firms can have significant impacts on the performance of a firm. Their importance, as well as the importance of fully understanding them by a multitude of financial and economic agents cannot be understated. Due to the evolving and potentially strengthening link between corporate governance and credit ratings, it is a crucial determinant to understand deeply by investors, analysts, researchers, and corporate management agents.
Although mechanisms of corporate governance and their aims, along with some relations to corporate credit ratings have been widely researched, there still exist important areas surrounding the topic which have not yet been explored. Importantly, prominent papers aiming to examine the link between corporate governance and credit ratings, such as that done by Ashbaugh-Skaife et al. (2006), have mostly delivered a single coefficient derived
from the respective datasets used in order to explain it. Although this is valuable research, and together with multiple other researchers running similar models on datasets of different years and countries we are enabled to construct a bigger picture about how corporate governance effects firm credit ratings; we are unable to see how this effect has changed over differing periods of times or firm characteristics. More recently, the study by Alali et al. (2012) conducted new research, splitting up firms into certain size categories, and assessing the differences in effects corporate governance had on the respective categories of firm size. From this new research, managers and investors alike can use it to their advantage by prioritising corporate governance strengths in smaller firms relative to larger ones. The correct prioritisation of the determinants of credit ratings, such as corporate governance, can allow firms to achieve higher ratings, attain more efficient capital structures (Kisgen, 2006), and uncapped further potential benefits.
The present research adds significantly to the existing literature and aids Alali et al. (2012) in their study by exploring another area where we may notice trends in this relationship, specifically, over time. Due to the ever-increasing relevance and implications of corporate governance, this study aims to firstly analyse past research and authors’ claims about its links with credit ratings by applying similar models and analysing outcomes on a more comprehensive dataset than those previously used. Secondly, the next aim is to deliver fresh results as to whether there exists a trend in this relationship over time, or whether significant macroeconomic events (specifically, the Great Recession) caused a temporary change in this relationship. Delivering results as to whether there are trends can assist numerous types of financial and managerial agents in planning and decision making – a firm CEO may target stronger firm governance characteristics following evidence of a trend showing corporate governance has a stronger influence on credit ratings over time, for example. The results can also help us explore the response to credit rating evaluations during the financial crisis of 2008. In this paper, additional models and variables will be used to explore these new areas of research.
The motive for investigating the impact of the Great Recession is due to the intense scrutiny corporate governance came under when weaknesses were revealed by the economic downturn. The OECD (2009) reported that failures and weakness in corporate governance practices were responsible to a significant extent for the financial crisis, and
in 2010 they published detailed recommendations businesses should follow to improve vital areas of their governance strategies.
Using ordered logistic regression models to regress credit ratings on its determinants, we obtain results showing whether there is a significant link between corporate governance especially, and the ratings. Further, with variants of the model, we show whether time has affected both credit ratings themselves and the relationship between corporate governance and credit ratings. Next, by trimming the data into a more balanced sub-sample, the effect of a present financial crisis on the relationship between corporate governance and firm credit ratings is evaluated. To conduct this study, data from North America is used. Data is collected over 23 years from 1991 to 2013 and 1,283 firms are included in the study.
Previous papers usually utilise just a few select years of data, and often from many years before the paper was published. For this study, a vast amount of data is collected over an impressive range of years, and the most modern data available is used, making it one of the most extensive and relevant data sets analysed for this area of research. A comprehensive study is presented, but it is one that’s also required in order to uncover such findings.
The remainder of this paper is structured as follows: Section 2 will provide an overview of existing literature within the area of research, including important contributions.
Section 3 offers an insight into the specifics of the data used. Section 4 presents the employed methodological approach to the analysis. Section 5 provides the results to our analysis and investigates the relationship between corporate governance and corporate bond ratings, and Section 6 follows by checking the robustness of the results obtained.
Lastly, Section 7 concludes the paper and offers the resulting implications for financial agents who the results concern.
2 Literature Review and Hypotheses
Whilst Horrigan (1966), who initiated this line of research, used an ordinary least squares regression model for investigating corporate bond characteristics, this method assumes the differences between each of the credit ratings are equal and the variable itself is on an interval scale (Alexander, 2013). However, credit ratings are not on a linear scale and the
distance between any two ratings is not necessarily equivalent. A slightly better method for our purpose is the multiple discriminant analysis model, which Pinches and Mingo (1975) employ, as well as Altman and Katz (1976). Yet, multivariate discriminate analysis is an unordered method. Ederington (1985) examined performances of different models and investigated their statistical fit of explaining bond ratings. The unordered logit and ordered probit models (which Feki and Khoufi (2015) used) outperformed methods of ordinary least squares and multiple discriminant analysis, and it was found that ordered logit models performed best for predicting ratings out of the used sample. Other, less popular techniques include using Bayesian networks (Wijayatunga et al., 2006), and support vector machines and neural networks (Huang et al., 2004).
The most prominent studies in the area, including those by Ashbaugh-Skaife et al. (2006) and Alali et al. (2012), use ordered logistic regression models. Alexander (2013) reasons that ordered logit models are used frequently and are the most suitable partly because of the a priori assumptions they possess. The ordered assumption of structure is present and relevant with ordered logit regression analysis, whilst it has an ability to adjust to precise features of the ratings scale. Most studies, such as the two prominent ones mentioned, assigned numerical values to seven different ordered categories of ratings. Murica et al.
(2014) followed suit, and additionally split the ratings into three new classes for technical reasons. Alali et al. (2012) and Gupta (2021) further grouped the ratings into two classes (creating a binary variable), and eight classes, respectively. All the mentioned divisions make intuitional sense and give valid regression models, whereas separate categories for each rating division would yield too much variance in the dependent variable, which is the reason fewer categories have been utilised by previous researchers. Following this evidence, this study also employs an ordered logistic regression model.
