The Informational Content Of Ratings Changes Finance Essay - Free Essay Example

This paper focuses on the informational value of all rating change-announcements made by Fitch, Moodys and Standard Poors in the period between 2002 and 2011 i.e. around the Financial Crisis for Benelux-companies listed on respectively the BEL20-, AEX- and LuxX-index. Both short and long term credit ratings are taken into account. Results show that downgrades are associated with a significant cumulative average abnormal stock price return (CAAR) of -3.72%. Controlling for the anticipation of rating changes and the disclosure of other pricing relevant news around the announcement yields that unanticipated downgrades and downgrades surrounded by concurrent news are accompanied by greater abnormal stock price performances of respectively -7.44% and -7.17%. Other evidence shows that only small capitalized and/or high volatility firms are associated with significant abnormal stock price return of respectively -4.78% and -5.81% around the announcement of a downgrade. In addition, this study emphasizes the impact of prior rating levels and finds that downgrades starting from a level below the investment grade-barrier are associated with a significant CAAR up to -16.48%. Throughout the whole research, upgrades tend to yield only minor and insignificant results. Table of Contents Introduction The three credit rating agencies were key enablers of the financial meltdown. The mortgage-related securities at the heart of the crisis could not have been marketed and sold without their seal of approval. Investors relied on them, often blindly. In some cases, they were obligated to use them, or regulatory capital standards were hinged on them. This crisis could not have happened without the rating agencies. Their ratings helped the market soar and their downgrades through 2007 and 2008 wreaked havoc across markets and firms. (January 2011, The Financial Crisis Inquiry Commission Final Report) Credit Rating Agencies (CRAs) are widely seen as one of the major causers of the recent global Financial Crisis. They are criticized on their core activity; rating the creditworthiness of countries, companies and financial products (e.g. CDOs, RMBS, etc.). By awarding too high ratings CRAs contributed for a large part to the rapidly growing market on subprime mortgages. Without these hi gh rating it is very unlikely that there would have been such a large demand for especially these, party bad, financial products and hence the financial breakdown perhaps could have been avoided. CRAs are originally set in place to reduce the information asymmetry between borrowers and lenders. By rating the creditworthiness of different entities and products, they provide an important piece of information to outside investors. In their presence, information about companies gets disseminated much quicker and investors are spared from having to put very costly and duplicative effort into gathering the information. Thus, CRAs seem to play an essential role in todays financial market making the credit rating industry to grow quite rapidly of the last few decades. This market on credit ratings is dominated by three agencies, namely; Fitch, Moodys and Standard Poors. This study focuses on stock price effects around the announcement of rating changes made by the three CRAs in the peri od surrounding the Financial Crisis  [1]  . The main goal of this research is to get an insight into the informational value  [2]  of these rating revisions and check whether the evidence found in previous literature also applies for the Benelux stock-market. Therefore the main question to be answered in this research is as follows: What is the informational value of rating changes and how does the stock market react to the announcement of such changes? In other words, the main question is whether investors view rating changes as credible pricing relevant signals. If so, they will react accordingly by which stock prices will be affected and hence rating changes convey informational value. In order to be able to formulate a proper answer to this question, the following sub-questions are constructed: What is the difference in stock price effects between the announcement of up- and downgrades and how did the recent global Financial Crisis impact these effects? Does a stand-alone rating change significantly impact the price of a companies stock? Do investors use rating changes as substitute pieces of information if their own supply falls short? The setup and construction of the hypotheses to be tested according to the questions stated above is discussed in Chapter 3. Subsequently the data and methodology used in this research are discussed in Chapter 4. Empirical results from the tested hypotheses will be presented in Chapter 5 from which conclusions and possible grounds for further research are provided in Chapter 6. The contribution of this paper to existing literature can be particularly found in the countries to be studied. As the methodology used and the hypotheses tested largely correspond to earlier studies on the informational value of rating changes, the only difference lies in the fact that this study uses data from (only) the Benelux while others investigate the US, the UK, Spain, Italy, Oceania or the whole of Europe. To my best knowledge, no study fully concentrates on these or other relatively small countries. Literature Review This chapter provides an overview of the evidence presented by earlier studies on the informational