The Impact of Credit Rating Downgrades and Rating Agency Competition

(1)

Master Thesis

The Impact of Credit Rating Downgrades and Rating

Agency Competition

December 2017

This thesis examines if acting more in favor of the issuer of a credit rating by a rating agency results in lower probability in being dropped by the issuer. I test what the effect is of delaying a downgrade by a credit rating agency, after a negative shock, on the probability of being dropped by the issuer. As well as what effect competition between rating agencies is on their behavior. Making use of two major drops in oil price, that acts as a exogenous shock, and a sample of oil companies, the amount of trading days till the first downgrade are used to measure the delay in downgrade. Oil companies are used as the returns of those companies depend mainly on the oil price. I find evidence that a credit rating agency that delays downgrading its credit rating is less probable of being dropped than an company that takes less time to downgrade. This result shows that rating agencies that act most in favor of the issuer by delaying a downgrade remains having the issuer as a client.

Author: Supervisor:

Chris Keizer Dr. Rafael Perez Ribas

(2)

Page | 1

Statement of Originality

This document is written by Student Chris Keizer 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.

(3)

Page | 2

1 Introduction

Credit rating agencies provide independent opinions and analysis on the creditworthiness and credit-risk of companies and their financial obligations. They asses credit ratings to companies in order to reduce information asymmetry between company and investor (Frost, 2007). From the 1970s, the business model of the credit rating agencies changed from investor-pay to issuer-pay in order to get rid of the free-rider problem. This change in revenue model has created conflicts of interest as rating agencies are dependent from issuers for revenue and rating agencies have to make independent analysis on their clients credit risk (Mathis et al, 2009). Many companies have multiple credit ratings which results in that rating agencies have incentive to act a bit in favor of the issuer in order to not lose it as a client.

This thesis investigates if delaying a downgrade in credit rating after a shock has influence on the probability of being dropped by the issuer. A shock that has negative effect on the creditworthiness of a company will affect the credit rating of that company in a negative manner. Therefore, a delay in downgrade after such shock will be in favor of the issuer and will reduce the probability of losing the issuer as a client. Thus, the hypothesis tested is that the probability of the credit rating being dropped will decrease with the amount of trading days delay in downgrade. In Addition, I hypothesize that competition between rating agencies for a company has influence on acting more in favor of the company.

I test these hypotheses using a sample of oil companies where a significant drop in oil price serves as a shock. There are two shocks used that had negative effect on the returns of oil companies and therefore on the creditworthiness of these companies. A regression model is used to estimate the effects of such delay by including the amount of trading days from the begin of the shock till the first downgrade. In addition there are multiple firm specific control variables and interaction variables included in the model. The findings of this research are in line with expectations that a delay in downgrading a credit rating after a negative shock results in reduced probability of the credit rating being dropped. Both after the first shock as after the second shock, the results show that

(5)

Page | 4 the amount of trading days till the first downgrade is negatively linked with the probability of being dropped. These results present evidence to not reject the hypothesis. There is no clear evidence found that competition between rating agencies has influence on the findings.

Prior literature mainly focused on examining if rating agencies act more in favor of the issuer or more in favor of the investor. Covitz and Harrison (2003) have found no evidence that rating agencies act systematically in favor of issuers. They have used anticipation of rating changes by the bond-market delay as a measure of credit-quality deterioration recognition. Other research by Bongaerts, Cremers and Goetzmann (2012) examined if companies shop for another rating if they have received a downgrade. Their research pointed out that around the investment grade - high yield grade border companies use another rating agency. Jiang et al. (2012) found in their paper that after the change in business model to issuer-pay, the rating agencies assign higher ratings to bonds that yield higher expected fees.

This research differentiates from previous literature in that I try to find what drives a company to drop one credit rating agency above another one and test what factors increase the probability of a rating being dropped. As well as what the effect of competition between rating agencies is on these factors. The finding that a delay, measured in this research by the amount of weeks between a negative shock and a downgrade in credit rating, reduces the probability of being dropped is in line with previous literature that the credit rating agency that acts more in favor of the issuer will keep it as a client.

The remainder of this thesis is organized as follows. Section two provides the related literature about credit rating agencies and derives the hypothesis. Followed by section three with an elaboration on the data used. Section four will discuss the methodology used in order to test the hypothesis. Afterwards, section five will present the results where section six provides a robustness check on the results. Section seven discusses the limitations and implications regarding this research and last section eight concludes.

(6)

Page | 5

2 Theory and Literature Review

2.1 Credit Rating

For companies to finance investments, new and ongoing projects and activities, bonds provide a critical mechanism. Companies raise more money through the bond market than by issuing equity. It is important for investors to know what the riskiness is for their investment as they want all the promised interest and principal payments associated with the bond issue. As is articulated by Moody (2003), the stated goal of rating agencies is to deliver an independent and objective credit-risk analysis.

Credit rating agencies provide such ratings that indicate the riskiness of an investment by predicting the probability of default for all kind of debt issuers and debt securities using a well-known single scale (Becker and Milbourn, 2011). The credit rating industry consists of worldwide more than 150 local and international credit rating agencies. Although there are so many rating agencies, three rating agencies are clearly the leading companies. These three rating agencies are Standard & Poor’s, Moody’s and Fitch, where S&P and Moody’s are the major two followed by Fitch. The credit rating industry can, therefore, be classified as an oligopoly (White, 2010).

The independent rating agencies examine the characteristics of the issue and a company’s financial outlook upon which they assign a rating that gives an indication of the risk of default corresponding to the firm’s bonds. These credit rating agencies essentially serve as an information intermediary that overcomes the information asymmetry between company and investor. They take into account both qualitative as well as quantitative information (Dittrich, 2007). It attempts to inform the investors what the likelihood is for them to receive their money plus interest back (Maher and Sen, 1997).

Credit rating agencies provide relevant information for both investors as corporations about the credit worthiness of the company. There have been many studies about the function of credit rating agencies in order to get a better understanding of how a credit rating agency works. It is found that credit rating agencies exhibit excess sensitivity to business cycle conditions. Besides this, most of the time ratings do not change and

(7)

Page | 6 rating agencies do not monitor the conditions of firms. Only to a greater or lesser extent at a particular time. These results suggest that credit ratings might not be optimal to use for risk management (Amato and Furfine, 2003).

2.2 Process

A company can obtain a credit rating by, prior to the issuance, contacting the bond credit rating agency and request to have a rating assigned to their new issue. These activities of credit rating agencies are of interest to issuers, investors, lenders and securities regulators. Their main objectives are protecting investors, make sure that securities markets are efficient, fair and transparent, and reducing the systemic risk. Credit rating agencies achieve this by offering an independent, informed analyses. One of the principles for the activities of a credit rating agency is that it should reduce the information asymmetry among lenders, borrowers and other market participants. Credit rating agencies reduce this information asymmetry by providing investors with a screening instrument and reveal hidden information. Information asymmetry is costly for the issuer, because investors will require a higher risk premium for their investment. Therefore reduces such a credit rating the cost of transaction for the issuer of a bond (IOSCO Statement of Principles Regarding The Activities of Credit Rating Agencies, 2003).