Although there is much research on credit ratings and their determinants, there is a lot of difference in models constructed and analysis outcomes in the existing literature. Thus, there aren’t many easy overall conclusions to reach regarding variables’ effects on credit ratings. Previous authors have arrived at varying results and thus it can still be strongly debated what determines corporate credit ratings. An obvious reason for the differences is the different models at the disposal of each previous author and these differences are compounded by differing data sets. It seems clear that the magnitude of the effect of
determinants on credit ratings changes from country-to-country, or year-to-year.
Therefore, a suggested approach of study may well be to not analyse the mean of these effects, but to analyse how they changed over time or over differing firm characteristics.
2.1 Determinants of Corporate Credit Ratings
As Alexander (2013) articulates, the three main categories of determinants are financial data, corporate governance mechanisms, and macroeconomic factors. From financial data, financial ratios are obtained, such as firm leverage levels and return on assets, used as proxies for respective determinants. Due to how publicly available financial ratios of corporations derived from financial statements are, coupled with the importance S&P puts on them to determine the resulting rating (Standard & Poor’s, 2010), in addition to their strong significance in financial models of determining credit ratings, it is no surprise that most studies place emphasis on these variables. The most widely accepted financial ratios – barring the two previously mentioned – for determining credit ratings are liquidity, interest coverage, capital intensity, and growth. Firm size is another key financial variable of interest, as it will be in the present study. Previous studies that use all, or a mix of most of these variables include those done by Alali et al. (2012), Murcia et al. (2014), Gupta et al. (2017), and Gupta (2021), amongst others.
As for corporate governance mechanisms, a few different calculations have been used previously with different conclusions drawn in magnitude, sign, and significance of the respective coefficients. Tarigan and Fitriany (2018) thoroughly analyse the effects of individual components of corporate governance by splitting it into sections of company structure, ownership structure, audit committee, and external auditor. These sections make up nine variables of interest, and controls are added for financial determinants. The results indicate that nearly all factors of corporate governance such as the number of directors, the number of commissioners who serve the company, and being audited by both a big 4 auditor and second tier auditor have positive and significant effects on corporate credit ratings. It can therefore be deduced that corporate governance strength helps firms to receive higher credit ratings. Further, the only variables that had a negative effect were ones that were more-subjectively beneficial for a firm such as the number of shareholders owning 5% or more of the outstanding shares. These results are firmly
backed up by the research of Alali et al. (2012), who used different summary governance score variables in their United States study. They used the Gov-Score of Brown and Caylor, which Brown and Caylor (2006, p. 410) describe themselves as “a summary governance measure based on 51 firm-specific provisions representing both internal and external governance”. It is a sophisticated mechanism but not without its limitations, which is why Alali et al. (2012) added robustness to their study by also testing the model with use of the G-index of Gompers et al. (2003) and the entrenchment index of Bebchuk et al. (2009) as alternative proxies for the level of firm governance. They found similar results with the alternative proxies to the one used in the main study (Gov-Score), which had positive and significant effects on credit ratings. Conversely, Murcia et al. (2014) did not find the effect of corporate governance on credit ratings to be significant, where they simply made corporate governance a dummy variable of value 1 if the firm was in Special Corporate Governance Level 2 and the Novo Mercado – which are both corporate governance categories with specific requirements in Brazil (Chavez and Silva, 2009) – and 0 otherwise. However, due to the simplicity of the variable used and its lack of relevance to the current study, being specific only to a country in a different continent;
the study of Murcia et al. (2014) provides little to suggest corporate governance does not have a significant impact on credit ratings in the United States, or anywhere else in the world other than Brazil. In fact, from the extant literature we can conclude that corporate governance does have a positive and significant effect on firm credit ratings (with the calculations for corporate governance used), we just can’t conclude upon what might change the magnitude of this effect. The present study will further explore this conclusion and analyse whether corporate governance has a significant effect on corporate credit ratings using a new summary governance variable derived from MISC data.
Another prominent study not yet mentioned includes that done by Bradley et al. (2008), who found that whilst a firm’s financial condition is a primary factor in determining its credit rating, aspects of corporate governance are also significantly related to credit ratings. They found qualities of corporate governance related to transparency, ownership structure, shareholder rights, board structure, and executive compensation to all significantly effect credit ratings when financial conditions were controlled for. From their study and the others mentioned, it can be inferred that corporate governance features and summary measures of governance are only secondary factors in determining credit
ratings. This can also be inferred from S&P’s (Standard & Poor’s, 2004) writing on corporate governance, but we still do not know if corporate governance is growing to become a primary determinant, or even just a determinant of more importance in general with regards to credit ratings. Realising whether this is the case or not is important to the overall understanding of corporate governance and especially its effects on credit ratings.
Given that S&P expected the link between the two to evolve, and the rapid change of attention towards the topic across the globe, it can be expected that corporate governance has become and will continue to become more important over time.
2.2 The Change of the Effect of Corporate Governance
There is little extant literature which investigates how the effect of corporate governance on credit ratings may change over time or via different firm characteristics or settings.
There is a clear use of investigating this to all of analysts, researchers, managers, investors, and other financial agents because it is then possible to spot trends; the current paper strives to significantly add to the literature in this area with regards to the change of the effect over time.
Alalit et al. (2012) extended their study to investigate how the effect of corporate governance changes due to firm size. After they concluded that both a higher level and an improvement in corporate governance were associated with more superior firm credit ratings, they found that this effect was accentuated for smaller firms relative to large firms. This is a strong addition to their study and to the literature, as rather than just stating the effect corporate governance has on credit ratings for their data set, an understanding can be derived about what determines the size of this effect and we can also understand that the effect is not going to be the same for firms in different scenarios or with different characteristics. Although they only used two categories of firms (small and large, with large being the top tercile and small being the bottom tercile of all firms according to size), they found significant results which helps many categories of personnel understand that smaller firms may benefit more with regards to credit ratings from focusing relatively more than large firms on the governance strategies of their corporation. As far as this study is aware, the research by Alali et al. (2012) is the only one which investigates partly into what determines the effect of corporate governance on credit ratings. There is thus
plenty of scope to add to the literature in this area and much more that can be reported about the change of the effect in question.