value of rating change-announcements. The first three sections discuss evidence on rating changes in general whereas section 2.4 to 2.8 discuss specific types of rating changes (e.g. rating changes for firms which are small or large capitalized, rating changes starting from a level below or above the investment grade-barrier, short and long term rating changes, etc.) in order to get a complete picture on the implications of a rating change-announcement. An overview of all findings in this chapter together with the studies providing the evidence is presented in Appendix (I). Evidence found on the reaction of stock prices to stronger signals  [3]  and empirical results on this are provided in Appendix (II) as the subject is considered to be of less importance in answering the main question of this research. No effects around rating changes As Calderoni et. al. (2009) states: The Information Asymmetry and Signaling Hypothesis (IASH) predicts that credit rating changes are private information-based signals concerning issuers prospects. As a consequence, downgrades should be bad news producing a price reduction, while upgrades should be good news resulting in a price increase. This hypothesis states that CRAs seem to possess private information about a firm and their rating actions should therefore provide the market with new information. If these CRAs actually are in the position of having privileged access to private information, the market should respond to rating changes as these convey information the market does not possess. First of all the market seems to be highly efficient in the sense that it very rapidly reflects new information into stock prices (Fama et. al., 1969). That is, when there is new information available concerning a certain stock, it is already reflected in its price before CRAs modify their cre dit ratings, if needed. Especially the type of information CRAs provide by their rating activities seems to be processed efficiently in the stock market. Hence, rating change-announcements cannot be used to efficiently time the market and provide investors with protection against potential losses (Linciano, 2004). Several other studies agree on this and reject that CRAs provide the market with new information. Examples of these studies are Brookfield and Ormrod (2000) and Kuhner (2001). These studies state that CRAs only provide information which the market already seems to possess and hence this information is already reflected in stock and bond prices before the rating is changed. In agreement with the previous, Linciano (2004) states the following: Overall, rating agencies do not seem to act on the basis of private information. This evidence, although corroborating the hypothesis that rating agencies act in line with the financial market regulation prohibiting selective disclosur e of significant corporate events, supports the argument that the information content of ratings is modest (p.13). In order to confirm the above, extensive research has been done to check whether rating changes affect stock and bond prices. Weinstein (1977) using monthly bond returns, Wakeman (1978) using monthly stock and weekly bond returns and Pinches and Singleton (1978) using monthly stock returns all find no significant reaction of bond or stock prices around the announcement of a rating change. As Weinstein (1977) concludes: There is some evidence of a price change during the period from a half to one-and-a-half years before the rating change. This price change is the result of information which eventually leads to the rating change, rather than the rating change itself (p. 345). Other evidence even suggests that stock and bond prices behave in opposite directions compared to the scope of the revision. They find statistically significant but weak evidence of a return rever sal around rating change-announcements [Pinches and Singleton (1978) and Glascock et. al. (1987)]. A possible explanation for these results is provided by Goh and Ederington (1993) and Kliger and Sarig (2000). Both studies argue that if CRAs foresee that wealth is being transferred from bond- to stockholders due to an increase in leverage, equity (bond) prices rise (fall) around downgrades. By this, bond and equity prices move in opposite directions of each other and therefore the total value of the firm remains unaffected. Asymmetry between up- and downgrades In contrast to the previous section and according to the IASH, many studies claim that CRAs actually do provide the market with new valuable information about a firms future prospects. For example, Holthausen and Leftwich (1986) and Schweitzer et. al. (1992) both conclude that changes in credit ratings are credible information signals and should therefore lead to significant abnormal stock price movements around the announcement. Other, non-US, studies [i.e. Barron et. al. (1997), Elayan et. al. (2003) and Romero and Robles-Fernandez (2007)] find somewhat similar results in respectively the UK, New-Zealand and Spanish stock market. A possible reason for the above results is presented by Holthausen and Leftwich (1986). They argue that CRAs possess inside information about a firm, the rating process usually includes discussions with management, visits to company premises, and forecasts of income statement and balance sheet data provided by the management (p. 61). Other studies [i.e. Griffin and Sanvicente (1982) and Ederington et. al. (1987)] claim that CRAs are used as an instrument to disclose sensitive information to the