When a company is requesting for a credit rating, it has to deliver all kinds of information to the rating agency. Then an analytical team will be assigned to the request and will conduct basic research about the individual issue characteristics and about the company. The analytical team will meet with the issuer to obtain any additional information they consider to be relevant. Based on this information, the analyst team prepares a rating report and will present this report to the rating committee. This rating committee will review the report and after they have done that they will set a final rating decision.

The issuer will then be informed about the rating by the rating agency and will receive the rating report. The issuer can request for a rating reconsideration if it presents new information that can lead to a rating adjustment. Finally, if the issuer has chosen to purchase the rating, the rating will be made publicly available so there is no non-public information hidden by the issuer of a bond (Maher and Sen, 1997). The rating agency

(8)

Page | 7 will publish the rating via a press release and puts it on its website (Sangiorgi, Sokobin and Spatt, 2009).

2.3 Conflict of Interest

The business model of these rating agencies is that the company that issues the bond also has to pay for the credit rating of that bond. This “issuer pays” model was introduced in the early 1970’s in place of the “investor pays” business model by the three largest rating agencies (White, 2010). The issuer only pays fees for the credit rating when the credit rating is being published

These new business models where the issuer pays the rating agency created new conflict of interest. A large part of the income of credit rating agencies comes from the issuers rather than from the investors. This generated a kind of dependence from rating agencies to issuers of bonds (Mathis, McAndrews and Rochet, 2009). This results in that a rating agency keep its rating more upward in order to keep the issuer happy. This is done to prevent the issuer, the client, from moving its business to another credit rating agency. These incentives can have influence on the subjective measurements of risk (Mason and Rosner, 2007).

One major example where credit rating agencies were affected by perverse incentives is prior to the financial crisis of 2008. Important financial institutions were not allowed to have assets without the highest credit rating of one of the three major credit rating agencies. Therefore, there was a high demand by those companies for high credit ratings. As rating agencies were paid by those financial institutions, profits depend on keeping those institutions satisfied. If a credit rating agency gave a realistic rating given the associated risk with their securities while other agencies did not, profit would dive from those agencies. This resulted in that financial institutions were shopping around agencies that gave the highest rating (Crotty, 2009).

2.4 Related research

This so-called rating shopping where companies keep the highest rating and withdraw the lower rating is examined by Skreta and Veldkamp (2009). They have developed a model where issuers are able to choose between potential credit rating agencies. Despite that raters are all trying to give a honest estimate of the creditworthiness of the

(9)

Page | 8 issuer, this leads to overly optimistic credit ratings. Thus if the raters make, on average, a correct estimate and the errors are symmetrically distributed, the issuers will choose to purchase the rating that is most optimistic. Bolton, Freixas and Shapiro (2012) find as well that issuers payment influence the precision of the ratings and that the entry barriers enhance the market power for the three major rating agencies.

For companies it is of importance to keep the highest ratings, as lower ratings have negative impact on the bond returns of those companies. Studies done by Hand, Holthausen and Leftwich (1992) and Hite and Warga (1997) test the effect of credit rating changes on bond returns. They show that credit ratings that are downgraded have significant negative effect on the bond returns of companies.

Covitz and Harrison (2003) have tested in their paper if the conflict of interest for a rating agency affects its actions. They empirically examines if rating agencies systematically acts in favor of the issuer or in favor of the investor. Their analysis indicated that rating changes were anticipated by the bond market, but they found no evidence that because of the conflict of interest rating agencies act in favor of the issuers. Bongaerts, Cremers and Goetzmann (2012) examined a rating shopping hypothesis as well, which proposes that conditional on receiving a downgrade on its credit rating an issuer shops for a better credit rating. They found in their research that in many cases the likelihood of getting Fitch as additional on S&P and Moody’s strongly associated is with those two rating agencies being just below the investment grade border. Cantor and Packer (1997) found the same in their research that a third rating agency assigns on average a higher rating. Their results suggest that the observed differences in rating mainly arises from differences in rating scale. Besides, they found little evidence that issuers decide to use a third rating because of the higher rating. Jewell and Livingston (1999) found in their research that a firm with credit ratings from S&P and Moody’s will have higher ratings from those two if a Fitch rating is used as well. Furthermore, they found that if a company uses Fitch additionally to ratings of S&P and Moody’s the company will receive less downgrades and more upgrades from those two rating agencies.

(10)

Page | 9

3 Data

3.1 Data

In order to conduct this research, data has to be collected about credit ratings. The data that are collected for this research is retrieved from several databases. I have collected up- and downgrade data for oil companies, with the exact date of a credit rating change. Furthermore, I have retrieved the oil price development as well.

The time range used in the sample is from January 1st_{1990 till December 30}th_2016.

This range includes multiple price swings in oil price and gives a significant amount of credit rating up- and downgrades. The credit ratings in this sample are from the three major credit rating agencies, namely Standard & Poor’s, Moody’s and Fitch. The required data is retrieved from Compustat, Thomson One and Datastream.

The data collected from Compustat include the CUSIP codes of the oil companies and the S&P credit ratings, restricted for only North American companies.

In order to only keep companies from the oil and gas extraction sector in the sample, it was filtered by SIC codes and only companies with SIC coded between 1300 and 1399 were used. Afterwards, the Standard & Poor’s Long Term credit ratings were retrieved for the remaining companies. This results in end of month dates of a credit rating change, as Compustat cannot give the precise date. Therefore, Thomson One is used to obtain the exact credit rating change date. Thomson One gives the credit rating dates for all the credit ratings a particular company has, so the ratings with dates from S&P, Moody’s and Fitch are all retrieved from this database. These dates can be retrieved by using the CUSIP code from Compustat. The dates has to be looked up for each company separately. Afterwards, the credit rating dates in the sample were categorized into upgrades and downgrades.

The financial data for the corresponding companies in the sample is retrieved from the Compustat Capital IQ database. This database contains the quarterly firm specific data that is used for the control variables in the regression. The values of these control variables are the most recent fiscal quarter end data before the start of the shock in oil

(11)

Page | 10 price. The first control variable is Leverage. This is the total debt of a company divided by the total assets. Size is the natural logarithm of the total assets of a company, which is used as a proxy for the firms size. The variable Net Income is the ratio net income over total assets. Int_cov is generated by dividing the operating income before depreciation by the interest expense of a company. Tangibility is the net property, plant and equipment divided by the total assets. Lastly, the variable Cash stands for the cash holdings over the total assets.

Lastly, the oil prices were retrieved from Datastream. The WTI, Western Texas Intermediate, is used as an indicator of the oil price. The WTI is the underlying commodity for oil future contracts on the New York Mercantile Exchange. This benchmark is chosen, because the oil companies in the sample are all North American and this oil benchmark is used primarily in the U.S..