There is currently no existing literature which aims to investigate the change of this effect over time or during certain economic events, and therefore it is hard to evaluate trends over periods, if they exist, and makes it impossible to predict how the importance of the determinant might change. Seeing as many significant macroeconomic events have occurred over the recent decades as well as corporate governance obtaining more importance in general, it is important to understand how the relationship between the two variables has changed, if at all. It is particularly important to those who it may affect directly (such as investors and managers) so that their strategies can be adapted, however it is also important for the literature too as we cannot expect this relationship to remain constant, but we are unsure how it changes. Finding a single figure for a given relationship over several years provides limited scope for analysis; but being able to deduce what affects this relationship would provide a deeper understanding about it and therefore a better understanding about how this relationship may change in the future. This thesis heavily relates to the previous literature by being focused on an area of much popularity and interest yet increases the breadth and capacity of information available by exploring a new angle on corporate governance and how it effects corporate credit ratings.
2.2 Hypotheses Development
Most of the literature in this study area has found corporate governance to have a positive and significant effect on corporate credit ratings within the respective data sets analysed.
The amount of evidence provided for this is vast and intuitively we expect that a higher level of corporate governance reduces firm agency problems as it improves the probability of managers acting in the interest of the firm shareholders and less on their private objectives which may harm firm performance. This is usually due to appropriate monitoring over firm managers and contract incentives provided to them in order to keep manager and shareholder incentives aligned. Therefore, a greater level of corporate governance should result in financial targets being met, meaning debt repayments to lenders will be met more easily and the prospect of lending credit is safer as the company borrowing is more reliable. As this is an important factor analysed in giving corporate
credit ratings and S&P puts a positive weight on corporate governance strengths when evaluating credit ratings, we are able to derive our first hypothesis:
H1: A higher level of corporate governance attributes to a higher level of firm credit rating.
It is not clear yet whether this study will likely find corporate governance to have greater influence on corporate credit ratings than past studies as the data set used includes both older (from 1993) and newer (up to 2013) data than has been generally previously used;
Alali et al. (2012) analyse the relationship between corporate governance and corporate credit ratings over the period of 2003-2005, whilst Ashbaugh-Skaife et al. (2006) analyse data from the 2003 proxy season covering firm governance structures of the previous fiscal year. However, what is likely is that we find a change in the relationship over time.
Corporate governance has come under more scrutiny and placed under more importance in general by policy influencers, bondholders and rating agencies over time due to weak governance being directly related to financial failures of businesses in the past. For example, the UNCTAD’s analysis puts partial blame on poor corporate governance practices for the cause of the crisis (UNCTAD, 2010). Corporate governance improvements generally occur from a backward-looking point of view and to avoid failures which have previously occurred (as hoped for from the OECD publishing corporate governance practice recommendations in 2010 after the financial crisis). We therefore expect the relationship between corporate governance and credit ratings to become stronger over time and for the reverse to never happen. It is both unlikely and hard to reason for less importance to be placed on corporate governance over time when analysing corporate credit ratings. The second hypothesis is therefore constructed as:
H2: The impact of corporate governance on corporate credit ratings has become stronger over time.
Further, the financial crisis of 2007-2008 was so severe that Americans had a net loss in wealth of $9.8 trillion (Merle, 2018). Other metrics observed include an increase in suicide numbers (Chang et al., 2013), and a decrease in fertility (Schneider, 2015). The adverse effects of the crisis are astronomical and uncountable. As corporate governance weaknesses were largely to blame for many firms failing to properly carry out business and legal processes (OECD, 2009), this study aims to investigate specifically whether the
relationship between corporate governance and corporate credit ratings shifted temporarily due to the Great Recession. This leads to the third hypothesis to test:
H3: The impact of corporate governance on corporate credit ratings became stronger during the years of the Great Recession.
3 Data Description
The present study employs a panel data set comprising of governance and financial variables, totalling 11,125 observations from 1,283 companies over 23 years between 1991 and 2013. As is the case for most studies analysing the effect of corporate governance on firm credit ratings, data for the country selected was mostly easy to access, but data from other countries was not incorporated into the main model nor was it used for any robustness checks later in the study. This paper focuses solely on the United States due to availability of data at present and the importance of choosing a high-economically developed country for this research. This study is more forward-looking than related studies, so using the most modern data is of more importance. As low-economically developed countries lag high-economically developed countries in many economic standards and processes, corporate governance mechanisms and thus improvements also lag in the same way, meaning that the newest data from a low-economically developed country won’t be as modern as the equivalent data from a high-economically developed country.
Table 1 shows that three data sources have been used (as well as the variable constructions from those sources), from which the respective datasets have been merged into a final data set which this study analyses. The firm credit ratings come from Standard & Poor’s (S&P Global Ratings), whilst all corporate governance data is available from MSCI (formerly KLD and GMI). Finally, a multitude of financial data was acquired from Capital IQ, which contains all appropriate statistics and measures needed in order to formulate proxies for the desired variables.
Table 2 below shows the summary statistics of the dependent and independent variables.
There are 11,125 observations for each variable as any data entries with missing observations were excluded from the data set.
Table 1: Variable Definitions and Sources
Variable Definition Source
Credit_Rating Ordinal Rating 1 S&P Ratings
Corporate_Governance Total Strengths – Total Concerns MSCI (KLD)
Profitability Net income ÷ Total Assets Capital IQ
Capital_Intensity PPE ÷ Total Assets 2 Capital IQ
Size Log(Total Assets) 3 Capital IQ
Leverage Long-term Debt ÷ Total Assets Capital IQ
Interest_Coverage EBIT ÷ Interest Expense Capital IQ
This table presents the variables used in the analysis and their definitions or calculations and the source of the data. (1) The credit rating is an ordinal scale ranging from 1 to 7, with 7 indicating the most superior ratings. The categories are shown in table 2. (2) PPE is Property, Plant, and Equipment. (3) The logarithm used is the natural logarithm.
