Credit rating changes and stock prices

(1)

CREDIT RATING CHANGES AND

STOCK PRICES

University of Amsterdam, Amsterdam Business School

Maser International Finance

Master Thesis

Author: Amanda Schaap

Student number: 10845313

Supervisor: Razvan Vlahu

Date: August 2017

(2)

Page | 1

Acknowledgement

I would like to take this opportunity to express my gratitude to my supervisor Razvan Vlahu. I am thankful to him for sharing his expertise, for his guidance along the way and the help where needed in the last phase of this Master study.

I would also like to thank my husband for supporting and encouraging me throughout the master study and during the process of writing this thesis.

(3)

Page | 2

1. Introduction

The role of credit ratings in the past few decades have become increasingly important in the financial market. A credit rating serves investors with increased insights in the credit risk profile of their investments, by indicating the creditworthiness of a company, a security, a money market instrument or a sovereign entity. Originated in the beginning of the twentieth century to provide information on customers and suppliers, the credit rating industry has developed into a fundamental aspect of modern finance. With a tendency to serve investors with increased insights on credit risk, credit rating agencies (CRAs) at the same time also act like “gatekeepers” to the capital market. Most of the corporate bonds that are issued have a credit rating (many have more than 1 rating) in order to access funding from capital markets. Depending on the level of rating, the cost for borrowing funds from the market is determined. In view of this one can imagine the importance of having a high credit rating on issued debt, especially of having a rating above investment grade. Many institutional investors like banks, insurance companies and pension funds have to comply with governmental rules and regulation that often restrict their investments to be in investment grade issues or higher. In order to guarantee a certain level of quality, these institutions may only use Nationally Recognized Statistical Rating Organizations (NSRSOs) which have been certified by the Securities and Exchange commission (SEC) and authorized to rate federal and state regulated entities (Scalet and Kelly, 2012). With only a hand full of NSRSOs, S&P and Moody’s being the two market leaders in the industry, these CRAs have a very powerful position in influencing the capital flow of funding. Nevertheless, the debacles with Enron in 2001 and Worldcom in 2002 have shown that credit ratings granted by the CRAs may in fact not correctly reflect the credit risk involved in those companies. The 2008 financial crisis has only strengthened the negative view on CRAs, as many prime AAA rated securities defaulted and the entire financial market collapsed as a result. While the past crisis itself was caused by a combination of factors rather than inaccurate credit ratings, many commentators agree that CRAs contributed to the severity of it and its widespread effects. Moreover, many observers question the efficacy of CRAs to accurately determine the default risk involved.

How could it happen that CRAs, being heavily regulated by the SEC, substantially misclassified the credit risk profile of these firms and so many financial securities? A prevailing explanation for the big failures of the CRAs is the issuer-paid model. In this business model the issuer of debt pays for the rating rather than the investors using those ratings. Because CRAs derive their revenues from these debt issuers, they are incentivized to act in the best interest of their clients and as such to inflate the credit ratings they assign, in order to satisfy their existing customers but also to attract new business.

(5)

Page | 4 Prior research find evidence that credit ratings have understated the credit risk involved in rated firms and securities during booming times, when investors are more trusting and the risk of reputational damage for the CRAs is low (Bolton et al., 2012). Alp (2013) shows that in the period 2002 to 2007 a structural break in more stringent credit ratings exists. Many theories have been brought forward to improve the current business model. One of the alternatives is for example an investor paid-model in order to resolve the conflict of interest that exists in the issuer-paid model. While prior studies show that investor-paid ratings appear to be more accurate than issuer-paid ratings (Cornaggia and Cornaggia, 2013; Xia, 2014), other research shows that investor-paid CRAs will not gain market share as long as the issuer-paid CRAs continue to operate in the market (Bongearts, 2014).

A question raised in the academic literature is whether a change in credit rating truly contains information value. According to the efficient market theory of Fama (1970), all information available is incorporated and reflected in the stock price. Assuming that rating agencies such as Standard & Poor (S&P) and Moody’s base their credit ratings for a company on information publicly available, one would expect that a rating change does not have an impact on stock prices. Nevertheless, research has shown that changes in credit ratings in fact do have an impact on stock prices and thus they are considered to convey new information to the market (Griffin and Sanvicente, 1982; Holthausen and Leftwich, 1986; Hand et al., 1992; Goh and Ederington, 1993; Dichev and Piotroski, 2001 and Boot et al., 2006).

Despite of insight in the issues that the issuer-paid model brings, a possible solution to the problem has not yet been found. Up to now the issuer paid-model continues to dominate the market and many investors, companies as well as sovereign entities are highly depended on the credit ratings they get served. In an interview Thomas Friedman brilliantly stated:

“There are two superpowers in the world today in my opinion. There’s the United States and there’s Moody’s Bond Rating Service. United States can destroy you by dropping bombs, and Moody’s can destroy you by downgrading your bonds. And believe me, it’s not clear sometimes who is more powerful” (Interview with Jim Lehrer, February 1996).

This paper examines the informational content of a credit rating and focuses on how the effect of a change in credit rating on stock prices evolved after the financial crisis of 2008. As such this paper contributes to the existing literature that is based on a period before or during the financial crisis, while I compare these periods with data from after the financial crisis. Standard event study methodology is used in order to find whether a change in credit rating leads to abnormal return

(6)

Page | 5 performance in stock prices. The analysis is extended with a multivariate regression analysis to find what firm specific characteristics might explain abnormal returns. This paper is structured as follows: section 2 provides the background information on the credit rating industry and its history. Section 3 consists of a review on existing literature and theories on the effect of CRAs and their ratings. Section 4 presents the hypotheses to be tested as well as the methodology and data used in the analysis. Section 5 outlines the results of the event study and multivariate regression analysis, while section 6 concludes.

(7)

Page | 6 3,4%

13,0% 34,4% 49,1%

Others Fitch Moody's S&P

2. Background information on CRAs and the industry

CRAs and “The Big 3”

CRAs evaluate bonds and other (fixed-income) financial securities in order to determine the ability of bond or security issuers to meet their financial obligation related to those instruments. Financial instruments rated by a CRA are often issued by major (global) corporations or by national governments. By means of an assessment on the instrument and its issuer, a CRA seeks to provide information to individual and institutional investors in order to increase insight in the credit risk involved, i.e. the likelihood that the issuer cannot repay the debt or recover losses in the event of a default. Next to a bond or security, CRAs also evaluate and grant credit ratings to a firm or government as a whole, thereby indicating their credit worthiness. The risk that is associated with the security or firm overall is reflected by a credit

rating. The industry is dominated by 3 large CRAs, (i.e. Standard & Poor’s (S&P), Moody’s and Fitch), which together cover around 96.5% of all outstanding ratings as at 31/12/2015 (2016 annual report of Nationally Recognized Statistical Rating Organization). Nevertheless, as can be seen in figure 1, the main dominant players of the industry are S&P (49.1%) and Moody’s (34.4%), with Fitch being third (13.0%).