There are two major downward shocks in oil price that are used as events in this study. These events start at 11th_{of July 2008 and the 20}th_{of June 2014.}

The eventual sample left me with 111 companies. Table 1 shows the distribution of the credit rating up- and downgrades split up by year and by Credit Rating Agency. This table shows that S&P gives the most upgrades and downgrades, followed by Moody’s, and Fitch the least. As is shown in table 1, is that there are significant more downgrades than upgrades for all three credit rating agencies. Besides this, it shows that most of the upgrades and downgrades are concentrated in 2015 and 2016.

(12)

Page | 11

Table 1. Distribution of Up- and Downgrades over the years for the credit ratings from Standard & Poor’s, Moody’s and Fitch.

S&P Moody's Fitch

Year Downgrades Upgrades Downgrades Upgrades Downgrades Upgrades

1990-2000 18 24 1 1 1 4 2001 6 5 0 1 1 1 2002 14 4 0 0 1 0 2003 3 4 0 1 1 2 2004 3 4 0 0 0 3 2005 6 4 0 1 1 1 2006 3 6 0 0 1 2 2007 2 12 0 4 1 4 2008 0 8 1 3 0 3 2009 8 5 5 1 4 0 2010 5 8 0 4 1 1 2011 3 6 0 0 2 0 2012 16 9 5 16 3 1 2013 2 10 3 4 1 1 2014 12 16 6 14 0 3 2015 49 9 40 6 0 1 2016 119 18 61 15 11 0 Total 269 152 122 71 29 27 3.2 Descriptive Statistics

Table 2 below shows the descriptive statistics of the variables used in the regression. The whole sample for the time windows of both the first and second shock consists of 210 observations for 102 companies. The binary dependent variable has 17 times the value of 1, which indicates that 17 of the 102 companies withdraws a credit rating after a negative shock in oil price. In these two time windows after a shock there are 126 rating agencies that downgrade their rating at least once. The variable DurationDG is equivalent to 37.515 and DurationUG to 14.078 with 124 and 49 observations, respectively. This means that it takes, on average, more than 37 weeks for a rating agency to downgrade its credit rating after a shock with negative effect on the creditworthiness of a company, and happened 124 times. It takes around 60 weeks, on average, before a rating agency has its first upgrade, where this happened 49 times. The size of the downgrade is on average a little more than 1 step down the rating scale. Leverage ratio is equal to 0.309, which indicates that the firms debt equals 30.9% of the firms total assets. The size of the firms in the sample is on average 8.518, which is

(13)

Page | 12 calculated as the natural logarithm of total assets of the firm, is approximately 5 billion. The variable Net Income has a mean of 0.012 and indicate that the net income is equivalent to 1,2% of the total assets. Furthermore, the interest coverage ratio is, on average, 15.378. This value means that the average firm in the sample is able to pay off the interest associated with their outstanding debt, whereby the value of 15.378 means that a company’s operating income before depreciation is more than 15 times as much as the interest payments. The tangibility ratio is 0.745 for the average firm in the sample. This ratio indicates that 74,5% of the totals assets consists of property, plant and equipment. Lastly, the cash ratio has value of 0.041 that indicates that cash and short term investments are on average 4.1% of the total assets for the average firm.

Table 2. Descriptive Statistics

Variables N Mean Min Median Max

Std. Dev. Dependent Variable Dropped 210 0.081 0 0 1 0.273 Explanatory Variables Downgraded 210 0.600 0 1 2 0.510 DurationDG 124 37.515 0 31.4 100.4 36.341 DurationUG 49 60.335 0 0 126.2 33.374 SizeDG 124 1.010 0 1 11 1.550 Control Variables Leverage 208 0.309 0.043 0.298 0.742 0.140 Size 208 8.518 5.209 8.323 12.156 1.423 Net Income 208 0.012 -0.046 0.009 0.136 0.024 IntCov 204 15.378 -10.202 7.5 190.476 29.103 Tangibility 208 0.745 0.134 0.778 0.967 0.177 Cash 202 0.041 0 0.026 0.274 0.052

(14)

Page | 13 Table 3 on the next page show the Pearson correlation matrix with the correlations between all variables used in this model. The correlation of the dependent variable Dropped is highly significant at the 1% level with five of the nine independent and control variables. The direction of those correlations is as expected as well. The dependent variable DurationUG and the control variables IntCov, Tangibility and Cash are, however, not significant with the dependent variable Dropped. To start with, the significant independent variable DurationDG has a negative coefficient which is in line with expectations that the probability of a credit rating being dropped decreases with the time till the first downgrade. The variable SizeDG has a positive correlation with Dropped and indicates that the size of the downgrade moves with the probability of a rating being dropped. The control variable Leverage is in line with expectations as well and has a positive coefficient. This means that a higher ratio of debt over total assets moves with increased probability of a firm withdraws a credit rating. The same applies to the control variables Size and Net Income, which are both highly significant and have both a negative coefficient. This indicates that the probability of a credit rating being dropped moves in the opposite direction of the firm’s size and with the ratio of net income over total assets.

(15)

Page | 14 Ta bl e 3. Pea rs on C or rel at ion Ma tr ix D rop ped D ow ng ra de d D ura tion D G D ura tion U G Levera ge Si ze N et In com e In tCov Ta ng ibi lit y Ca sh Si zeD G D rop ped 1 D ow ng ra de d 0.165* * 1 D ura tion D G -0.296* ** -0.128 1 D ura tion U G -0.133 0.518* ** -0.105 1 Levera ge 0.298* ** 0.419* ** -0.236* ** 0.329* * 1 Si ze -0.202* ** -0.252* ** 0.232* ** -0.182 -0.480* ** 1 N et In com e -0.192* ** -0.186* ** -0.014 0.035 -0.368* ** 0.125* 1 In tCov -0.114 -0.259* ** -0.097 -0.292* * -0.513* ** 0.402* ** 0.585* ** 1 Ta ng ibi lit y 0.072 0.170* * 0.036 0.024 0.173* * -0.047 -0.232* ** -0.062 1 Ca sh -0.148 -0.09 -0.077 -0.025 -0.199* ** -0.061 0.396* ** 0.228* ** -0.457* ** 1 Si zeD G 0.290* ** 0.073 -0.029 -0.005 0.157 -0.208* * -0.174* -0.089 -0.050 -0.041 1 The sy mb ol s * ** , * *, * r epr es en t s ta tistic al sign if ic an t a t t he 1%, 5% a nd 10 % l ev els , r es pe ctive ly

(16)

Page | 15

4 Empirical Method

In order to test what the probability is of a company to drop a credit rating from one of the credit rating agencies, a simple regression model is used. This probability is tested primarily as a function of the time it takes for a credit rating agency to downgrade its credit rating. This time is measured as the amount of weeks after a exogenous shock with negative effects for the returns of oil companies.

These exogenous shocks in oil price that have a negative effect on the returns of oil companies has to be determined. The oil price development presented in figure 1 shows that there are two major downward shocks in oil price. The highest point just before the drop is of interest to use as the starting point of the downward shock. Therefore, the exact date of these points has to be specified.

To determine these dates as the starting point of the shock, I looked at what the trend of oil price over the time period of 1990 till 2017 is. This trend will produce values of the oil price accordingly to this trend. Afterwards, the actual oil prices are subtracted by the value of the trend at that moment. That will result in residual e. This residual is then divided by the standard deviation of these residuals in order to calculate a z-score. The equation is shown below.