Table 2: Summary Statistics
Variable Obs Mean Std. Dev. Min Max
Credit_Rating 11,125 3.850 1.189 1 7
Corporate_Governance 11,125 -0.356 0.719 -4 2
Profitability 11,125 0.044 0.051 -0.067 0.141
Capital_Intensity 11,125 0.317 0.244 0.008 0.790
Size 11,125 8.566 1.248 6.506 10.981
Leverage 11,125 0.260 0.154 0.028 0.600
Interest_Coverage 11,125 7.414 9.382 -1.074 36.002
This table presents the distribution of institutions’ credit ratings and the determinants of credit ratings. We report the number of observations, mean, median, standard deviation, min, and max values of the estimates. Credit_Rating = the created ordinal variable taking values 1 through to 7 where a higher value is a more superior rating, Corporate_Governance
The data set being used here is more comprehensive than most previously used by other authors, and the time span especially is adequate for noticing trends (if any) over time.
Murcia et al. (2014) used a data set comprising of 49 companies totalling 153 observations for their Brazilian study, whilst Sareen and Madhu (2015) included 252 companies that were assigned long term ratings in 2012 for their Indian study. Our dataset is more like that of Lin et al. (2018), who used a data set comprised of 12,024 firm-year observations from 1,475 firms over the years of 2005 to 2013 for their successful study on Taiwan involving corporate governance and corporate social responsibility.
The dependent variable, Credit_Rating, is the ordered variable taking values from 1 to 7 (also seen as the min and max values) that have been assigned to groups of ratings, forming categories. A mean value of 3.850 implies a credit rating in the BBB+ to BBB- category, which holds 33% of the total observations. A standard deviation of 1.189 shows that the data is concentrated around the mean, with neighbouring categories (3 and 5) also having a lot of observations and the two extreme values (1 and 7) containing much fewer observations (see Table A1). Corporate_Governance has a negative mean value of -0.356, showing firms on average had more ‘concerns’ than ‘strengths’ which are used to construct this score, with a relatively high standard deviation. The lowest corporate governance score recorded is -4 whilst the highest is 2, so a negative mean is not surprising. On average, firms within the dataset are profitable with a value of 4.4%, the standard deviation for this variable is low (5.1%), however due to winsorization at the 5% level this makes sense. A mean Capital_Intensity of 0.317 is low, showing a lot of firms have a lot of intangible assets, with the extreme values being 0.008 and 0.790. On average, firms are only 26% leveraged, with some firms taking on close to zero long-term debt and some being up to 60% leveraged. Interest_Coverage has a mean of 7.414, indicating that on average, firms have much more income relative to their interest
expenses (by a factor of 7.414), although the minimum value recorded is negative (-1.074), whilst the maximum value recorded is 36. A score as high as 36 indicates that
the firm is either highly profitable or has low interest costs, or both.
In the Appendix section of this paper, Table A1 provides a view of how the credit rating values are distributed within their ordinal categories and over the years analysed. From 1991 to 2001, the mode value for the Credit_Rating variable is 5, indicative of the A+ to A- category; from 2002 onwards the mode is 4, which is the overall mean for the data set.
Values 6 and 7 generally lose observations as time advances, even though there are more total observations in the later years. This indicates either corporations generally found it harder to meet the conditions of these categories or S&P made the criteria stricter. Every other value of Credit_Rating generally gained observations as time went on.
Table 3 contains results of the correlations between each of the variables. Although the value the correlation between Corporate_Governance and Credit_Rating is negative, it is of small magnitude and statistically insignificant even at the 10% level, disallowing us to conclude on the relationship yet. To achieve an insignificant result is within expectations as it’s known that the S&P place more importance on the financial variables (Standard &
Poor’s, 2010), which we can see have stronger and direct relationships, all returning highly significant values. Financially stronger-performing firms have a directly greater ability to meet debt obligations as they possess the means to, corporate governance weaknesses can then start to take away from this strength, so with significant weaknesses Table 3: Pearson’s Correlation Coefficients
Variable (1) (2) (3) (4) (5) (6) (7)
(1) Credit_Rating 1.000
(2) Corporate_Governance -0.010 1.000
(3) Profitability 0.366*** 0.009 1.000
(4) Capital_Intensity -0.007*** 0.117*** -0.065*** 1.000
(5) Size 0.475*** -0.130*** 0.037*** 0.006 1.000
(6) Leverage -0.448*** 0.047*** -0.298*** 0.263*** -0.221*** 1.000
(7) Interest_Coverage 0.386*** -0.038*** 0.586*** -0.222*** 0.118*** -0.547*** 1.000
This table presents the Pearson’s correlation coefficients. Credit_Rating = the created ordinal variable taking values 1 through to 7 where a higher value is a more superior rating, Corporate_Governance = the summary governance score measure where the value indicates the cumulative score of a
summation of corporate governance strengths and weaknesses, Profitability = Net income ÷ Total Assets, Capital_Intensity = PPE ÷ Total Assets, Size = Log(Total Assets), Leverage = Long-term Debt ÷ Total Assets and Interest_Coverage = EBIT ÷ Interest Expense. *, **, *** shows significance at 0.05, 0.01, 0.001 levels, respectively.
in corporate governance firms may still deserve a high credit rating. On the contrary, if firms are entirely unable to meet their debt obligations they agreed to, and this is visible in the financial data, they are strictly not entitled to a positive credit rating, regardless of how strong their corporate governance is. Additionally, the Pearson’s correlation matrix displays only univariate, linear analysis results without any controls, so analysing the direct relationship between corporate governance and firm credit ratings has minimal use apart from a purely data-oversight perspective. When we perform a logit regression analysis controlling for the other variables, we can expect a more accurate representation of the association between stronger levels of corporate governance and the respective credit ratings rewarded. It is much more logical and within expectations to see positive and significant correlations between Credit_Rating and Profitability, Size, and Interest_Coverage. The same can be said about the negative correlation observed between Leverage and Credit_Rating; having more debt relative to total assets makes it directly harder for a firm to meet debt obligations and therefore a lower Credit_Rating is expected.