Figure 1. Market share large CRAs

The leading CRAs use an alphabetic rating scale for their credit ratings such as AAA, BBB, BB (S&P), and Aaa, Baa, Ba (Moody’s). For an issuer overall, S&P for example describes their highest rating scale AAA as “extremely strong capacity to meet its financial commitment”. The lowest rating scale D is described as “in default on one or more of its financial obligations”.1_{Both S&P and Moody’s define}

their ratings as forward-looking opinions about the (relative) credit risk of an obligor. As such it is not an investment advice or a guarantee of credit quality. Within the rating scale, a distinction is made between investment grade ratings and speculative grade ratings. This distinction is rather important as it (highly) influences the investment decision of many investors. For S&P ratings from BBB- and higher have investment grade, while for Moody’s investment grade is applicable for ratings Baa and higher. Appendix 1 shows a complete overview of all rating scales from S&P and Moody’s.

(8)

Page | 7

Standard & Poor’s Global Ratings

S&P Global Ratings is part of the financial enterprise S&P Global Inc., which next to the credit rating division includes the S&P Dow Jones indices (like the well-known S&P 500), S&P Global Markets Intelligence and S&P Global Platts. Formerly known as McGraw Hill Financial Inc., in April 2016 the company changed its name to S&P Global Inc. The S&P rating business in 2016 had more than 1 million credit ratings outstanding on governments, corporates, financial entities and structured finance products. S&P Global Ratings operates in 28 countries around the world with approximately 1,500 credit analysts and $46.3 trillion in rated debt. Over the full year of 2016 S&P rating business earned revenues of $2,535 million, which reflects 44.8% of S&P Global’s total revenues of $5,661 million. 2

Moody’s Investors Services

Moody’s Investors Services (provider of credit ratings, research and risk analysis), together with Moody’s Analytics (provider of software, advisory services, research and financial risk management) belong to Moody’s Corporation. Formerly a division of Dun & Bradstreet, Moody’s spun off in 2000 to become a freestanding company. Moody’s Corporation operates in 36 countries worldwide, with around 10,700 employees and revenues of $3,604 million in 2016. Over 2016 Moody’s rating business was responsible for total revenues of $2,340 million, which reflects 64.9%. 3

Fitch Ratings

Fitch Ratings is part of the financial information services company Fitch Group. Next to Fitch Ratings the group consist of Fitch Solutions (provider of credit market data), BMI Research (independent provider of country risk and industry analysis) and Fitch Learning (training and professional development firm). The group operates in more than 30 countries and because it is not a public company, no financial information is provided on its website. 4

Brief history on CRAs

The first bond rating agency was established by John Moody in 1909. The historical development of the credit rating industry before that time is extensively described in the working paper by R. Sylla (2001). His work starts by explaining that already in the beginning of the seventeenth century a

2_{www.spglobal.com}_{(annual report 2016)} 3_{www.moodys.com}_{(annual report 2016)} 4_{www.fitchratings.com}

(9)

Page | 8 Dutch government bond market existed and that the Dutch East India Company were the first to issue common stock in order to raise capital. In the following centuries, financial systems developed more into the modern form as we know it with key components like elements of a banking system, a central bank, stable money and a securities market. During the nineteenth century more and more countries developed financially and the international bond market grew significantly. The United States differed from the European countries due to its large geographical landscape. With large scale projects, mainly in the railroad industry in order to establish an economy amongst states, the need to raise capital increased substantially. This led to the development of a domestic and international market for bonds in U.S. railroad corporations. As is also pointed out by White (2013), it is not by accident that John Moody, the founder of the first bond rating agency, was an American and that his ratings were first only covering bonded debt of U.S. railroad companies. However, the innovation of John Moody was actually a combination of developments that had happened before 1909. During the nineteenth century credit reporting agencies were created to provide information on suppliers and customers. A well-documented example is that of the Mercantile Agency, founded in 1841 by Lewis Tappan. Tappan had compiled an extensive number of records on the creditworthiness of his customers. He decided to specialize on the provision of commercial information and provide the information to subscribers. His company later became R.G. Dun and Company in 1959, which in 1933 merged with Bradstreet Company (that operated in a similar business). Going back to the railroad industry, as it was the first big business in America, a specialized publication was brought to the market called The American Railroad Journal. Henry Varnum Poor became the editor of this journal in 1849 and he published information like assets, liabilities and earnings of the railroad companies for the use of investors. In 1916 the Poor Company entered into the rating business as well and in 1941 it merged with Standard Statistics (another ratings and information company) to form Standard and Poor’s.

Further explained by White (2013) is that the rating industry significantly changed in the year 1936. In that year the federal bank regulators decided that regulated banks could only invest in bonds that were investment grade. The rationale behind this development is intuitive as it reflects prudential regulation in order to keep banks solvent. Nevertheless, this regulatory change had a major influence in the rating industry, because many investors were now mandated to use credit ratings for their bond investments, which prior to this change was entirely a voluntary matter. Another important change that White (2013) points out occurred in the early 1970s. Although Moody’s originally created a business model that was focused on investors, in which the investor purchased a credit rating (investor-paid model), in the early 1970s the market gradually switched to a business model in which the issuer pays for the rating (issuer-paid model), and which still dominates the market today.

(10)

Page | 9 Although exact reasons for this switch have not been thoroughly analyzed, he points out two important developments: (1) low cost photocopying conquered the market, which might have feared CRAs in that it would undermine their revenues, (2) due to expanded regulation, bond issuers really needed credit ratings in order to get their bonds in the portfolio of regulated institutions (banks and at that time also insurance companies). With the issuer-paid model potential conflicts of interest were created, which is to be further explained in the next chapter.

Regulation of CRAs

Many large institutional investors, like pension funds and mutual funds, but also banks and insurance companies, have to comply with governmental regulations that restrict their investment decision. As such, these institutions have to invest in at least investment grade instruments. Should it be the case that an issue afterwards is downgraded, adjustments to the portfolio need to be made. This means that CRAs have quite a lot of power as they strongly influence the pool of investments available to these large institutions.

Because of this powerful position and the significant consequences a rating change (downgrade) could have in financial markets, CRAs are heavily regulated by the United States Security and Exchange Commission (SEC). The SEC’s mission states that it seeks to “protect investors, maintain fair, orderly, and efficient markets, and facilitate capital formation”. In order to constitute a certain level of quality in credit ratings that banks and insurance companies were obliged to use in their investment decisions, only Nationally Recognized Statistical Rating Organizations (NRSROs) could be used. These NRSROs concerns CRA firms that received SEC certification, and the authority to rate federal and state regulated entities (Scalet and Kelly, 2012).

As a response to the high profile accounting scandals of Enron in 2001 and Worldcom in 2002, the Securities and Exchange Commission (SEC) tightened regulation around credit ratings, in order to increase the accuracy of ratings issued by CRAs and to reduce the conflict of interest following the issuer-pays model. As a result the Credit Rating Agency Reform Act was enforced in 2006. This act aimed to improve the quality of rating agencies, by means of fostering accountability, transparency and competition in the CRA industry, in order to protect investors and the public interest (Public Law 109-291, 2006).

Following the 2008 financial crisis, which had proven that the Credit Rating Reform Act had had little impact in preventing or limiting the severity of the crisis, the Dodd-Frank Act was introduces in 2010.