market. By disclosing this sensible information through rating revisions, rather than through public disclosure, they prevent competitors from getting an informational advantage. Moreover, Ho and Michaely (1988) argue that rating changes merely based on public information could also be informative because of the costs associated with the collection of this information. These costs sometimes exceed the benefit and therefore market prices will probably not reflect all publicly available information. CRAs, which may get access to this information much more efficient, can therefore provide new valuable information to the market through their rating activities. To check whether the informational content of rating change-announcements is already reflected in bond and stock prices, a wide scope of research has been done starting a long time ago. Ex amples of these studies are Griffin and Sanvicente (1982) using stock returns and Ingram et. al. (1983) using bond returns. These studies provide significant evidence of abnormal stock price performance surrounding the rating change. Another study by Katz (1974) even claims that this abnormal performance stays on for a long period  [4]  after the announcement. The above studies, among others, come up with different explanations for the conflicting findings presented in section 2.1 and 2.2. As can be concluded from these studies, the main reason is related to the methodology that is used  [5]  . Most studies investigating the informational content of rating revisions distinguish between up- and downgrades when talking about stock price effects. As predicted by the IASH, both up- and downgrades should be surrounded by significant stock price reactions consistent with the direction of the rating change. Many studies [e.g. Dichev and Piotroski (2001), Vassalou and Xing (2005) and Halek and Eckles (2010)] reject this hypothesis and argue that only downgrades are surrounded by abnormal returns while upgrades only yield insignificant results. The asymmetry seems not only to persist in the US, but also in the UK- (Barron et. al., 1997) and Australian-market (Matolcsy and Lianto, 1995). Recent studies have tried to explain why there seems to be a difference in results between up- and downgrades. Jorion and Zhang (2006) perhaps explain this best by arguing that the asymmetry is caused by different rating levels prior to the announcement. More on this influence of prior rating-levels can be found in section 2.7. Another possible explanation is presented by Dichev and Piotroski (2001). They argue that good and bad news differ from each other as they have different information implications. Their evidence suggests that downgrades suffer from an underreaction; after the initial negative reaction to the bad news, prices keep decreasing. This evidence is consistent with findings presented by Matolcsy and Lianto (1995). They argue that these findings could be consistent with the propositions that good news travels fast compared to bad news, or that equity holders are more concerned with a downgrade than upgrades (p. 901). Another study claims that issuers tend to shift their information releases to only good news. Hence, bad news is rather scarce and downgrades should therefore be a more credible signal to other market participants (Ederington and Goh, 1998). In contrast to the mentioned asymmetry between up- and downgrades, several studies claim that upgrades actually do affect market prices. Past evidence presented by Schweitzer et. al. (1992) already comes up some marginally significant abnormal returns around upgrades and more recent literature by Kliger and Sarig (2000) and Romero and Robles-Fernandez (2007) state that not only downgrades seem to convey new information to the market. Besides actual up- and downgrades, CRAs can add rati ngs to the CreditWatch-list  [6]  . Micu et. al. (2006) investigate whether these additions to the CreditWatch-list also impact share prices and come up with evidence suggesting that all types of rating change-announcements, thus also reviews, significantly impact market prices. Recent studies agree on this but argue that not actual downgrades but only reviews for downgrade cause stock prices to change significantly [Linciano (2004) and Norden and Weber (2004)]. The latter study claims that within a combined analysis  [7]  actual downgrades experience no significant abnormal returns, whereas reviews for downgrade still do. They also argue that these reviews are anticipated much earlier  [8]  , whereas for actual downgrades most reactions seem to appear close before the event. More on this anticipation can be found in section 2.4. Elayan et. al. (2003) agrees on the above evidence but concludes that not only reviews for downgrade but also reviews for upgrade significantly impact market prices. Their evidence is consistent with an earlier study by Barron et. al. (1997) who investigates the UK-stock market and states that also placements on the CreditWatch-list for positive reasons are surrounded by significant positive abnormal stock price performances. Decline of trust after the Financial Crisis As discussed in the previous section, CRAs are important players in todays markets. They tend to have access to private information which they gained through their rating-activities. Therefore, theoretically, they can be seen as institutions helping to get rid of information asymmetries between security issuers and other stock market-participants (Kuhner, 2001). Other studies (e.g. Wakeman, 1981) argue