Some correlations displayed in the table are slightly higher, for instance those between Interest_Coverage and Profitability, as well as Interest_Coverage and Leverage. High intercorrelations between independent variables can challenge the reliability of results obtained through the regression analysis and is undesired (Daoud, 2017). However, to formally determine if the correlations pose a threat to the reliability of the results, the multicollinearity is further checked by analysing the variance inflation factors (VIF) values at a later stage.
4 Methodology
This study utilises an ordered logistic model, also referred to as the proportional odds model, with year and industry fixed effects. In line with Ashbaugh-Skaife et al. (2006), who recommends the model, and Alali et al. (2012), we agree this method is most appropriate for the research in question largely due to the dependent variable being ordered yet non-numerical. First deliberated by McCullagh (1980), the model is an extension of the standard logistic regression model, with specific use cases for ordered dependent variables which are dichotomous (Alali et al., 2012). As we are splitting the
dependent variable into seven separate groups, it is the most applicable method for the present study.
The credit ratings data used in the present study is that of Standard & Poor’s for US companies. Table 4 shows - following Ashbaugh-Skaife et al. (2006) and Alali et al.
(2012) – how the independent variable has been split up with different ratings being grouped together and assigned integer values on the ordinal scale. For additional analysis checks, two categories for the type of rating (speculative and investment) have also been created themselves assigned to one of these categories.
Our corporate governance variable, Corporate_Governance, is an index calculated by subtracting the total number of observed corporate governance related concerns from the total number of observed corporate governance strengths based on data from MSCI. Each provision is given a value of 1 if criteria is met for either a strength or weakness; 0 otherwise. There is a total of 18 firm-specific provisions which make up the entirety of the Corporate_Governance variable, with 8 of these being strengths, and 10 being weaknesses. Table A2 in the Appendix elaborates on the definitions of each of these provisions. It is a similar index to the Gov-Score of Brown and Caylor (2006), which is a summary governance measure based on 51 provisions. Gomper’s G index (Gompers et al., 2001) is another similar index, based upon 24 unique governance rules.
Table 4: Credit Rating Classifications
Credit Rating Ordinal Scale Credit Rating Type
CCC+, CCC, CCC-, CC, D, SD 1 Speculative
B+, B, B- 2 Speculative
BB+, BB, BB- 3 Speculative
BBB+, BBB, BBB- 4 Investment
A+, A, A- 5 Investment
AA+, AA, AA- 6 Investment
AAA 7 Investment
This table presents the credit rating classifications which institutions can be assigned based on their credit rating. The ordinal scale takes values from 1-7, where a higher value indicates a more superior credit rating, and type of credit rating is divided into two categories – Speculative and Investment. For example, a B+ corporate credit rating is grouped with B and B- and assigned ordinal value 2, with orders 1, 2, and 3 being grouped into the Speculative type category.
To isolate the causal effect that the level of corporate governance has on corporate bond ratings, control variables are required. This study incorporates firm performance (profitability), capital intensity, firm size, leverage, and interest coverage into the model as relevant and necessary controls. The required financial data to form these ratios is detailed in Table 1. The controls act as proxies for a firm’s probability of default, and their inclusion is important for evaluating the causality between the level of corporate governance and credit ratings.
With the data collected for 1,283 companies over 23 years (1991 – 2013), containing their credit ratings, corporate governance scores, and financial data (used to formulate proxies for desired variables; listed in Table 1), some minor data manipulation is applied. Firstly, all financial firms are excluded from the data set due to significantly different industry regulations and characteristics. Second, data entries with missing observations are excluded from the data. Lastly, the remaining data is winsorized at the 5% level to remove outliers which may disproportionately affect the analysis results. Ordered logit regression analysis is carried out to test the causal impact of firm governance score on firm credit ratings. Specifically, the following equation is formed:
(1) Where,
(2) In order to test the first and main hypothesis, the regression above is estimated. For the hypothesis to be accepted, b1 needs to be a positive and significant coefficient, at least at the 10% level of significance. Having stronger significance demonstrates that the results generated for the current sample will have stronger applicability to the general population, so a reasonable level of significance is required to draw any conclusions. As for the control variables, positive coefficient signs are expected on all but one of them, with Leverage having the only predicted negative coefficient.
𝐶𝑟𝑒𝑑𝑖𝑡_𝑅𝑎𝑡𝑖𝑛𝑔𝑖𝑡= b0 +b1𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒_𝐺𝑜𝑣𝑒𝑟𝑛𝑎𝑛𝑐𝑒𝑖𝑡+Sb𝑗(𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖𝑡) +e𝑖𝑡
Sb𝑗(𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖𝑡)
=b2𝑃𝑟𝑜𝑓𝑖𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑖𝑡+b3𝐶𝑎𝑝𝑖𝑡𝑎𝑙_𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦𝑖𝑡+b4𝑆𝑖𝑧𝑒𝑖𝑡+b5𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡 +b6𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡_𝐶𝑜𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡
When long-term debt increases relative to total assets, a firm will have more financial demands relative to its capabilities and therefore will find it harder (or at least no easier) to repay the debt (Bhojraj and Sengupta, 2003). Therefore, as this ratio increases, it becomes more unlikely the debt gets repaid on time or at all, making leverage levels an appropriate proxy for default risk. Due to the known importance of this ratio in S&P ratings, being purely financial data (reference?), the increased potential difficulty for a firm to repay its debt should be reflected upon in general with lower rating scores. The prediction of a negative relationship is supported by much of the literature with authors including but not limited to Gupta et al. (2017), Ashbaugh-Skaife et al. (2006), and Alali et al. (2012) all reporting negative and highly significant coefficients.