𝑍 − 𝑠𝑐𝑜𝑟𝑒 =𝑋𝑡− E[𝑋𝑡] 𝜎

Where in this equation 𝑋_𝑡 stand for the oil price at time t and E[𝑋_𝑡] for the expected oil price given the trend line. The standard deviation is noted as 𝜎 in this equation. Thereafter, every day’s closing oil price has a certain z-score. The dates with the most significant, in this case positive, z-scores, just before the drop in oil price, are chosen as the starting dates. The two starting points are the 11th_{of July 2008 and the 20}th_{of June}

2014. The time windows used are from the starting point of the shock till two years after that particular date. This time frame is chosen as in a maximum of two years the drop of a credit rating agency can be linked to the downgrade of a credit rating after the downward shock in oil price.

(17)

Page | 16

Figure 1. Oil price development from the 1st_{of January 1990 till the 31}st_{of December 2016. The red vertical lines represent} the starting dates of the downward shocks in oil price. These dates are found by looking at the most significant z-scores just before the shock.

On Figure 1 above, the red vertical lines represent the event dates where the z-score had the highest most significant value. These are the points in time that were used as starting dates of the shock. As you can see on the figure, both starting dates are at the top of the oil price before the shock followed by a sharp drop in oil price.

To estimate what the effects on the probability of dropping a credit rating agency, the below regression model is used. The regression equation is constructed as follows:

Pr (𝐷𝑟𝑜𝑝𝑝𝑒𝑑 = 1) = 𝛽0+ 𝛽1𝐷𝑜𝑤𝑛𝑔𝑟𝑎𝑑𝑒𝑑𝑖𝑡+ 𝛽2𝐷𝑢𝑟𝑎𝑡𝑖𝑜𝑛𝐷𝐺𝑖𝑡+ 𝛽3𝐷𝑢𝑟𝑎𝑡𝑖𝑜𝑛𝑈𝐺𝑖𝑡

+ 𝛽₄𝑆𝑖𝑧𝑒𝐷𝐺 + 𝛽₅𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖𝑡+ 𝛽₆𝑆𝑖𝑧𝑒_𝑖𝑡+ 𝛽₇𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒_𝑖𝑡+ 𝛽₈𝐼𝑛𝑡_{𝑐𝑜𝑣𝑖𝑡} + 𝛽₉𝑇𝑎𝑛𝑔𝑖𝑏𝑖𝑙𝑖𝑡𝑦_𝑖𝑡+ 𝛽₁₀𝐶𝑎𝑠ℎ_𝑖𝑡+ 𝜀_𝑖

The dependent variable Dropped in this model is a binary variable that equals 1 if a company drops a credit rating agency and 0 if it does not drop a credit rating agency. The main explanatory variables are Downgraded , DurationDG and DurationUG. The variable Downgrade is a dummy variable that equals 1 if a credit rating agency has a downgrade in the two year time window. The other explanatory variable DurationDG stands for the amount of trading days from the negative shock in oil price until the first downgrade by a credit rating agency. Whereas, the variable DurationUG stands for the amount of weeks from the start of the shock until a company has its first credit rating upgrade. The variable SizeDG measures the size of the downgrade in credit rating. So it

0 50 10 0 15 0 O lie pr ijs 1/1/1990 1/1/1995 1/1/2000 1/1/2005 1/1/2010 1/1/2015 Date

(18)

Page | 17 displays the amount of steps down the rating scale after the downgrade. Furthermore, the control variables are included in the model.

In order to estimate the effects of competition on how a rating agency behaves after the shock in oil price, the following regression model is used:

Pr (𝐷𝑟𝑜𝑝𝑝𝑒𝑑 = 1) = 𝛽₀+ 𝛽₁𝐷𝑜𝑤𝑛𝑔𝑟𝑎𝑑𝑒𝑑_𝑖𝑡+ 𝛽₂𝐷𝑢𝑟𝑎𝑡𝑖𝑜𝑛𝐷𝐺_𝑖𝑡+ 𝛽₃𝑆𝑖𝑧𝑒𝐷𝐺

+ 𝛽₄𝐷𝑢𝑚𝑚𝑦𝐶𝑅𝐴 + 𝛽₅𝐷𝑢𝑚𝑚𝑦𝐶𝑅𝐴 ∗ 𝐷𝑜𝑤𝑛𝑔𝑟𝑎𝑑𝑒𝑑 + 𝛽₆𝐷𝑢𝑚𝑚𝑦𝐶𝑅𝐴

∗ 𝐷𝑢𝑟𝑎𝑡𝑖𝑜𝑛𝐷𝐺 + 𝛽7𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 ∗ 𝐷𝑜𝑤𝑛𝑔𝑟𝑎𝑑𝑒𝑑 + 𝛽8𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒

∗ 𝐷𝑢𝑟𝑎𝑡𝑖𝑜𝑛𝐷𝐺 + 𝛽9𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 ∗ 𝑆𝑖𝑧𝑒𝐷𝐺 + 𝛽10𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 ∗ 𝐷𝑢𝑚𝑚𝑦𝐶𝑅𝐴

+ 𝛽_𝑖𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝜀_𝑖

This regression equation is the same as the previous one, except that a dummy variable and multiple interaction variables are included. The dummy variable DummyCRA takes value of 1 if a company has more than one credit rating. This dummy variable is interacted with the core explanatory variables in order to estimate what the effect of competition is on these variables. In Addition, the variable Leverage is interacted with these core explanatory variables as well to test what the effect of leverage is. The control variables used are the same as in the previous regression formula.

(19)

Page | 18

5 Results

In this section the empirical results of the regression outputs are presented and discussed. First, the effect on the probability of a credit rating agency being dropped as a function of time till the first downgrade is tested. Second, the effect of competition on the probability of dropping a rating agency is estimated.

5.1 Primary results

Table 4 on the next page shows the regression output of the merged sample. This sample consists of all observations for both two shocks in oil price. It measures the impact of the duration, in weeks, it takes for a rating agency to downgrade its rating after the negative shock, the amount of weeks until the first upgrade, if a rating agency have downgraded its credit rating, the size of the downgrade and various control variables on the probability of having a credit rating dropped. The underlying sample of this regression only includes companies that have more than one credit rating. The regression results include in general the coefficients for all variables, the standard errors in parentheses and the how much the model fits its data. Column (1) only includes the dummy variable Downgraded which has the value of 1 if a rating agency has downgraded its data, and all control variables in regression model. In column (2), (3) and (4) the duration till the first downgrade, the duration till the first upgrade and the size of the downgrade are included in the model, respectively.