that CRAs do not merely act as information intermediaries, but also play the role of reputable auditor  [9]  . When CRAs are not able to fulfill this role properly anymore, for whatever reason, confidence in their work will decline and it will obviously lead to weaker stock price effects around the announcement of a rating change. Why CRAs may be taken less seriously today is best explained by Calderoni et. al. (2009), who states the following: The recent financial turmoil has highlighted a severe decline in the confidence that investors are giving to the activity performed by Cr edit Rating Agencies. The harsh debate followed to the recent bankruptcy of Lehman Brothers that remained investment-graded until the day it asked protection under Chapter 11 has cast serious doubts about the role that these Agencies can perform in modern financial markets (p. 2). Pre-announcement stock returns As can be concluded from section 2.1, market prices do not seem to react to rating changes. This should either imply investors do not care about rating revisions or they already anticipated the rating change. This early anticipation could be a result of rating changes from which the informational content  [10]  is already known to the market upfront and hence is already reflected in stock price returns before the actual rating change is announced. Several studies find no significant stock price reaction prior to rating change-announcements [e.g. Linciano (2004) and Romero and Robles-Fernandez (2007)]. The study by Linciano (2004) argues that the absence of pre-announcement abnormal returns, even for the contaminated subsample, however, might be an indirect evidence of a timely action of the rating agencies when they move on the basis of a news which is already in the public domain (p. 14). This absence of anticipation by investors leads to stronger effects according to Gropp an d Richards (2001). Hand et. al. (1992) adds to this statement by arguing that this reaction is only present for unanticipated downgrades, while for upgrades no significant excess returns are observed. In contrast to the above, other studies claim that there actually is evidence of abnormal performance in the period preceding the rating change [Steiner and Heinke (2001) and Hull et. al. (2004)]. Other studies test this statement and present evidence of significant positive (negative) abnormal returns around upgrades (downgrades) preceded by abnormal returns in line with the direction  [11]  of the rating change [Wansley and Clauretie (1985) and Goh and Ederington (1999)]. Norden and Weber (2004) argue that this anticipation starts approximately 90-60 days before the event. In addition to this, Goh and Ederington (1999) conclude that the market reaction is also stronger if the firm has experienced negative pre-downgrade abnormal returns. (p. 101). They do not find this evidence re garding upgrades. Concurrent news disclosures Other evidence presented in section 2.1 shows that the market very rapidly adjusts to new information and hence it will be reflected in stock prices rather quickly (Fama et. al. 1969). So if new information about a certain company becomes public and is processed into its stock price before the actual rating change is announced, CRAs can only add value  [12]  to this and hence the stock price will not change significantly. Rating revisions seem to follow information releases which are already publicly known and therefore CRAs do not provide the market with new information and no abnormal returns will be observed around these rating change-announcements [Weinstein (1977) and Wakeman (1978)]. Several studies agree on this by stating it in a somewhat different way. As Linciano (2004) proves; the abnormal returns present in periods surrounding rating changes are a reaction to concurrent news disclosures rather than to rating change-announcements. So rating changes per se do not a bnormally influence stock prices. Other studies by Hand et. al. (1992) and Galil and Soffer (2011) provide evidence of abnormal returns indeed disappearing around downgrades, but abnormal returns around upgrades seem to persist (Galil and Soffer, 2011) or even become more positive (Pinches and Singleton, 1978) when controlling for concurrent news disclosures. Schweitzer et. al. (1992) disagrees on the above by first stating that downgrades experience significant negative abnormal returns while upgrades are associated with positive but insignificant abnormal returns. Next they conclude that the results hold even when observations with potentially confounding events are removed from the sample (p. 249). This indicates that concurrent news disclosures around rating change-announcements do not impact effects associated with these rating revisions. Degree of publicly known information As can be concluded from the previous evidence presented by several studies, stock price effects around rating change-announcements depend largely upon the information eventually leading to the rating revision. The question here is if this information is already incorporated into stock prices by the market before CRAs revise a certain companies credit rating. When this is not the case, stock prices will logically become more affected by the announcement compared to rating changes which are already expected by the market. Hsueh and Liu (1992) and Schweitzer et. al. (1992) provide evidence of this statement by concluding that stock price effects around rating changes are only present in markets were little information about