(11)

Page | 10 This act added regulation to that already existing in order to promote the financial stability in the United States by means of improving accountability and transparency of the financial system.

(12)

Page | 11

3. Literature review

Theoretical literature

Many theories have been proposed in the years after the high profile accounting scandals and the recent financial crisis, which try to explain why CRAs clearly incorrectly assessed the credit risk of companies or financial securities. One of the most discussed theories relates to the conflict of interest that CRAs may have, as a result of the issuer-pays model. As explained by Scalet and Kelly (2012) the conflict of interest can occur in at least two ways: (1) when the CRA is paid by the issuer of the security, the CRA might want to act in the best interest of the client rather than accurately rate the security; (2) in the case that the CRA earns significant fees for consulting services related to the structuring of securities that in turn will be rated by the CRA, the latter might want to maximize consulting revenues rather than to produce an accurate rating. Both these conflicts imply that CRAs are incentivized to inflate credit rating in such situations.

This is consistent with work from Bolton et al. (2012) who find that credit ratings were inflated during booming times when investors are more trusting. They consider three sources of CRA conflicts: (1) the conflict of understating risk to attract new business (2) the ability of issuers to purchase only the most favorable rating (3) the trusting nature of investors. Their most important finding is that a duopoly market actually contributes less to an efficient market than a monopoly market. This seems counterintuitive at first as one would expect that increased competition improves rating quality (as was intended by the Credit Rating Agency Reform Act), yet this result is explained by the fact that issuers have the opportunity to shop for the best rating.

Skreta and Veldkamp (2009) study the relationship between rating shopping and inflated credit rating. They show that a combination of complex assets and the possibility to shop for a credit rating can in fact produce rating inflation. The intuition is when issued ratings differ from each other, the more incentive an issuer has to only disclose the most favorable one. For simple assets this is not the case because CRAs then have nearly identical forecasts. Furthermore, they state that the link of rating shopping and complex assets can work both ways. When issuers want to shop for ratings they might want to further increase complexity of their assets in order to have a broader choice of ratings (as the ratings will more likely differ from each other).

Becker and Milbourn (2011) also examine the effect of increased competition on the credit ratings market. They study how the entry of Fitch to the market affects rating quality from the incumbent (S&P and Moody’s). The find evidence of lower rating quality when completion increases. They first

(13)

Page | 12 find that when completion increases, ratings from S&P and Moody’s rise (moving closer to the AAA segment). Secondly, credit ratings become less informative when competition increases, as less variation in bond yields can be explained by S&P and Moody’s. In addition, they find that the ability to predict default decreases when Fitch has an increased market share (although this test only includes data from S&P).

Contrary to the aforementioned outcomes that rating shopping would contribute to inflated credit ratings, Griffin et al. (2013) find that before the credit crisis, CDO’s rated by both S&P and Moody’s defaulted more often than CDO’s rated by just one of the two. This contradicts the theory of “rating shoppers” having less accurate ratings, as they would pick only the most favorable rating.

Bongearts et al. (2012) examine how the use of multiple CRAs will be affected by the widespread dependency on credit ratings. They find that for the same bond at the same point in time, credit ratings differ significantly, with Fitch being on average clearly more positive than S&P and Moody’s. This in line with Fitch playing a strategic role when S&P and Moody’s could withhold investment grade status, i.e. that Fitch is available to “push” a bond into investment grade when S&P and Moody’s are on opposite sides of this border (bonds are classified as follows: for issues with 2 ratings the lower rating is used while for issues with 3 ratings the middle rating is used). In addition, while CRAs in general provide useful information to the market, the writers do not find evidence that Fitch provides additional information incorporated in bond prices, relative to the information of S&P and Moody’s already included in prices. As such they find strong evidence in that Fitch ratings have a regulatory certification effect. Because the likelihood of having a Fitch rating is strongly linked to S&P and Moody’s being on the other side of investment grade and non-investment grade, Fitch ratings are sought to be a “tiebreaker” in these situations.

Cornaggia and Cornaggia (2013) compare in their study the timeliness and stability of credit ratings from a traditional issuer-paid CRA (Moody’s) and a subscriber-paid CRA (Rapid Ratings). Their results show that Moody’s has a strong preference for rating stability rather than timeliness. While Rapid Ratings is more likely to reverse a downgrade than Moody’s, the latter is significantly slower in identifying default risk. According to the writers, Moody’s bias towards rating stability exceeds the threshold what is needed to mitigate rating volatility, wherefore they conclude that Moody’s has relatively informative ratings as a result of the issuer-paid model. This result stems from the fact that Rapid ratings provides advanced notice of defaults and as such it provides significant loss avoidance compared to Moody’s.

(14)

Page | 13 He Qian and Strahan (2012) find that there is a robust relationship between issuer size of a Mortgage Backed Security (MBS) and the market price of these securities, conditional on credit ratings. Their results suggest that especially during boom periods investors accounted for the risk that credit ratings of large issuers were more inflated than those of small issuers. This is in line with Bolton et al (2012) who find that a CRA is more prone to inflate a rating when an issuer is more important, for instance when it is a repeat issuer or when it has larger issues.

The issuer-pays model is often criticized because research has shown that it contributed in different ways to inflated credit ratings, which in turn played a key role in the past financial crisis. In a response to the issuer-pays model which is still strongly dominating the market, policymakers have come up with alternative models in order to improve the CRA industry. A frequently discussed alternative by academics is the paid model. Xia (2014) finds that the entry of an investor-paid CRA to the market significantly improved the rating quality of S&P, in which the latter became more responsive to a firm’s credit risk and its rating change incorporate more information content. Bongearts (2014) tests next to the investor-paid model the potential of other alternative business models such as investor-produced model and mandatory co-investment by CRAs. He finds that the potential of these alternative models is limited and even if welfare improves, demand for such alternatives is limited when issuer-pays CRAs still operate in the market. This is explained in that investor-paid CRAs will face a cost-disadvantage as long as issuer-paid CRAs operate in the market as well, wherefore investor-paid CRAs will not even try to enter the market. Mandatory co-investment would increase rating accuracy as inaccurate ratings would lead to losses for the CRA when it has a stake in the issue. Nevertheless, because issuers prefer high over accurate ratings, they would prefer a CRA that does not have a stake in the product.

An efficient market contributes to the allocation of capital in that capital flows to the “good” investments that are expected to realize higher returns compared to the “poor” investments. The efficient markets theory according to Fama (1970) states that a market is efficient when prices always “fully reflect” the information available. The theory consist of three forms, (1) the weak form in which the available information reflects (just) historical prices, (2) the semi-strong form in which the available information reflects all the information that is publically available, (3) the strong form in which the available information reflects all information (also non-public) that is relevant to price formation. Under the assumption that CRAs base their credit ratings on information that is publically available, one would expect that the information is already incorporated in the price and as such, a change in a credit rating does not result in a price change. In line with efficient market theory, the

(15)

Page | 14 information content of a credit rating change has been widely studied by academics, in order to establish whether a credit rating change in fact does convey new information to the market.