In column 2 and 3, the coefficient of the variable Downgraded is significant at the 5% and 1% level, respectively, and indicates that a credit rating being dropped is positively related with probability of the credit rating being dropped. The amount of trading days till a credit rating is being downgraded is negatively related to the probability of a credit rating being dropped. This is significant at the 5% and 1% levels as well, and is as expected and pro the hypothesis. In the fourth column, the size of the downgrade is included as well. The coefficient of this variable positive and highly significant at the 1% level, and has value of 0.0397. This means that for each step down on the rating scale after a downgrade in credit rating change, the probability of dropping that credit rating increases with nearly 4%. This variable implicitly says that a credit rating is

(20)

Page | 19 downgraded by a rating agency. The dummy variable Downgraded moves after including the downgrade size from highly significant to not significant at all. The leverage ratio remains in all three regressions highly significant and has a positive coefficient. This means that a higher level of debt over total assets results in higher probability of a credit rating being dropped.

Table 4. The regression output of the effects of duration till first downgrade on the probability of a credit rating being dropped for the two samples of both oil shocks combined.

This table shows the results from the regression model examining the effects of the duration from the beginning of the shock until the first downgrade of a credit rating and multiple control variables on the dependent variable, the probability of a credit rating being dropped. The underlying dataset of this regression includes data of both two shocks in oil price. Column (1) shows the model where the effect of a downgrade on the probability of being dropped is estimated. In column (2), the amount of trading days till the first downgrade of a credit rating is included in the regression. Column (3) includes the amount of weeks from the beginning of the shock till the first upgrade. Lastly, in column (4) the size in rating scale of the downgrade is included in the model. All four models include credit rating agency fixed effects. The standard errors are in parentheses. The symbols ***, **, * represent statistical significant at the 1%, 5% and 10% levels, respectively.

(1) (2) (3) (4)

Dropped Dropped Dropped Dropped

Downgraded 0.0173 0.142** 0.186*** 0.0978 (0.0400) (0.0690) (0.0582) (0.0689) DurationDG -0.00204** -0.00268*** -0.00217** (0.000924) (0.000786) (0.000901) SizeDG 0.0397*** (0.0131) DurationUG -0.000928* -0.000927* (0.000499) (0.000539) Leverage 0.464*** 0.414** 0.481*** 0.405** (0.172) (0.172) (0.146) (0.169) Size -0.0119 -0.00946 -0.456 -0.00316 (0.0172) (0.0170) (0.695) (0.0167) Net Income -1.583 -1.542 0.000814 -0.990 (0.994) (0.984) (0.000752) (0.975) IntCov 0.00155* 0.00130 -0.0263 0.000877 (0.000883) (0.000881) (0.0964) (0.000867) Tangibility -0.0940 -0.0719 -0.508 -0.0304 (0.119) (0.118) (0.364) (0.116) Cash -0.547 -0.555 -0.00189 -0.481 (0.426) (0.422) (0.0139) (0.411) Constant 0.109 0.0937 -0.0324 0.0130 (0.196) (0.194) (0.159) (0.191)

Credit rating agency fixed

effects Yes Yes Yes Yes

Observations 234 234 234 234

R-squared 0.096 0.119 0.157 0.174

(21)

Page | 20

5.2 Secondary results

The regression in this section will be more focused on the effects of competition between rating agencies on the probability that a company drops a credit rating.

Therefore, the underlying sample consists of all companies that have one or more credit ratings. Besides, the leverage ratio is interacted with the core explanatory variables to see what the effect this company specific value is on those variables. Table 5 on the next page shows the regression results, where in column (1) only the core explanatory variables are included. The second column includes the dummy variable DummyCRA that takes value of 1 if a company has more than one credit rating and 0 if only one credit rating, and the interaction variables between this dummy variable and the

explanatory variables. In column (3) the interaction variables between leverage and the explanatory variables are included. The regression in column (4) has all variables included.

The results of column (1) show, just like in table 4, that the a downgrade in credit rating positively related is with probability of a credit rating being dropped and the amount of weeks negatively related. The size of the downgrade has as well a positive coefficient that indicates that a bigger downgrade increases the probability of that credit rating being dropped. All coefficients are highly significant. Regression of column (2) yields no significant coefficients, except for the size of the downgrade and the control variable Leverage. The interaction variables between the dummy variable if a company has more than one credit rating are all not significant. This means that competition does not have significant influence on rating agencies acting more in favor of the issuer and, therefore, make difference in the probability of being dropped. Column (3) shows what the effects of leverage are on the core variables. The interaction variable between Leverage and Downgraded is highly significant at the 5% level and indicates that companies with high leverage that get downgraded have increased probability of dropping a credit rating. The interaction between the leverage ratio and the amount of weeks till the first

downgrade has a negative coefficient and is significant at 1%. This means that a delay in downgrading a credit rating has increased effect in decreasing the probability of being dropped for companies with high levels of leverage.

(22)

Page | 21

Table 5. The regression output of the effects on the probability of a credit rating being dropped for the two samples of both oil shocks combined.

This table shows the results for the effect of competition on the probability of dropping a credit rating as well as the interaction variables between leverage and the core explanatory variables. Column (1) shows the model where the effect of a downgrade on the probability of being dropped is estimated. In column (2), the dummy variable that equals 1 if a company has more than one credit rating is included along with the interaction variables of this dummy variable. Column (3) includes the interaction variables of leverage. Lastly, in column (4) all variables are included in the model. All four models include credit rating agency fixed effects. The standard errors are in parentheses. The symbols ***, **, * represent statistical significant at the 1%, 5% and 10% levels, respectively.

(1) (2) (3) (4)

Downgraded 0.136** -0.0304 -0.224 -0.325 (0.0595) (0.163) (0.175) (0.224) DurationDG -0.00275*** -0.000130 0.00361 0.00476 (0.000775) (0.00226) (0.00230) (0.00291) SizeDG 0.0360*** 0.0382*** -0.0961* -0.0907* (0.0115) (0.0117) (0.0501) (0.0506) DummyCRA 0.0467 0.0780 (0.0591) (0.0983) DummyCRA*Downgraded 0.181 0.129 (0.171) (0.173) DummyCRA*DurationDG -0.00299 -0.00177 (0.00243) (0.00238) Leverage*Downgraded 1.137** 1.091** (0.486) (0.493) Leverage*DurationDG -0.0167*** -0.0157** (0.00626) (0.00638) Leverage*SizeDG 0.280** 0.272** (0.110) (0.111) Leverage*DummyCRA -0.171 (0.313) Leverage 0.428*** 0.427*** 0.0148 0.153 (0.144) (0.145) (0.224) (0.332) Size 0.00303 0.00397 0.00546 0.00635 (0.0138) (0.0138) (0.0134) (0.0135) Net Income -0.489 -0.504 -0.150 -0.131 (0.682) (0.687) (0.663) (0.678) IntCov 0.000798 0.000756 0.000144 8.00e-05 (0.000740) (0.000745) (0.000757) (0.000771) Tangibility -0.0502 -0.0422 -0.00748 -0.00943 (0.0941) (0.0962) (0.0915) (0.0942) Cash -0.548 -0.540 -0.470 -0.488 (0.358) (0.361) (0.347) (0.352) Constant -0.0551 -0.105 0.00387 -0.0635 (0.158) (0.164) (0.154) (0.172)

R-squared 0.180 0.190 0.250 0.256

(23)