securities is available. Since fewer analysts are interested in small firms and hence they are less followed  [13]  , the market is generally in greater need of information about these firms. A companies credit rating is one piece of information which is avail able at all times and hence changes for small firms will logically be associated with greater abnormal stock price performances. Many recent studies [e.g. Dichev and Piotroski (2001), Elayan et. al. (2003) and Calderoni et. al. (2009)] provide somewhat similar evidence, but argue that the reason for the outperformance is different. For example, Dichev and Piotroski (2001) claim that small firms are low-graded more often than big firms. Therefore  [14]  prior rating-levels could explain why small firms experience greater abnormal returns around credit rating-revisions. Matolcsy and Lianto (1995) investigated the Australian market on stock price effects around rating change-announcements and provide a somewhat contrasting conclusion by stating that there seems to be no difference between small and big markets. They provide evidence consistent with studies investigating the much bigger U.S. market and therefore it can be concluded that it is not clear that credit rating research in smaller markets will generate different results (Elayan et. al., 2003, p. 338). Other markets where participants are in greater need of information are markets where conditions are less certain (i.e. markets where high levels of volatility are observed). When conditions are less certain rating change announcements are less likely to be anticipated, and hence convey more information when the market as a whole is in greater need of information (Hsueh and Liu, 1992, p. 225). Their study further investigates the subject matter and comes up with proof of significant stock price effects around rating changes only for firms which experience high levels of volatility. Prior rating-levels Several studies presented above state that stock price effects around rating changes depend upon the rating-level prior to the announcement. Jorion and Zhang (2006) claim that prior ratings actually are the most important variables needed to analyze abnormal returns around rating revisions. Along the same line with other studies [i.e. Kliger and Sarig (2000) and Dichev and Piotroski (2001)] they conclude that lower prior rating-levels are generally associated with larger abnormal returns. As they claim these effects are present for upgrades as well as downgrades, Dichev and Piotroski (2001) present evidence of greater effects of low-rated debt only around downgrades. Furthermore, Jorion and Zhang (2006) claim that when prior rating-levels are taken into account, the investment grade-barrier  [15]  -effect seems to disappear, as results become insignificant. After stating that the prior rating-level needs to be taken into account while analyzing stock price effects around rating changes, the question why this prior rating is so important can be answered according to a theoretical explanation. In their paper, Jorion and Zhang (2006) best explain this as follows: For example, a downgrade from AA- to A+ should have much less information content than a downgrade from BB- to B+. In the former case, the probability of default is very small and is hardly affected. The second case, however, represents a much larger increase in default probability, is reflected in larger changes in bond yield spreads, and should have a larger impact on stock prices. Accordingly, if downgrades more often start from lower ratings than upgrades, it is not surprising to observe an overall stronger stock price effects for downgrades (p. 4). In addition they find that their sample distribution of prior rating-levels for up- and downgrades differs. When controlling for these prior ratings, also upgrades tend to be associated with significant results. These findings therefore largely expla in the earlier described asymmetry of stock price effects between up- and downgrades. Short and long term credit ratings Generally, studies investigating announcement effects around rating changes pay little attention to rating revisions regarding short term credit ratings. Still there are some studies investigating short term rating revisions and provide evidence of their importance. First, Nayar and Rozeff (1994) find that there are actually no differences between abnormal returns around short and long term credit ratings. They argue that downgrades are associated with significant negative abnormal returns while upgrades seem to give no reaction, which is consistent with evidence presented in section 2.2. A later study by Barron et. al. (1997) states that if any effects are observed concerning changes in short term credit ratings, they are counter-intuitive and insignificant. This would imply that insignificant positive abnormal returns are associated with downgrades while insignificant negative abnormal returns are associated with upgrades, which contradicts earlier findings. Hypotheses Previous chapter reviewed literature on the effect of several rating change-announcements on stock prices of different companies. This chapter will discuss and create hypotheses to check statements in that review. The first section presents two hypotheses constructed to test whether rating changes are associated with abnormal returns around the announcement and the difference in results between revisions occurring before and after the Financial Crisis. But as the main question arising from existing literature