Empirical literature

One of the first researches on the information content of a credit rating using stock data was done by Pinches and Singleton (1978). In their study they examine whether a credit rating change was anticipated for by the market before the rating event, based on monthly stock returns. They find strong support in that both an increase as well as decrease in a credit rating resulted in abnormal common stock return before the announcement of the rating change. Nevertheless, after the rating change normal returns were observed, although with slight tendency for reversal in the month the rating change occurred. From these findings they conclude that the information content of a bond rating change is very small and that the stock market appears to be highly efficient in processing the information coming from a rating change (both upgrades and downgrades).

In line with research of Pinches and Singleton (1978), Griffin and Sanvicente (1982) also used monthly stock returns in their study on the information content of a bond rating change, looking at the 11 months before and the month during a rating reclassification. They find that bond downgrades do convey new information to common stock holders and significant abnormal returns were observed for the month of the rating change and the 11 preceding months of the event. For upgrades however, price adjustments were not statistically significant in the month of the announcement, while in the 11 preceding months positive abnormal returns were observed for the upgraded firms.

The first to use daily stock return in their research instead of monthly stock return were Holthausen and Leftwich (1986). Because of the use of daily stock returns they could use a smaller timeframe contributing to a more powerful test. In their study they narrow down their focus on a time period of a 2-day window. Negative abnormal stock returns are observed for bonds that were downgraded by either S&P or Moody’s, while little evidence is found for abnormal returns in the case of a rating upgrade. Interestingly they report that negative abnormal returns are statistically significant when a downgrade occurred across rating classes (from AA to A), while no significant abnormal returns are detected for downgrades within rating classes (from AA to AA-).

Hand et al. (1992) examine daily excess returns for both bonds and stocks, associated with a rating change by S&P and Moody’s. Consistent with previous studies they find significant negative excess

(16)

Page | 15 returns for downgrades and weaker positive (not significant) average excess returns for upgrades measured on stocks prices, but the same outcome is found for bond prices. They also use a non-contaminated sample (non-non-contaminated in relation to other news around the event date), which does result in significant positive excess bond returns for upgrades, while this is not the case for stock returns.

Another study on the reaction of stock return following a bond rating change was done by Goh and Ederington (1993). While they find a significant negative reaction on stock prices following a downgrade, they claim that this reaction is not to be expected for all downgrades. A distinction is made between downgrade as a result of deterioration in the firm’s financial prospect and downgrades due to an increase in leverage. They argue that an increase in leverage in fact has positive implications towards stockholders, as action is taken to move wealth from bondholders to stockholders. They find support in their results that a downgrade based on deteriorated financial prospect shows a significant negative stock reaction, while this is not found in the case leverage is increased. Based on the aforementioned they conclude that rating changes should not be treated homogeneous.

Dichev and Piotroski (2001) carried out a long run study on stock return following bond rating changes. They examine stock return reactions over a period from 1970 to 1997 using all rating changes from Moody’s in that same period. In line with other studies they find no reliable abnormal returns for upgrades, while for downgrades substantial negative abnormal returns are observed. The underperformance of stock is most pronounced for the first month following a downgrade and lasts at least a year. In addition, this result is most pronounced for small firms and low-credit-quality firms. In order to explain these results, they document that downgrades are strong predictors of future deterioration in earnings.

According to Vassalou and Xing (2003) the outcome of significant negative abnormal stock return following a downgrade and the lack of significant stock return in the event of an upgrade depends on the method used to compute abnormal returns, specifically that large variations in default risk are not taken into account. They compute a default likelihood indicator and when returns are adjusted for default risk but also for book-to-market ratio and size, the negative abnormal returns as observed in other studies disappear and only remain in long horizons. When also considering occurrence of a subsequent downgrade following the initial downgrade (many firms face another downgrade within 3 years after the initial downgrade), abnormal returns in the long horizon disappear as well or become insignificant. The absence of significant abnormal returns for upgrades is according to the

(17)

Page | 16 writers explained by the minimal variance in default risk observed around upgrades. Based on these results they argue that credit ratings and downgrades have a disciplinary effect on the management of companies and provide a monitoring service for investors.

Boot et al. (2006) also document a significant negative effect on stock prices as a result of a downgrade in credit rating, while upgrades on credit ratings do not show a significant reaction. They claim that credit ratings could play a role as “focal points” in that if sizable investors (e.g. pension funds) base their investment decision on the credit rating, other investors will rationally follow. They find that the effect of a downgrade is even more informative after a credit watch procedure, which according to their theory allows for an implicit contract between the CRA and a firm, where the latter is supposed to take specific actions to prevent a possible downgrade. This finding is interesting as it contradicts the results of others that have studied the effect of a downgrade following an addition to the credit watch procedures, which showed that such rating changes did not lead to significant abnormal returns (Followill and Martell 1997).

(18)

Page | 17

4. Hypotheses, methodology and data

Hypotheses

Considering the powerful position of CRAs and following tightened regulations after the accounting scandals, it makes sense that investors were expecting a credit rating to accurately reflect the risk involved. Yet the last financial crisis has shown that credit ratings in fact were not correctly reflecting credit risk involved. In view of this, this paper is aimed at answering the following central question:

How did the relationship between a credit rating change and stock prices evolve following the 2008 financial crisis?

In order to answer the central question outlined above, three different time periods will be compared to show a possible change in the relationship of rating changes on stock data. A period before the financial crisis (2005-2007), a period during the financial crisis (2008-2010) and a period after the financial crisis (2011-2016). In line with the results of prior research it is expected that a significant negative abnormal returns to be observed as a result of a downgrade, while for upgrades positive abnormal returns (although not significant) are generally observed (Holthausen and Leftwich, 1986; Hand et al., 1992; Goh and Ederington, 1993; Dichev and Piotroski, 2001; Boot et al., 2006). In view of these findings the following hypothesis is proposed:

H1: There is a significant effect on stock prices following a change in credit rating in the period; - before the financial crisis, 2005-2007

- during the financial crisis, 2008-2010 - after the financial crisis, 2011-2016

As mentioned before, institutional investors like pensions funds, banks and insurance companies have to comply with governmental regulations. This restricts their investment decision and requires that investments are made in issues that have investment grade status. The distinction between investment grade and non-investment grade is therefore an important factor in the investment decision of institutional investors, and as such it could explain the abnormal returns when a rating moves into non-investment grade or vice versa. In view of the aforementioned, the following hypothesis is proposed:

H2: The effect of a credit rating change on stock prices is more pronounced for rating changes moving in or out investment-grade status.