Page | 22 The interaction between leverage and the size of the downgrade yield a positive and significant coefficient that indicates that the effect of size of the downgrade on the probability of the credit rating being dropped is higher for companies with higher levels of leverage. The significance of the core explanatory variables, on the other hand, move from highly significant in the first regression to not significant at all when the

(24)

Page | 23

6 Robustness Test

The regression analysis from table 4 shows the estimation results of the effects on the probability of a credit rating being dropped. The underlying sample for that regression consists of a merged sample of the combined samples for both two shocks in oil price. Thus, table 4 does not make a distinction between the two shocks in oil price. Therefore, as a robustness check, the same test is done on two separate samples where the

underlying set of companies are split up for each of the two events. This is done to see whether the main findings are not driven by specific events. The time window of the two samples are two years after the start of the shocks beginning at the 11th_{of July 2008}

and the 20th_{of June 2014. First, the separate tests are done to estimate the effects of a}

delay in downgrade on the probability of a credit rating being dropped. Second, the two separate tests are done to estimate the effect of competition on the delay in credit rating downgrade and, therefore, on the probability of a rating being dropped.

6.1 Estimation of delay in downgrade

In this section the robustness test for the regression of table 4 is done. It is tested if the effects of a delay in downgrade, the time till the first upgrade and the size of the

downgrade have different effects on the probability of a rating being dropped for both separate shocks in oil price.

6.1.1 Regression results for the first oil shock

Table 6 on the next page presents the regression output for the first shock in oil price starting at the 11th_{of July 2008. The same regression is performed as in the regression}

of table 4 with a time window of two years.

The results in the second, third and fourth column suggest that a credit rating that has been downgraded is positively linked with probability of that credit rating being dropped at the 1% significance level. This is in line with expectation that rating agencies that downgrade their credit rating have increased probability in being dropped. This coefficient moves from not significant to significant at the 1% level after including the amount of time until first downgrade and stays significant after time till the first

(25)

Page | 24 upgrade and size of downgrade are included in the model. The coefficient of 0.181 of the dummy variable Downgraded in column 3 shows that the probability of a credit rating being dropped is 18.1% higher if a rating agency downgrades its credit rating than if it does not downgrades its rating.

Table 6. The regression output of the effects of duration till first downgrade on the probability of a credit rating being dropped after the first shock in oil price.

This table shows the results from the regression model examining the effects of the duration from the beginning of the shock until the first downgrade of a credit rating and multiple control variables on the dependent variable, the probability of a credit rating being dropped. The underlying dataset of this regression includes all North-American oil companies that had a credit rating in the two-year period from 11-07-2008 till 11-07-2010. Column (1) shows the model where the effect of a downgrade on the probability of being dropped is estimated. In column (2), the amount of trading days till the first downgrade of a credit rating is included in the regression. Column (3) includes the amount of weeks from the beginning of the shock till the first upgrade in the model. Lastly, in column (4) the size in rating scale of the downgrade is included in the model. All four models include credit rating agency fixed effects. The standard errors are in parentheses. The symbols ***, **, * represent statistical significant at the 1%, 5% and 10% levels, respectively.

(1) (2) (3) (4)

Downgraded 0.0513 0.184*** 0.181*** 0.217*** (0.0367) (0.0562) (0.0570) (0.0654) DurationDG -0.00404*** -0.00400*** -0.00290* (0.00136) (0.00138) (0.00169) SizeDG -0.0880 (0.0791) DurationUG -0.000393 -0.000357 (0.00102) (0.00101) Leverage 0.191 0.183 0.206 0.192 (0.177) (0.164) (0.176) (0.176) Size 0.0209 0.0159 -1.473 0.0140 (0.0164) (0.0153) (0.901) (0.0156) Net Income -1.185 -1.421 0.000804 -1.179 (0.947) (0.882) (0.000776) (0.937) IntCov 0.000597 0.000739 -0.121 0.000639 (0.000807) (0.000750) (0.0965) (0.000788) Tangibility -0.0733 -0.126 0.0619 -0.0882 (0.100) (0.0947) (0.336) (0.101) Cash 0.0935 0.0665 0.0152 0.0649 (0.359) (0.333) (0.0156) (0.335) Constant -0.172 -0.0834 -0.0837 -0.0929 (0.184) (0.173) (0.175) (0.174)

R-squared 0.118 0.256 0.259 0.279

(26)

Page | 25 The coefficients of the variable DurationDG, which measures the amount of weeks between the beginning of the oil shock till the first downgrade, is in both the second as

the third regression negative and highly significant at the 1% level with value of -0.004 in the third regression. Thus, delaying one trading day in downgrading the credit

rating results in decreased probability of being dropped. This coefficient is, as well, pro expectations and pro the hypothesis that a credit rating agency who delays a downgrade on its rating has lower probability of being dropped. The firm specific control variables are from the first regression in column 1 insignificant and remain insignificant after in column 2, 3 and 4 more explanatory variables are included. Including the size of the downgrade and the time till the first upgrade after the shock does not result in more significant results. This is in contrast with the regression results of column 4 of table 4 where the size of the downgrade is highly significant and the duration till the first upgrade mildly significant. The core regression results behave, on the other hand, the same as in table 4.

6.1.2 Regression results for the second oil shock

Table 7 presents the regression results for the second oil price shock. The same regression model is used as in the regression of table 4. The difference is the time window that is used, namely in table 7 the two-year time window from the beginning of the second drop in oil price at the 20th_{of June 2014 till the 20}th_{of June 2016 is used.}

It can be concluded from the table below that there is a positive relationship between a rating agency that downgrades its credit rating and the rating agency being dropped. In the second and third column, the variable Downgraded has a positive coefficient and is significant at the 5% level. In the third regression where the time till the first upgrade is included, it has a coefficient of 0.242 which indicates that the probability of a credit rating being dropped increases with 24.2% if that rating is downgraded. The effect of trading days until the first downgrade after a shock is in all three regressions negative and highly significant at the 1% level. This means that if credit rating agencies delay their downgrading and, therefore, have more weeks between the shock and the downgrade have less probability of being dropped by a company. This is in line with expectation and pro hypothesis that a credit rating agency that delays the downgrade of a credit rating has lower probability of being dropped by a company. The coefficient in

(27)

Page | 26 the third column suggests that the probability of a credit rating being dropped decreases with 0.00377% for every week a downgrade is delayed. Just as in the regression of table 4, the size of the downgrade is positive and highly significant which means that a the probability of the credit rating being dropped increases with the size of the downgrade.

Table 7. The regression output of the effects of duration till first downgrade on the probability of a credit rating being dropped after the second shock in oil price.

This table shows the results from the regression model examining the effects of the duration from the beginning of the shock until the first downgrade of a credit rating and multiple control variables on the dependent variable, the probability of a credit rating being dropped. The underlying dataset of this regression includes all North-American oil companies that had a credit rating in the two-year period from 20-06-2014 till 20-06-2016. Column (1) shows the model where the effect of a downgrade on the probability of being dropped is estimated. In column (2), the amount of trading days till the first downgrade of a credit rating is included in the regression. Column (3) includes the amount of weeks from the beginning of the shock till the first upgrade in the model. Lastly, in column (4) the size in rating scale of the downgrade is included in the model. All four models include credit rating agency fixed effects. The standard errors are in parentheses. The symbols ***, **, * represent statistical significant at the 1%, 5% and 10% levels, respectively.