Profitability is expected to have a positive coefficient due to importance of this financial ratio to a firm’s survival and security. It is a suitable proxy for default risk due to how much harder it becomes for a firm to pay its debt when profitability is lower or negative.
Profitability is expected to be a leading determinant of credit ratings, and this should be reflected in the following results.
Another proxy used for default risk is the measure of a firm’s tangibility levels, referred to as Capital_Intensity in the present study. Defined as the value of property, plant, and equipment divided by total assets, a firms tangibility levels can reflect its abilities to liquidate assets when necessary in order to meet debt obligations, making it more likely debt will be repaid. This reduces the probability of a firm’s default and increases the credit rating.
Size, the natural logarithm of total assets, is an important control to include in the model because logically, large firms are expected to have less risk of default than small firms (Bhojraj and Sengupta, 2003). Generally, a smaller firm may have more limited ways of meeting debt obligations and may have higher costs of debt relative to the size of debt too, increasing the difficulty of debt repayment as well as the default risk, reducing the credit rating grade. We therefore control for firm size in order to better analyse the causal effect of corporate governance levels on credit ratings.
Lastly, interest coverage (Interest_Coverage) adequately acts as a proxy for default risk as it directly demonstrates the ability of a firm to meet its demands of debt repayment. It has a positive predicted coefficient sign. The ratio shows how many times greater a firm’s
earnings before interest and taxes are than its interest expenses. Firms with higher interest coverage levels can meet debt obligations more easily and are therefore assumed to have greater credit rating scores, making Interest_Coverage a necessary control.
The control variables stated and explained above are important for establishing causality rather than simply causation in the relationship of primary importance. Although causality requires that all other factors are controlled for, this is empirically impossible due to the number of total factors which can credit ratings, accompanied by the differing levels of importance each factor has for different types of firms. Nonetheless, in this paper, the most important determinants of credit ratings, i.e., business risk and financial risk elements, have been controlled for. In addition, industry and firm fixed effects are included, and robustness checks are later carried out to validate the claim of causality, with the use of clustered standard errors, multicollinearity checks, and testing on subsamples of the data. As a result, a level of causation, rather than just correlation, can be concluded upon.
Testing the first hypothesis of this paper will strengthen and add to the presently available literature by applying a comprehensive methodology of analysis to a new, modern, and wide-spanning data set. But it is the testing of the second and third hypothesis which explores a new avenue within the topic and will allow us to have a greater understanding of the relationship between corporate governance and firm credit ratings.
Testing our second hypothesis allows us to investigate whether there is a trend in how corporate governance affects credit ratings over time. In order to test this hypothesis, some further steps need to be taken in the formulation of the model. To assess the change in relationship, a time variable needs creating, Time, as well as an interaction term for this variable with the main variable of interest, corporate governance. The term Corporate_Governance*Time is added to the model, giving a new regression equation:
(3) Where,
𝐶𝑟𝑒𝑑𝑖𝑡_𝑅𝑎𝑡𝑖𝑛𝑔𝑖𝑡
= b0 +b1𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒_𝐺𝑜𝑣𝑒𝑟𝑛𝑎𝑛𝑐𝑒𝑖𝑡+b2𝑇𝑖𝑚𝑒𝑖𝑡
+b3𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒_𝐺𝑜𝑣𝑒𝑟𝑛𝑎𝑛𝑐𝑒 ∗ 𝑇𝑖𝑚𝑒𝑖𝑡+Sb𝑗(𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖𝑡) +e𝑖𝑡
Sb𝑗(𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖𝑡)
= b 𝑃𝑟𝑜𝑓𝑖𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦 +b 𝐶𝑎𝑝𝑖𝑡𝑎𝑙_𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 +b 𝑆𝑖𝑧𝑒 +b 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒
b2 is the coefficient for the Time variable. It demonstrates the relationship between the advancement of time and credit ratings whilst accounting for other factors. The variable takes values of 1-23, assigned for each of the years in our sample. A positive value here would show that that as time has advanced through our sample years, credit ratings have increased. However, by just observing Table A1 we can see this is not the case and in general credit ratings have decreased. It is therefore expected that in the regression results for equation (3), b2 is both negative and significantly so. In order to target our hypothesis septically, b3 is the coefficient of primary interest. This value, if significant, will allow us to conclude upon whether there has been a change in relationship over time and the sign of the coefficient allows us to deduce whether corporate governance has become a factor of increasing (positive sign) or decreasing (negative sign) impact on credit ratings.