(19)

Page | 18 Prior research has also shown that certain firm characteristics influence the effect of a rating change on stock prices and explain abnormal return performance. As explained above, financial institutions have to comply with regulations around their investment decision. When downgraded it becomes more expensive to borrow money, which is fundamental for their operations. Moreover, the public highly trusts financial institutions to operate and invest in a save manner. When trust in banks for example is broken, it counteracts the bank to continue its operation in the same way as people will be less likely to trust that bank with their money. As such one would expect that a rating change has a stronger effect for financial institutions than for non-financial institutions. Goh and Ederington (1993) included leverage in their study and find that an increase in leverage does not result in abnormal stock returns. Nevertheless, from a first intuition an increase in leverage means that the probability of a company defaulting on its debt increases as well. It is therefore interesting to assess whether leverage might explain abnormal return in the analysis. Firm size seems to be an important characteristic as well. Dichev and Piotroski (2001) and find that negative abnormal returns following a downgrade are more pronounced for small firms, while Vassalou and Xing (2003) find that negative abnormal returns disappear when corrected for firm size. Considering that a downgrade in the rating means that is will become more expensive to extract funding because the cost of capital increases, it makes sense that smaller firms take a harder hit when funding costs increase. Vassalou and Xing (2003) also included the book-to-market ratio in their study and when adjusting for this ratio they find that abnormal returns disappear. A book-to-market ratio indicates whether a security is undervalued or overvalued. When book-to market ratio is low (higher market value than book value), a security is overvalued. In case book-to market ratio is high (higher book value than market value), a security is undervalued. Companies with a high book-to-market ratio are also referred to as value stocks or growth stocks, as it is expected that the company’s market capitalization is to grow to its potential. In line with findings from Vassalou and Xing (2003) it is interesting to consider this firm characteristic as well. Lastly, a company’s profitability might be an explanatory factor for abnormal return. When a company has low profitability it has more difficulty in meeting its financial obligations and as such the probability of default increases. This firm characteristic is therefore considered. In view of the aforementioned, the following hypothesis is proposed:

H3: Abnormal return performance is explained by the following firm characteristics; - industry

- leverage ratio - firm size

- book to market ratio - profitability

(20)

Page | 19

Methodology

In this thesis standard event study methodology is used in which abnormal returns will be compared to “normal returns”. In order to find normal returns, often used models are the market-model or the market-adjusted model. As evidenced also by Brown and William (1985), in general the outcome of both models are similar. Considering that with the somewhat simpler market-adjusted model parameter estimation is not required, this model will be used to find abnormal returns.

The model is constructed as follows:

𝑅𝑅

𝑖𝑖𝑖𝑖

= 𝑅𝑅

𝑚𝑚𝑖𝑖

+ 𝜀𝜀

𝑖𝑖𝑖𝑖

(1)

𝑅𝑅

𝑖𝑖𝑖𝑖 = Actual stock return at time t

𝑚𝑚𝑚𝑚

= S&P 500

𝑅𝑅

𝑚𝑚𝑖𝑖

=

Actual market return at time t 𝑖𝑖 = company

𝜀𝜀

𝑖𝑖𝑖𝑖

=

The abnormal return 𝑚𝑚 = day

Because this thesis focuses on companies that are listed on the United States stock exchange, the benchmark that is used for the market is the S&P 500 index. The index has a broad scope and is often used as a benchmark for the U.S. market.

For the normal stock return and market return the log returns are calculated by means of the following formula:

𝑅𝑅

𝑖𝑖𝑖𝑖

= 𝑙𝑙𝑙𝑙

_𝑃𝑃𝑃𝑃_𝑡𝑡−1𝑡𝑡

(2)

𝑅𝑅

𝑖𝑖𝑖𝑖 = Actual stock return at time t

𝑃𝑃

𝑖𝑖

=

The price of the stock at time t

𝑃𝑃

𝑖𝑖−1

=

The price of the stock 1 day prior to time t

The stock price reflects the adjusted closing prices of the stocks that are included in the sample, because it corrects historical prices for events such as dividend payments or stock splits.

When all the above information is obtained and calculated, the abnormal return can be calculated as follows:

(21)

Page | 20 The intuition of measuring abnormal returns is that one can assess the impact of an event. By subtracting the normal return that is unconditional to the event from the actual return, the abnormal return that is conditional to the event remains. This way the abnormal return measures whether a change in stock price is associated with the event at time t.

As Kothari and Warner (2004) explain, an event study aims at establishing whether the return at time t of a cross-sectional distribution is abnormal. In most studies the focus is almost always on the mean of the abnormal return distribution. This makes sense if one wants to establish if the event, on average, is associated with a change in security price. The cross-sectional mean abnormal return (also called the average abnormal return) for a sample of N securities at time t is calculated as follows:

𝐴𝐴𝐴𝐴𝑅𝑅

𝑖𝑖𝑖𝑖

=

_{𝑁𝑁 � 𝐴𝐴𝑅𝑅}

1

𝑖𝑖𝑖𝑖 𝑁𝑁 𝑖𝑖=1

𝑚𝑚𝑚𝑚

= S&P 500

(4)

𝑖𝑖 = company 𝑚𝑚 = day

After an aggregation across firms, the abnormal return needs to be aggregated across time. Kothari and Warner (2004) further explain that it is interesting to find whether the mean abnormal return around the event date is equal to zero because (1) the event can be (partially) anticipated on which means that some of the abnormal return that is related to the event occur pre-event date, (2) abnormal returns post event date provide information on market efficiency (when there is no abnormal return after the event this meets the efficient market hypothesis). The cumulative abnormal return is defined as follows:

𝐶𝐶𝐴𝐴𝑅𝑅

𝑖𝑖

= � 𝐴𝐴𝑅𝑅

𝑖𝑖𝑖𝑖 1 𝑖𝑖=−1

(5)

In this thesis an event window of 3 days (-1/+1) is focused on because a short window will be more powerful in associating abnormal return performance to the event measured. When a longer event window is used, the probability that abnormal performance is not linked to the event it concerns increases. Nevertheless, considering that anticipation to the event can occur but moreover, the speed of information revealed in prices is an unknown factor, an event windows ranging up to 11 days (-5/+5) are measured and included as well.

(22)

Page | 21 In order to establish whether the abnormal return across firms and across event window on average systematically differs from zero, we lastly need to compute the cumulated average abnormal return:

𝐶𝐶𝐴𝐴𝐴𝐴𝑅𝑅 =

_{𝑁𝑁 � 𝐶𝐶𝐴𝐴𝑅𝑅}

1

𝑖𝑖 𝑁𝑁 𝑖𝑖=1

(6)

With the CAAR as performance measure one can statistically test the null hypothesis that abnormal return in the event window is equal to zero. The test statistic is calculated as:

𝑚𝑚

_{𝑠𝑠𝑖𝑖𝑠𝑠𝑖𝑖𝑖𝑖𝑠𝑠𝑖𝑖𝑖𝑖𝑠𝑠}

= √𝑁𝑁

𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶_𝜎𝜎

(7)

Where

𝜎𝜎

the standard deviation is of the cumulated abnormal returns on the cross-sectional distribution of the sample and defined as:

𝜎𝜎 = �

_{𝑁𝑁 − 1}

1

�

(𝐶𝐶𝐴𝐴𝑅𝑅

_𝑖𝑖

− 𝐶𝐶𝐴𝐴𝐴𝐴𝑅𝑅)

𝑁𝑁

𝑖𝑖=1

(8)

After the event study a multivariate regression is performed in order to establish which firm characteristics might explain the abnormal return performance. In line with the expectation from prior research that significant abnormal returns are only found for downgrades, the regression will only be performed on abnormal returns from downgrades as well. The regression will include the variables that have been mentioned in the previous section:

Leverage ratio is determined as total debt / total assets. Because an increase in leverage means a higher probability of default, it is expected that leverage will have a negative effect on abnormal returns and that the sign of the coefficient will be negative. Profitability ratio is calculated as net profit / total revenues. An increase in profitability is a positive indication because it means that a company is better able to meet its financial obligations, and therefore the sign of the coefficient is expected to be positive. Book-to-market ratio is determined as book value / market value. As explained earlier, a high book-to-market ratio indicates that a security is undervalued and as such it expected that this would negatively influence abnormal returns with a negative sign on its coefficient. In order to determine whether a firm is considered small or not, a dummy variable is

(23)

Page | 22 created with 1 for small firms and 0 for large firms. In line with the work of Vassalou and Xing (2003) small firms are defined as the first quartile of the sample based on market capitalization. Because prior research shows that abnormal returns are more pronounced for small firms, it is therefore expected that this will negatively influence abnormal returns with a negative coefficient sign. The last variable to be included in the regression is a dummy variable, indicating whether a firm is financial (1) or financial (0). Financial firms are more restricted in their investment decision then non-financial firms and it is therefore expected that the effect of a downgrade is stronger. In view of this the sign of the coefficient is expected to be negative.

Data

Because S&P and Moody’s are the leading CRAs in the rating business, this thesis is focused on rating changes of either S&P or Moody’s or both. As mentioned earlier, the scope of the thesis is the United States market and as such all companies that are listed on a stock exchange and which are rated by S&P and/or Moody’s are included in the sample. The sample periods reflect a period before the financial crisis (2005-2007), during the financial crisis (2008-2010) and after the financial crisis (2011-2016). Because the focus lies on the United States market, the S&P 500 index is used as a benchmark for calculating the market return. The data regarding all companies that are stock listed in the United States was obtained from the Orbis database. The Bloomberg database was used to obtain which of these companies were rated by at least S&P or Moody’s and for which a rating action was performed in the period 2005-2016. The stock price data and data regarding the variables of the regression were obtained from the Bloomberg database as well.

With regard to the credit ratings, only the long term issuer ratings from both agencies have been considered in the sample because these ratings indicate a CRAs opinion on the creditworthiness of an issuer as a whole (rather than the likelihood of default on a specific security).

Table 1 provides an overview of rating actions per period and split by the CRA performing the rating action. The total sample comprises of 504 downgrades performed and 459 upgrades. It can easily be observed that S&P performed more rating changes than Moody’s in all periods and for both upgrades and downgrades. It should be noted however that the ratings included in the sample only reflect long term issuer ratings, for which less data was available from Moody’s. Interestingly the results in table 1 show that most downgrades were done in the period of the financial crisis (2008-2010), while the least upgrades were performed in that same period. While this seems intuitive and in line with previous research that show that credit ratings appeared to be inflated before the

(24)

Page | 23 financial crisis, clearly the CRAs were more inclined to downgrade firms in the period 2008-2010. In addition, in the period after the financial crisis (2011-2016) the number of rating actions clearly reduced considering that the last period comprises of 6 years while the other two periods both reflect 3 years.

In table 2 and table 3 descriptive statistics are presented for the variables that are included in the multivariate regression. Table 2 shows the dummy variables and how these are distributed amongst the sample. Only the sample for downgrades is presented as the multivariate regression is only performed on downgrades as well. Small firms are defined as the first quartile of the sample based on market capitalization and as such represent 25% of the 504 downgrades. One can observe that the smallest firm still represents a company with a substantial size, $207 million dollar. This also appears to be a non-financial firm. The largest firm on the other hand has a total market capitalization of $346 billion. Table 3 shows there is much variation on the remaining variables as well. Especially profit margin is widely dispersed amongst the sample.

This table shows statistics of the dummy variables by means of market capitalization expressed in millions of dollars. Table 1. Descriptive statistics rating changes

Downgrades Upgrades S&P 102 133 Moody's 32 36 Total 134 169 S&P 145 75 Moody's 64 7 Total 209 82 S&P 113 178 Moody's 48 30 Total 161 208 2005-2016 Total 504 459 Period 2005-2007 Period 2008-2010 Period 2011-2016

Table 2. Descriptive statistics firm specific regression variable

N Minimum Maximum Mean

Financial firms 144 225 262320 37838

Non-financial firms 360 207 346648 22364

Small firms 126 207 3355 1952

(25)

Page | 24

5. Empirical results

Preliminary results of the event study

Table 4 presents the descriptive statistics for upgrades that were performed in the period before the financial crisis, during the financial crises and the period after. The cumulative abnormal return for an event window of 3 days and 11 days is presented. The mean, which in fact is equal to the cumulative average abnormal return (CAAR), is positive for all 3 periods. From the table it can be concluded that on average the most abnormal return is observed during the financial crisis period, in the 3 day event window at 0.36% as well as an in the 11 day event window at 0.65%. It can also be noted that the variation in stock return is the highest in the period during the crisis, which can be derived from the standard deviation of 3.34% and 6.63% for the 3 day event window and 11 day event window respectively.

This table shows the cumulative abnormal return (CAR) calculated per company in the sample for upgrades, for an event period of 3 days and 11 days. The three main periods are presented separately to show different characteristics.

In table 5 the cumulative abnormal return for downgrades is shown for all three periods. The mean is negative for the period before the crisis and during the crisis, for both event windows that are presented. However, for the period after the financial crisis the average abnormal return performance is positive. This is rather surprising because negative abnormal return is expected after a downgrade in credit ratings, in line with prior research from among others Holthausen and Leftwich (1986), Hand et al. (1992), Goh and Ederington (1993) and Dichev and Piotroski (2001). Corresponding to what was observed for upgrades, the highest abnormal return performance occurred in the period during the crisis, with an average of -4.65% for the 3 day window and an average of -4.90% in the 11 day window. Also the most variation in stock return is observed in this

Table 3. Descriptive statistics other regression variables

N Minimum Maximum Mean

Debt-to-asset ratio 504 0,000 0,964 0,291

Book-to-market ratio 504 -0,259 7,949 0,777

Profit margin 504 -27,247 1,013 -0,043

Table 4. Descriptive statistics upgrades

N Minimum Maximum Mean Std. Deviation Variance

CAR [-1,1] 2005-2007 169 -7,94% 10,74% 0,23% 2,26% 0,0005 CAR [-1,1] 2008-2010 82 -11,55% 10,23% 0,36% 3,34% 0,0011 CAR [-1,1] 2011-2016 208 -19,68% 13,34% 0,32% 3,09% 0,0010 CAR [-5,5] 2005-2007 169 -9,02% 13,83% 0,34% 3,54% 0,0013 CAR [-5,5] 2008-2010 82 -16,71% 25,79% 0,65% 6,63% 0,0044 CAR [-5,5] 2011-2016 208 -18,86% 17,71% 0,35% 4,52% 0,0020

(26)

Page | 25 period with a standard deviation of 18.74% and 22.94% in the 3 day and 11 day window respectively. It is interestingly to observe that the variation in downgrades is much higher than what is observed for upgrades. One can easily see that the minimum of abnormal return for downgrades during the financial crisis is much more negative than for upgrades, -11.55% compared to -151.45% in the 3 day event window. When zooming in on the data it is found that the -151.45% abnormal return relates to a downgrade on September 15, 2008 (performed both by S&P and Moody’s) for American International Group. One day later on September 16, 2008 the United States Government bailed out the company for an amount of $85 million. This way preventing that the company, at that time one of the world’s biggest insurers, would go bankrupt. Nevertheless, both S&P and Moody’s downgraded their ratings from high grade to upper medium grade ratings (from AA- to A- by S&P and from Aa3 to A2 by Moody’s). This means that 1 day before the bailout of the United States Government, American International Group was still rated with investment grade ratings.