(1) (2) (3) (4)

Downgraded -0.0266 0.231** 0.242** 0.178 (0.0621) (0.108) (0.107) (0.108) DurationDG -0.00366*** -0.00377*** -0.00371*** (0.00126) (0.00126) (0.00123) SizeDG 0.0384** (0.0151) DurationUG -0.00108 -0.00105 (0.000662) (0.000649) Leverage 0.483* 0.283 0.314 0.305 (0.253) (0.256) (0.255) (0.250) Size -0.0209 -0.0199 -1.062 -0.0111 (0.0240) (0.0234) (1.538) (0.0230)

Net Income -1.841 -1.662 -6.98e-05 -0.651

(1.544) (1.503) (0.00156) (1.515) IntCov 0.000996 0.000137 -0.0994 -0.000223 (0.00158) (0.00156) (0.189) (0.00153) Tangibility -0.210 -0.145 -1.033 -0.0568 (0.192) (0.188) (0.654) (0.186) Cash -1.167* -1.142* -0.0196 -0.945 (0.673) (0.654) (0.0232) (0.641) Constant 0.340 0.343 0.306 0.192 (0.290) (0.282) (0.281) (0.279)

R-squared 0.107 0.161 0.179 0.218

(28)

Page | 27 Furthermore, all of the control variables in the regression of column 3 are not significant at all. The leverage ratio and the cash ratio in the first regression, where only the Downgraded dummy variable is included, are significant at the 10% level. After including the duration till the first downgrade in the regression, the Leverage variable moves to not significant anymore and stays at this level at the further regressions. The regression of table 7 yield similar results as in the regression of table 4. The core regression coefficients have the same relationship regarding the probability of dropping of a credit rating agency and the coefficient are of nearly the same size as in table 4. The regressions of table 6 and 7 can be interpreted as evidence of validity of the results of the regression of table 4. The results of both shocks in oil price taken as separate underlying samples yield the same results as those samples merged into one. It can, therefore, be concluded that the main findings are not driven by specific events.

6.2 Estimation of effects of competition

This section investigates if the effect of competition between rating agencies differs from the results of table 5 if it is tested for both shocks in oil price separately. First the results for the first shock is tested followed by the results for the second shock.

6.2.1 Competition effects for first shock

The regression results of table 8 present the estimation of the effects of competition on the duration till a rating agency downgrades its credit rating and the size of the

downgrade. In addition, it shows the effect of leverage on these variables as well. This regression is for the two year sample of the first shock in oil price starting at the 11th_of

July 2008. The interaction variables between the dummy if a company has more than one credit rating are all not significant at all. This indicates that competition between rating agencies does not have influence on the behavior of the rating agencies. The interaction between leverage and the core explanatory variables yield in both the third and fourth column highly significant results and behave in the same way as in the regression with the merged sample. Except the interaction with the size of the downgrade. In this regression it has a negative value that indicates that for the first shock in oil price the effect of the size in downgrade results in lower probability that a company will drop a credit rating if it has higher levels of leverage.

(29)

Page | 28

Table 8. The regression output of the effects on the probability of a credit rating being dropped for the first shock in oil price.

This table shows the results for the effect of competition on the probability of dropping a credit rating as well as the interaction variables between leverage and the core explanatory variables for the two year period after the first shock in oil price from 11-07-2008 till 11-07-2010 . Column (1) shows the model where the effect of a downgrade on the probability of being dropped is estimated. In column (2), the dummy variable that equals 1 if a company has more than one credit rating is included along with the interaction variables of this dummy variable. Column (3) includes the interaction variables of leverage. Lastly, in column (4) all variables are included in the model. All four models include credit rating agency fixed effects. The standard errors are in parentheses. The symbols ***, **, * represent statistical significant at the 1%, 5% and 10% levels, respectively.

(1) (2) (3) (4)

Downgraded 0.217*** 0.375 -1.851*** -2.623*** (0.0537) (0.250) (0.330) (0.540) DurationDG -0.00300** -0.0105 0.0256*** 0.0361*** (0.00115) (0.00945) (0.00558) (0.0108) SizeDG -0.0775** -0.106 0.584*** 0.963*** (0.0385) (0.0679) (0.181) (0.330) DummyCRA 0.0177 -0.0121 (0.0373) (0.0628) DummyCRA*Downgraded -0.151 0.432 (0.231) (0.277) DummyCRA*DurationDG 0.00786 -0.0111 (0.0100) (0.0123) Leverage*Downgraded 6.545*** 7.568*** (1.036) (1.121) Leverage*DurationDG -0.0976*** -0.0989*** (0.0206) (0.0232) Leverage*SizeDG -1.939*** -2.944** (0.699) (1.110) Leverage*DummyCRA 0.111 (0.245) Leverage 0.0943 0.106 0.0759 0.00436 (0.127) (0.129) (0.116) (0.217) Size 0.0141 0.0137 0.0129 0.0137 (0.0119) (0.0121) (0.00943) (0.00922) Net Income -0.412 -0.435 -0.248 -0.431 (0.429) (0.449) (0.339) (0.367) IntCov 0.000145 0.000148 -0.000120 -4.40e-05 (0.000496) (0.000511) (0.000401) (0.000418) Tangibility -0.0803 -0.0726 -0.0328 -0.0482 (0.0752) (0.0777) (0.0613) (0.0607) Cash -0.0319 -0.00534 0.302 0.325 (0.271) (0.280) (0.223) (0.222) Constant -0.0814 -0.0995 -0.110 -0.0993 (0.135) (0.139) (0.106) (0.114)

R-squared 0.254 0.267 0.560 0.611

(30)

Page | 29 The core explanatory variables are in the first column all significant and all in the

direction what is expected. After including the interaction variables these core

explanatory variables stay highly significant, but the coefficient moves in the opposite direction so that these are opposite of expectation and hypothesis. These results could indicate that there are event specific factors that influence the findings, but the sample size for this regression is very minimal which could explain the difference from the regression of the merged sample.

6.2.2 Competition effects for second shock

The regression results for the two year sample of the second shock in oil starting at the 20th_{of June 2014 is presented in table 9. The core explanatory variables of the}

regression in column 1 behave in the same manner as in the regression of the merged sample and is pro hypothesis. This regression shows as well that there is no competition effect visible. The estimates of the interaction variables with the leverage ratio behave indeed the same, even if all variables are included in the sample. The results of table 9 are all nearly the same as in the regression of table 5 with the merged sample. There are, therefore, no clearly event specific factors that influence the findings.

(31)

Page | 30

Table 9. The regression output of the effects on the probability of a credit rating being dropped for the second shock in oil price.