is if the rating change-announcements per se provide the market with new valuable information, section 3.2 will discuss three hypotheses constructed to test the informational content of rating changes. The last section discusses the hypothesis constructed to check if the earlier explained reason for the asymmetry between up- and downgrades, presented by Jorion and Zhang (2006) and described in section 2.7, also holds in this study. Rating changes and the Financial Crisis Different studies, whether recent or some time ago, come up with rather mixed results regarding announcement effects around changes in credit ratings. Some studies argue that rating changes do not abnormally influence stock prices while others argue that CRAs actually do provide the market with new valuable information and hence stock prices will be affected. Furthermore, there are studies that provide evidence of an asymmetry in results between up- and downgrades. They argue that downgrades seem to be associated with abnormal stock price returns while no significant effects can be observed around upgrades. To check whether rating changes indeed are surrounded by abnormal returns and if the asymmetry between up- and downgrades also applies for Benelux-stocks, the first hypothesis to be tested will be as follows: Downgrades, compared to upgrades, are associated with greater abnormal stock price returns surrounding the announcement. Also placements on the CreditWatch-list are considered to be rating changes as can be concluded from section 2.2. The first hypothesis will also be tested in each of the following hypotheses (i.e. hypotheses II to VI) where the difference in effects between up- and downgrades will be discussed. According to many critics, CRAs played an important role in the recent global Financial Crisis by assigning incorrect ratings to a variety of financial products. Because of their bad work, it may be that nowadays investors have less confidence in these CRAs and hence do not react to rating changes to the same degree as they did before the turmoil. To test whether this loss of confidence is reflected in abnormal stock price performances around the announcement of rating changes, the following hypothesis is constructed: Rating changes occurring after the Financial Crisis experience a smaller abnormal stock price performance around the announcement. Together, the two hypotheses above should give a clear insight into the impact of rating change-announcements on company stock prices and whether the Financial Crisis changed the credibility investors attribute to the work done by CRAs. Results will be presented and discussed in section 5.1. The informational content of rating changes As it is possible that rating changes are anticipated by the market, stock prices prior to the announcement could follow a pattern according to the direction of the rating change. Some studies do find this abnormal stock price performance prior to the revision while others argue that this cannot be observed. If this pre-announcement pattern is present the possibility exists that the market already incorporated the information eventually leading to the rating change into the stock price and hence this price will not abnormally be affected anymore by the announcement per se. To test this latter statement the following hypothesis is constructed: Rating change-announcements preceded by abnormal stock returns in line with the direction of the rating change experience smaller announcement effects. Other evidence presented in the previous chapter suggests that abnormal returns might be caused by other events rather than due to the rating change itself. When news-announcements contain pricing relevant information and they are disclosed closely around the announcement of a rating change, it could be that these concurrent news disclosures create abnormal returns while the rating change per se does not influence the price of a stock. To check whether rating change-announcements in themselves abnormally impact the price of a stock, the next hypothesis to be tested will be as follows: Concurrent news-disclosures occurring in the period from two days before till two days after the announcement of a rating change weaken results concerning the abnormal stock price performance. Next, section 2.6 shows that if rating changes are associated with abnormal returns around the announcement it depends largely upon the degree of information available to the market about the particular stock (Hsueh and Liu, 1992). Stocks for which the market is in greater need of information (i.e. stocks of small and/or high volatility firms) should be accompanied by greater price movements around the announcement of a rating change. To test whether small and/or high volatility firms indeed are associated with greater abnormal returns, the following hypothesis is constructed: Firms which are small capitalized and/or experience low levels of volatility are associated with greater abnormal stock price returns around the announcement of a rating change. The above described hypotheses together should give a good understanding about the informational value of rating change-announcements. Results regarding these hypotheses according to the tests presented in the next chapter will be discussed in section 5.2. Asymmetry and prior rating-levels From section 2.7 it can be concluded that the rating level prior to the rating change is a variable that needs to be included while analyzing the abnormal performance of a stock price around the