This table shows the cumulative abnormal return (CAR) calculated per company in the sample for downgrades, for an event period of 3 days and 11 days. The three main periods are presented separately to show different characteristics.

Results of event study for upgrades

In table 6 the outcome of the cross-sectional t-test for upgrades is presented for a range of event windows. The average abnormal return is positive for all event windows and for all 3 periods. Finding positive abnormal return is in line with expectations based on results of prior research. Looking at the event window of focus, the 3 day event window, the positive average abnormal return performance is not significant in the 3 different sample periods. This is also in line with outcomes of prior research from Griffin and Sanvicente (1982), Holthausen and Leftwich (1986), Hand at al. (1992), Goh and Ederington (1993), Dichev and Piotroski (2001), Boot et al. (2006), that all did not find significant abnormal stock returns for upgrades. Contrary to these findings, there are however event windows that do show significant abnormal return in the table below. In the period 2005-2007 I find CAAR at 0.45% that is just significant, for the 6 days event window starting 5 days prior to the event date. This could indicate that investors anticipated the upgrade prior to the date of upgrade. In that same period significant abnormal return is observed for a 2 day event window starting the day

Table 5. Descriptive statistics downgrades

N Minimum Maximum Mean Std. Deviation Variance

CAR [-1,1] 2005-2007 134 -16,86% 10,47% -0,03% 3,51% 0,0012 CAR [-1,1] 2008-2010 209 -151,45% 23,04% -4,65% 18,74% 0,0351 CAR [-1,1] 2011-2016 161 -12,74% 8,90% 0,26% 3,37% 0,0011 CAR [-5,5] 2005-2007 134 -29,47% 11,09% -0,18% 5,42% 0,0029 CAR [-5,5] 2008-2010 209 -152,58% 40,08% -4,90% 22,94% 0,0526 CAR [-5,5] 2011-2016 161 -17,41% 17,03% 0,75% 5,43% 0,0029

(27)

Page | 26 of the event and the next day, indicating that investors react immediately after the event. Yet similar results for these event windows are not observed in the other periods. Nevertheless, in the period 2008-2010 significant abnormal return is observed for the 6 day event window up to 5 days after the event, which is in line with the previous finding that investors seem to respond after the upgrade rather than anticipating the upgrade.

Results of event study for downgrades

In table 7 the results of the cross-sectional t-test for downgrades is shown. Compared to the outcome for upgrades, the results in the different event windows is much more aligned per period. The table shows that in the period before the financial crisis there is no significant abnormal return for all the event windows, with even positive abnormal return for the window [-5,0] and [-1,0]. However, during the financial crisis all event windows show significant negative abnormal returns, which again is in line with research from Griffin and Sanvicente (1982), Holthausen and Leftwich (1986), Hand at al (1992), Goh and Ederington (1993), Dichev and Piotroski (2001), Boot et al. (2006). One could conclude that during the financial crisis, investors perceived a downgrade as new information to the market and reacted to it accordingly. This finding is rather interesting when taking into account that before the financial crisis CRAs already became under scrutiny following the accounting scandals of Worldcom and Enron, while CRAs played a key role in the financial crisis itself because many securities appeared to be inflated (Scalet and Kelly, 2012; Bolton et al., 2006; Xia, 2014; Bongearts, 2014). One explanation is that investors realized that ratings were inflated and as such strongly responded to downgrades knowing that the former rating was likely not correct. Another explanation lies with the larger investors like pension funds and other financial institutional investors like banks and insurance companies, which by regulation often need to sell their stock which become below investment grade. Another interesting finding is that in the period after the financial crisis only positive abnormal returns are found for all event windows, which is contrary to the expectation following prior research findings. For the event window [-5,5] and [0,5] the positive

Table 6. Outcome t-test upgrades at 95% significane level

[-5,5] [-5,0] [-1,0] [-1,1] [0,1] [0,5] 2005-2007 CAAR 0,34% 0,45% 0,19% 0,23% 0,36% 0,19% p-value 0,220 0,051 0,202 0,180 0,017 0,406 2008-2010 CAAR 0,65% -0,37% 0,18% 0,36% 0,47% 1,31% p-value 0,378 0,484 0,567 0,337 0,106 0,014 2011-2016 CAAR 0,35% 0,28% 0,12% 0,32% 0,20% 0,07% p-value 0,271 0,228 0,448 0,139 0,279 0,771

(28)

Page | 27 abnormal return is even significant. A possible explanation could be found in line with the findings from Goh and Ederington (1993), in which they conclude that an increase in leverage actually has positive implications for stockholders, because wealth is moved from bondholders to stockholders. Conditional on an increase in leverage, which is a reasonable scenario for companies in a period after crisis, this could explain why a downgrade is clearly perceived positively by investors in 2011-2016.

Results on the effect of moving in or out investment grade status

As was explained in the previous section, many institutional investors have to comply with governmental regulations and as such their investment decision is often restricted to investment grade issues. One would therefore expect that a downgrade for instance has more effect on the price of stock when this would mean that an investment portfolio needs to be adjusted. Unfortunately the sample does not contain downgrades that moved from investment grade into non-investment grade, wherefore it was not possible to compare abnormal return performance to that of downgrades not crossing the investment grade border. Out of the 459 upgrades in total, 87 upgrades moved into investment grade status which are compared to 372 upgrades that did not, the results are shown in table 6. Because stock returns are not normally distributed a Mann-Whitney U test is used to compare whether cumulate average returns significantly differ for upgrades that move into investment grade and upgrades that do not. Instead of comparing the mean of both groups, mean ranks are compared which are based on the median of both groups. From the table below it can be observed that for all event windows p-value > 0,05 and as such abnormal return performance appears not to be different for upgrades that push a credit rating into investment grade. This finding is not very surprising because an upgrade of a credit rating does not have similar consequences as for a downgraded rating. However, it does mean that the investment pool enlarges, because institutional investors then may invest in those instruments as well. One explanation as to why this is not observed in the results, is that large financial institutions likely invest in instruments that do not have credit ratings close to the investment grade border, because it would mean that the likelihood

Table 7. Outcome t-test downgrades at 95% significane level

[-5,5] [-5,0] [-1,0] [-1,1] [0,1] [0,5] 2005-2007 CAAR -0,36% 0,11% 0,04% -0,11% -0,22% -0,55% p-value 0,447 0,761 0,877 0,726 0,424 0,147 2008-2010 CAAR -4,84% -4,21% -4,04% -4,62% -3,20% -3,25% p-value 0,002 0,005 0,000 0,000 0,001 0,002 2011-2016 CAAR 0,75% 0,22% 0,06% 0,26% 0,21% 0,54% p-value 0,080 0,489 0,787 0,327 0,331 0,099

(29)

Page | 28 of having to make adjustments to the portfolio is much larger than when investing in upper medium or high grade ratings.