This table shows the results for the effect of competition on the probability of dropping a credit rating as well as the interaction variables between leverage and the core explanatory variables for the two year period after the first shock in oil price from 20-06-2014 till 20-06-2016 . Column (1) shows the model where the effect of a downgrade on the probability of being dropped is estimated. In column (2), the dummy variable that equals 1 if a company has more than one credit rating is included along with the interaction variables of this dummy variable. Column (3) includes the interaction variables of leverage. Lastly, in column (4) all variables are included in the model. All four models include credit rating agency fixed effects. The standard errors are in parentheses. The symbols ***, **, * represent statistical significant at the 1%, 5% and 10% levels, respectively.

(1) (2) (3) (4)

Downgraded 0.156 0.0310 -0.307 -0.581 (0.0988) (0.288) (0.278) (0.384) DurationDG -0.00328*** -0.00110 0.00400 0.00745 (0.00113) (0.00369) (0.00346) (0.00477) SizeDG 0.0337** 0.0350** -0.103* -0.113* (0.0138) (0.0141) (0.0620) (0.0628) DummyCRA 0.0645 0.216 (0.114) (0.174) DummyCRA*Downgraded 0.125 0.322 (0.298) (0.318) DummyCRA*DurationDG -0.00237 -0.00426 (0.00386) (0.00393) Leverage*Downgraded 1.271* 1.211* (0.715) (0.723) Leverage*DurationDG -0.0174** -0.0161* (0.00855) (0.00866) Leverage*SizeDG 0.293** 0.321** (0.136) (0.138) Leverage*DummyCRA -0.523 (0.468) Leverage 0.418* 0.438* -0.0723 0.315 (0.219) (0.223) (0.410) (0.539) Size -0.00359 -0.000682 0.00142 0.00591 (0.0199) (0.0205) (0.0197) (0.0202) Net Income -0.800 -0.722 0.00715 -0.00133 (1.394) (1.411) (1.385) (1.402) IntCov 0.000217 0.000143 -3.50e-05 -0.000382 (0.00146) (0.00147) (0.00151) (0.00154) Tangibility -0.0985 -0.0995 -0.0731 -0.0786 (0.145) (0.154) (0.142) (0.152) Cash -1.021* -1.018* -0.964* -0.997* (0.564) (0.585) (0.550) (0.569) Constant 0.0960 0.0130 0.170 -0.0241 (0.240) (0.267) (0.241) (0.283)

R-squared 0.196 0.201 0.258 0.271

(32)

Page | 31

7 Discussion

The results of the regression to estimate the effect of a delay in downgrading the credit rating by a rating agency on the probability that a credit rating is being dropped, indicate that a delay is positively linked with the probability of being dropped. Besides the amount of trading days till the first downgrade, if a company has a downgrade on its credit ratings at all and the amount of weeks until the first upgrade in credit rating are included in the estimation. Although these are factors that are of influence on the probability of a credit rating being dropped, there are more unobserved reasons why a company chooses to drop a particular rating agency above another. Such reason could be behind the scenes meetings with the rating agencies where the bargaining outcome has influence on the decision to drop a particular rating agency.

The two underlying samples for the regressions of both the first shock and second shock in oil price include 32 and 70 companies, respectively. Those two samples are constructed for companies with more than one credit rating, where the major part consist of companies that do not withdraw a credit rating in the time window from the beginning of the shock till two years afterwards. Especially during the time window of the first shock in oil price there are three companies that drop a credit rating. This results in having a small N and could therefore give a lack of power for the regression. The same applies for the regression to test the effect competition. The sample for the first shock consist of 49 companies and the second shock of 91 companies, where the during the time window of the first shock there still are only three ratings that were dropped. Furthermore, the samples are restricted for oil companies only. So the sample is selected and the results are, therefore, only applicable for actions of rating agencies on oil companies.

(33)

Page | 32

8 Conclusion

This research examined what the effect of delaying a downgrade of credit rating by a rating agency is on the probability of the credit rating being dropped and what the effect of competition between rating agencies is. Because of the issuer-pay business model of credit rating agencies, it is in interest of a rating agency to keep the issuer happy so it will not withdraw its credit rating. This resulted in a conflict of interest for rating agencies where on the one side they are dependent on issuers of credit ratings for revenues and on the other hand to give an objective credit-risk analysis. Therefore, credit rating agencies tend to delay downgrading their credit rating after a negative shock in order to prevent that their rating will be dropped. The research into this topic is scarce, as most of the previous literature investigates which rating is higher after an issuer uses two of more credit ratings and if because of the conflict of interest the rating agency acts in favor of the issuer.

This research performs a regression to test the effects of delaying a downgrade after a negative shock in oil price on the probability of that rating being dropped by the issuer. I find that the probability decreases with the amount of trading days the rating agency delays the downgrade. In the period of 1990 till 2017 there are two major drops in oil price. This regression model is used for both two shocks and yield highly significant results that are pro the hypothesis of decreased probability of the rating being dropped when the downgrade is delayed. The regression results show, additionally, that having a downgrade in credit rating results in increased probability of being dropped by the issuer. Therefore, it can be concluded that it is in favor of a rating agency to delay a downgrade in credit rating as the probability decreases that it loses the issuer as a client. There is no clear evidence found about the effect of competition on the behavior of rating agencies.

Limitations of this research are that only oil companies are used as underlying companies in the sample. Besides, there are various other unobserved factors that could influence the decision in dropping a particular credit rating.

(34)

Page | 33

Reference list

Amato, J & Furfine, C (2004), “Are credit ratings procyclical”, Journal of Banking & Finance, p. 2641-2677.

Becker, B. & Milbourn, T. (2011), “How did increased competition affect credit ratings?”, Journal of Financial Economics, p. 493-514.

Bolton, P & Freixas, X. & Shapiro, J. (2012), “The Credit Ratings Game”, The Journal of Finance, p. 85-111.

Bongaerts, D., Cremers, M., Goetzmann, W. (2012), “Tiebreaker: Certification and Multiple Credit Ratings”, The Journal of Finance, p. 113-152.

Cantor, R. & Packer, F. (1997), “Differences of opinion and selection bias in the credit rating industry”, Journal of Banking and Financ, p. 1395-1417.

Covitz, D. & Harrison, P. (2003), “Testing Conflicts of Interest at Bond Ratings Agencies with Market Anticipation: Evidence that Reputation Incentives Dominate”, Federal Reserve Board, Available at SSRN:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=512402

Crotty, J. (2009), “Structural causes of the global financial crisis: a critical assessment of the ‘new financial architecture’”, Cambridge Journal of Economics, 33, p. 563-580.

Dittrich, F. (2007), “The Credit Rating Industry: Competition and Regulation”, University of Cologne - Department of Economics, Available at SSRN:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=991821

Frost, C. (2007), “Credit Rating Agencies in Capital Markets: A Review of Research Evidence on Selected Criticisms of the Agencies”, Journal of Accounting, Auditing & Finance, p. 469-492.

Hand, J & Holthausen, R.W. & Leftwich, R (1992), “The Effect of Bond Rating Agency Announcements on Bond and Stock Prices”, The Journal of Financ, p. 733-752.

The Impact of Credit Rating Downgrades and Rating Agency Competition

Master Thesis