The last analysis to be carried out is the investigation of the third hypothesis – whether corporate governance affects firm credit ratings differently during times of a financial crisis. Just like the previous hypothesis, additional terms need to be incorporated into the original model. A dummy variable Crisis is included, taking the value 1 if the year is 2008 or 2009, and 0 otherwise. Additionally, an interaction term with the corporate governance variable, Corporate_Governance*Crisis, is included. The size and significance of the coefficient for this variable will allow us to deduce whether corporate governance affects credit ratings differently during a financial crisis. We test this over the Great Recession and exclude data before 2003 to generate a more balanced data set as well as excluding other significant economic events such as the dot-com bubble. The resulting equation to estimate is:
(5) Where,
𝐶𝑟𝑒𝑑𝑖𝑡_𝑅𝑎𝑡𝑖𝑛𝑔𝑖𝑡
= b0 +b1𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒_𝐺𝑜𝑣𝑒𝑟𝑛𝑎𝑛𝑐𝑒𝑖𝑡+b2𝐶𝑟𝑖𝑠𝑖𝑠𝑖𝑡
+b3𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒_𝐺𝑜𝑣𝑒𝑟𝑛𝑎𝑛𝑐𝑒 ∗ 𝐶𝑟𝑖𝑠𝑖𝑠𝑖𝑡+Sb𝑗(𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖𝑡) +e𝑖𝑡
Sb𝑗(𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖𝑡)
= b4𝑃𝑟𝑜𝑓𝑖𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑖𝑡+b5𝐶𝑎𝑝𝑖𝑡𝑎𝑙_𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦𝑖𝑡+b6𝑆𝑖𝑧𝑒𝑖𝑡+b7𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡 +b8𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡_𝐶𝑜𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡
(6)
5 Results
So far, the only results displayed have been summary statistics and a Pearson’s correlation matrix. Although they both provide valuable insights regarding the data and correlations within it, they provide little information about the causal effects of the variables within the models. Especially for the corporate governance variable, which so
Table 5: Logistic Regression Results of the Effects of Corporate Governance on Firm Credit Ratings
Variable Dependent Variable = Credit_Rating
Model 1 Model 2 Model 3
Corporate_Governance 0.061
(1.54)
0.113***
(2.74)
Profitability 2.694***
(3.59)
2.628***
(3.50)
Capital_Intensity 2.740***
(7.46)
2.753***
(7.49)
Size 2.238***
(34.42)
2.248***
(34.49)
Leverage -5.171***
(-15.06)
-5.187***
(-15.10)
Interest_Coverage 0.075***
(14.50)
0.075***
(14.45)
Industry Dummies Included Included Included
Year Dummies Included Included Included
Observations 11,125 11,125 11,125
Chi-squared 1008.60*** 2524.26*** 2527.49***
Pseudo R-squared 0.429 0.358 0.358
This table presents regression results from estimating equation (1) using data from S&P, MSCI, and Capital IQ.
Corporate_Governance = the summary governance score measure where the value indicates the cumulative score of a summation of corporate governance strengths and weaknesses, Profitability = Net income ÷ Total Assets, Capital_Intensity = PPE ÷ Total Assets, Size = Log(Total Assets), Leverage = Long-term Debt ÷ Total Assets and Interest_Coverage = EBIT ÷ Interest Expense *, **, *** shows significance at 0.10, 0.05, 0.01 levels, respectively. The parentheses contain z-values.
Industry and year dummies are included. Model 1 contains Corporate_Governance as the only independent variable. Model 2
far, has been found to have no direct correlation with higher credit ratings. This outcome was to be expected with a more complete, applicable model now required with controls to have an idea about the independent causal impact corporate governance has on credit ratings.
Table 5 contains regression results for models 1, 2, and 3, where the independent variables are Corporate_Governance only, controls only, and both together, respectively. The output results for model 3 come from equation (1) composed earlier, where our interest lies. The coefficient for Corporate_Governance is positive and highly significant (at the 1% level of significance), demonstrating that a stronger level of corporate governance does increase firm credit ratings. This supports our first hypothesis, as well as strengthens many results of previous studies. Using a new method (data from MISC) to calculate a corporate governance score, whilst implementing a new model, and applying the model to a bigger data set both observations-wise and timewise is hugely validating for previous studies which have looked for and found similar results. The Table 5 results show that our model is valid whilst even incorporating data from 1991. The results show that a one- unit standard deviation increase in our primary variable, Corporate_Governance, translates to 0.113 of a standard deviation increase in the Credit_Rating variable.
The results are also shown to exhibit high explanatory power, with a McFadden Pseudo R square value of 0.358 for model 3. Menard (2000) concluded that McFadden’s was the favoured pseudo R2 index (in a study comparing five different pseudo R2 indices) when working in a logistic regression context, and therefore it is the used index here. The explanation by McFadden (1977, p. 35) himself demonstrates just how strong the present values are: “Those unfamiliar with the 𝜌2 index should be forewarned that its values tend to be consistently lower than those of the R2 index… values of .2 to .4 for 𝜌2 represent an excellent fit”.
It is worth noting that in model 1 above, corporate governance does not have a significant coefficient and therefore, just like with the insignificant result in the correlation matrix between Corporate_Governance and Credit_Rating, no concrete conclusion can be drawn from such a simple result. The model 1 and 2 result adds further to the idea that corporate governance is a secondary variable when S&P evaluates credit ratings and adds further to the conclusion that appropriate controls are required.
Table 6: Logistic Regression Results of the Effects of Corporate Governance, Time, Crisis, and Their Interaction Variables on Firm Credit Ratings
Variable Dependent Variable = Credit_Rating
Model 4 Model 5
Corporate_Governance 0.261*
(1.80)
0.172***
(2.74)
Time -0.280***
(-23.59)
Corporate_Governance*Time -0.090
(-1.06)
Crisis -1.755***
(-10.73)
Corporate_Governance*Crisis -0.068
(-0.57)
Profitability 2.646***
(3.52)
1.565 (1.63)
Capital_Intensity 2.751***
(7.48)
1.898***
(3.68)
Size 2.251***
(34.50)
2.700***
(29.29)
Leverage -5.175***
(-15.06)
-4.600***
(-10.00)
Interest_Coverage 0.075***
(14.46)
0.081***
(11.53)
Industry Dummies Included Included
Year Dummies Included Included
Observations 11,125 8,287
Chi-squared 2528.24*** 1293.72***
Pseudo R-squared 0.357 0.384
This table presents the regression results from estimating equations (3) and (5) using data from S&P, MSCI, and Capital IQ. Corporate_Governance = the summary governance score measure where the value indicates the cumulative score of a summation of corporate governance strengths and weaknesses, Profitability = Net income ÷ Total Assets, Capital_Intensity = PPE ÷ Total Assets, Size = Log(Total Assets), Leverage = Long-term Debt ÷ Total Assets and Interest_Coverage = EBIT ÷ Interest Expense *, **, *** shows significance at 0.10, 0.05, 0.01 levels, respectively. The parentheses contain z-values. Industry and year dummies are included. Model 4 contains the variable TIME, and its interaction with Corporate_Governance. Model 5 contains the Crisis variable (a dummy equal to 1 if YEAR = 2008 or
Table 6 provides the results regarding the second and third hypotheses. Model 4 incorporates the new Time variable and its interaction with our corporate governance index variable, which allows us to conclude whether there is a trend in the relationship between corporate governance and credit ratings over time, and the direction of this trend.