Results on the multivariate regression

In table 9 the outcome of the multivariate regression is presented including the variables that have been mentioned in the previous section. As can be observed, almost all of the variables in the model appear to be not significant with p-value > 0,05. Although the expectation for book-to-market ratio was that it would have a negative sign, for half of the event windows this variable appears to have a positive effect, albeit at a non-significant level. This could be explained by the fact that the sample comprises of relatively large companies (market cap > $207 million). Companies with low book-to-market ratio are considered growth firms and are in general less mature than the companies included in the sample. Profit margin on the other hand behaves in line with what was expected and except for the event window [-5,0] the coefficient in other event windows is positive. Only for the event window [0,5] the coefficient is significant with p-value 0,002. One would expect that a company with lower profitability is more likely to be downgraded. That it appears to be not a significant contributor to the abnormal return performance might be explained in the way CRAs decide to perform a rating action. As is also explained by Cornaggia and Cornaggia (2013), issuer-aid CRAs prefer rating stability over rating timeliness. This implies that when a company is deteriorating in means of profitability, the CRA not immediately downgrades. This could be an explanation as to why the variables is not significant in most periods, while significant in the event window [0,5]. Unlike the intuition that a higher debt ratio would negatively influence stock return, this is not supported by the outcome of the regression as all event windows show a positive coefficient. Although not significant in any of the event windows presented, this finding could support the theory of Goh and Ederington (2003) who argue that an increase in leverage has positive implications for stock holders (because wealth is moved from bond holders to stock holders). In addition, it does make sense that a higher leverage ratio not directly implies a higher probability of default. Many companies for instance increase their level of debt as it contributes to a lower cost of capital (external capital is cheaper than own capital). Furthermore, in the period after the financial crisis the economy started to improve again and this corresponds with a timing in which companies invest

Table 8. Outcome Mann-Whitney U test at 95% significance level showing mean ranks upgrades

[-5,5] [-5,0] [-1,0] [-1,1] [0,1] [0,5]

2005-2016

Upgrade into IG 233,85 242,72 225,55 235,57 232,34 220,91

Upgrade not into IG 229,10 227,03 231,04 228,70 229,45 232,13

(30)

Page | 29 more in order to grow their business. It makes sense that a higher debt ratio as purpose for investment in the company’s business has a positive effect rather than a negative effect. For the variable financial firms (dummy variable indicating whether a firm in the sample is a financial institution or not), all event windows show a significant coefficient. In line with the expectation that negative abnormal returns would be higher for financial firms, the sign of the coefficient is indeed negative. Because all the event windows indicate a significant coefficient, most event windows with p-value < 0,001, it can be concluded that a company’s industry (financial / non-financial) truly explains part of the abnormal returns observed. One could conclude that this finding is explained by the regulatory requirements financial institutions have to comply with. Yet the regulation in many cases prevents these institution to invest in instruments below investment grade, which is not the case in this sample. Another explanation could be that financial institutions highly depend on the trust of the public and as such do not want to have investments in the portfolio that have a deteriorated prospect. This could lead to reputational damage when and as such it is much saver to invest only in the higher grade rating segment. For the dummy variable small firms/large firms I find no support in that abnormal returns are explained and all the event windows indicate it is not significant either. Following findings of previous studies that abnormal returns are more pronounced for small firms than for larger firms, the intuition was a negative coefficient for this variable. Nevertheless, the majority of event windows actually indicates a positive effect for small sized companies. This finding might be the result of how the sample is made up. Especially because the smallest company has a market capitalization of $207 million, one might argue whether this truly represents a small cap company.

The model overall explains the variation in negative abnormal returns ranging from 2.8% to 5.5%, which is indicated by the R². This means that actually a very small part of the variation is explained by the variables included in the model and as such, the majority of variation in abnormal returns is more explained by variables that are not included in the model.

(31)

Page | 30 In extension to the regression analysis in which all variables were included in the model, I also performed a regression in which the variables are separately included. The result of this analysis is shown in table 10. As the focus of this paper is on the 3 day event window, the CAR for this period is used. The main reason for separately measuring the variables on the independent variable (abnormal returns) is because the dummy variable industry appears to be the only variable that has a significant effect on abnormal return performance. This could mean that other variables might also significantly explain some of the variation if the dummy variable industry is not included. Nevertheless, as can be observed from the results in table 10, the outcome of the second regression is similar to the first regression. When the variables are separately measured to explain the abnormal return, the dummy variable industry still is the only variable that is significant, with p-value < 0,001. The other variables behave similar as in the first regression and all do not have a significant impact on the abnormal return performance. Interestingly the variable book-to-market ratio has a negative influence on abnormal returns when included separately in the regression, while it had a positive sign when it was included with the other variables in the regression. The negative sign is in line with the intuition that a higher book-to-market ratio indicates undervalued stocks and thus would negatively influence abnormal returns. That the variables other than industry are not significant when separately included in the regression is probably explained by the fact that only 4.7% of the variation in abnormal returns

Table 9. Outcome multivariate regression on CAR from downgrades

[-5,5] [-5,0] [0,5] [-1,1] [-1,0] [0,1] Variables Book-to-market ratio 0,002 0,000 -0,001 0,008 -0,001 0,005 0,881 0,982 0,863 0,352 0,942 0,405 Profit margin 0,004 -0,006 0,013 0,004 0,003 0,004 0,490 0,291 0,002 0,442 0,534 0,327 Debt-to-asset ratio 0,055 0,035 0,056 0,034 0,035 0,035 0,212 0,398 0,058 0,338 0,262 0,205 Financial firms -0,052 -0,056 -0,027 -0,062 -0,048 -0,046 0,003 0,000 0,017 0,000 0,000 0,000 Small firms -0,008 0,003 -0,011 0,011 0,009 0,002 0,633 0,830 0,323 0,402 0,423 0,826 Constant -0,020 -0,012 -0,018 -0,019 -0,015 -0,015 0,239 0,446 0,108 0,151 0,211 0,156 N 504 504 504 504 504 504 R² 0,028 0,034 0,055 0,047 0,047 0,046

(32)

Page | 31 is explained in the 3 day event window when including all variables (see table 9), while the variable industry alone explains 4.2% of the variation in abnormal returns (see table 10).

Table 10. Outcome multivariate regression on CAR [-1,1] from downgrades

Variables Book-to-market ratio -0,010 0,148 Profit margin 0,004 0,388 Debt-to-asset ratio 0,040 0,268 Financial firms -0,057 0,000 Small firms 0,009 0,507 Constant -0,021 -0,019 -0,030 -0,003 -0,021 0,001 0,001 0,011 0,683 0,001 N 504 504 504 504 504 R² 0,001 0,001 0,002 0,042 0,001

Credit rating changes and stock prices