Although the coefficient for the interaction component of model 4 is negative with a value of -0.09, this value is not significant even at the 10% level. Rather than having a trend over time to evaluate, the results show that there is not a significant trend within the dataset analysed for the relationship between corporate governance and credit ratings.
This is likely an incredibly complicated relationship and one that is hard to decipher in order to isolate the exact changes. From year-to-year and firm-to-firm the impact corporate governance has on credit ratings is likely to change vastly. Additionally, after significantly bad cases of weak corporate governance (such as those in the 2008 crisis) leading to adverse financial aftermath including wider impacts to society, regulation and expectations will change to enforce stronger levels of corporate governance. It is impossible to find a trend over time if the importance of corporate governance is decided by those events alone or largely. Interestingly, the model shows that when we effectively control for time, the impact corporate governance has on credit ratings increases, with a significant coefficient (at the 10% level) of 0.261. The time variant itself, with a highly significant coefficient of -0.280, shows what we already know from Table A1; there is a general trend of decreased credit ratings, holding everything else constant, over time. The magnitude of the effect indicates that on its own, time has the effect of decreasing the credit rating ordinal category by roughly 1, every 4 years that pass.
Model 5 allows us to isolate the impact that a present crisis has on the credit ratings themselves and the relationship between corporate governance and credit ratings. The results are similar to those presented in model 4: the additional variable here, Crisis, has a negative and highly significant value, but its interaction term with our corporate governance variable has no significance. Our third hypothesis states that we expected to see corporate governance having more of an impact on credit ratings during the years of the financial crash. With the insignificant result we conclude that there is no significant change regarding the magnitude of the impact corporate governance has on credit ratings during the years of financial crisis. However, with the Crisis variable itself having a highly significant result of -1.755, we show that a financial crisis in general, having
controlled for other affects, adversely impacts credit ratings. The magnitude suggests that during a financial crisis, credit ratings will decrease by almost two orders, e.g., from the category of A+, A, and A-, to BB+, BB, and BB-. However, this result is not so simple to observe from the data layout of Table A1 like it is for the results in model 4, but this is due to many other factors also impacting credit ratings at the same time. It is therefore highly likely that there are other factors working in favour of positive credit ratings during those years and the impact of a financial crisis during 2008 and 2009 combatted these positive impacts and resulted in small changes those years. Regardless, in general it would be expected to be able to observe the impact in Table A1, and additionally for credit ratings to be more sensitive to corporate governance levels during a financial crisis.
Just like models 1-3, models 4 and 5 have strong pseudo R2 results with values of 0.357 and 0.384, respectively, indicating that the explanatory power of those models is high.
This is no surprise after the related results in models 1-3, and there were only slight changes to the model in Table 6. As well as exploring new variables of Time and Crisis, along with their interaction terms, these models serve as robustness to the findings in model 3, which was the first complete model analysed. It strengthens the original finding that corporate governance does impact firm credit ratings as those findings remain significant even under some model adjustments and the explanatory power of all the models remained high.
5.1 Discussion of Results
Although the research, analysis, and results provided so far have demonstrated that there is no change in the relationship between corporate governance and credit ratings, both over time in general and during the specific years of the Great Recession, it has not been needed to for the study to be a success.
We firstly, and importantly show that the level of corporate governance, shown by a cumulative index resulting from the addition and subtraction of independent strengths and weaknesses, does significantly impact corporate credit ratings. We side with the literature and add robustness to the general area of research. The results add support to those found by Ashbaugh-Skaife et al. (2006), who find the governance score of Gompers et al. (2003) to have a positive effect on corporate credit ratings. Alali et al. (2012) found the same results as Ashbaugh-Skaife et al. (2006) using the same governance score and cement the
results by replicating the findings whilst using the Gov-score of Brown and Caylor (2006), and the entrenchment score of Bebchuk et al. (2009).
The present findings for the control variables selected also match the expectations given for them, with expectations drawn from existing research and theory. Profitability, Capital_Intensity, Size, and Interest_Coverage all have positive and highly significant effects on Credit_Rating in all relevant models (apart from the lack of significance in the Profitability variable in model 5). Whereas Leverage has a negative and highly significant effect on the Credit_Rating variable in models 2-4 (all where it was included). The findings of relations between the control variables selected and Credit_Rating are equivalent to those of Ashbaugh-Skaife et al. (2006). The difference in these controls is that they used return on assets (ROA) to proxy for performance as one of the proxies for default risk, whereas the present study uses profitability for the same measure. They included additional control variables too, but the strength of our results demonstrates they weren’t needed in the present models. Alali et al. (2012) included near-identical controls to Ashbaugh-Skaife et al. (2006) and included all the controls that our present study does.
Again, we find identical results for all control variables in terms of sign and significance, apart from in their model with an alternative dependent variable (a dummy variable of 1 if the firm was an investment grade, and 0 otherwise).
6 Robustness Analysis
Tables 7 – 10 contain results of multiple robustness checks for the analysis carried out so far. The last of those is a multicollinearity test, whereas the first three contain further regression analysis results.
Table 7 contains regression results for model 7, which has the same model specification as model 3, the main model, with the only difference being that standard errors are now clustered at the firm level. Clustering standard errors becomes useful when observations within a data sample may be subdivided into smaller-sized groups, called clusters. It allows for regression parameters’ standard errors to be estimated if this is the case. Using clustered standard errors is an applicable robustness check here in case there is clustering at the firm level, which wouldn’t be surprising given we can expect different observations