announcement. On top of that, Jorion and Zhang (2006) claim that it is even the single, most important variable and that it significantly changes the outcomes of their study when controlling for these prior rating-levels. They claim that a lower level of prior rating is associated with a greater abnormal stock price performance because of greater increases (decreases) in default probabilities to upgrades (downgrades) starting from a lower level. This evidence could explain why there seems to be an asymmetry between up- and downgrades. If, for example, downgrades more often start from a low rating-level  [16]  results will be influenced and presented evidence is contaminated. To test whether a low-start indeed is associated with a greater abnormal stock price performance around the annou ncement of a rating change, the following hypothesis is constructed: A rating change starting from a level below the investment grade-barrier is associated with a greater abnormal stock price return around the announcement. Results regarding this last hypothesis will be presented and discussed in section 5.3. The next chapter discusses the construction of the dataset and the methodology used to check the correctness of the hypotheses presented above. Data and Methodology In order to be able to test the previous presented hypotheses a dataset is constructed. Which data is included and how these are gathered will be explained in the first section. Section 4.2 will explain the event study-methodology that is chosen to check the correctness of the hypotheses. Dataset construction Rating changes used in this research are gathered from the Bloomberg-database. All rating changes made by the three major CRAs for Benelux-companies listed on respectively the BEL20-, AEX-, and LuxX-index during the period between January 2002 and December 2011 are collected. If a certain company is listed during, for example, the years 2002 to 2005 but delisted afterwards, the changes in credit rating made after delisting are excluded from the sample. Both long and short term credit ratings  [17]  are considered in this research. A list of all rating types included in the dataset can be found in Appendix (III). Ratings which are issued, withdrawn or remain stable (i.e. implying no real change in credit rating can be observed) are not considered rating changes and hence they are excluded from the sample. Changes from and to CreditWatch (e.g. from AA+ *- to AA+) are also considered rating changes as can be implied from the evidence presented in section 2.2. The first hypothes is to be tested leaves the dataset unchanged, only the lines for which the CAAR (-1,+1) cannot be calculated  [18]  are deleted from the sample. How this CAAR (-1,+1) is calculated will further be explained in section 4.2. In order to test the second hypotheses the column Crisis is added and presents whether the rating change occurred before or after the date considered to be the beginning of the Financial Crisis, which is January the 1st of 2008. For the third hypothesis the column In Line is added to the dataset and shows whether the rating change-announcement is preceded by abnormal returns in line with the direction of the rating change. These preceding abnormal returns are calculated in a window from 29 to two days before the announcement (i.e. the period between the event and estimation window as can be inferred from figure 1). A further explanation of calculations concerning abnormal returns will also be provided in the next section. The fourth hypothesis needs concurrent news-disclosures which are gathered from www.FD.nl  [19]  and screened on their importance regarding a companies stock price. The column that reads the name News shows whether an article disclosed in the period between two days before and two days after the rating change is announced contains pricing relevant information. A list with a sample of news-announcements which are considered to be of importance for a companies stock price can be found in Appendix (IV). Next, whether a company is small or large capitalized and whether it experiences low or high levels of volatility is presented in the seventh and eight column as is presented in Appendix (V) that shows a sample of the total dataset used. Market capitalization and level of volatility are needed for hypothesis V and are extracted from Datastream. Dividing lines used to the determine levels of capitalization and volatility are respectively a market capitalization of 10 million and an average annual stock price volatility o f 30%. The last column shows whether a company is considered to have low-graded debt before the rating is changed which is needed for hypothesis VI. The investment grade-barrier is used as a dividing line here. All together, a final dataset of 368 unique rating changes  [20]  is constructed. Table 1 below shows descriptive statistics concerning the sample distributions. Table 1: Sample descriptive statistics This table shows the number of up- and downgrades present in each subsample for the different hypotheses presented in chapter 3 consisting of all rating changes made by Fitch, Moodys and SP in the period between 2002 and 2011 for Benelux-companies listed on respectively the BEL20-, AEX- and LuxX-index (A further explanation of sample-names presented below can be found in section 4.1) Sample Upgrades Downgrades Total Before Crisis 89 88 177 After Crisis 50 123 173 In Line 78 138 216 Not in Line 61 73 134 With News 58 110 168 Without News 81 101 182 Small Cap 65 141 206 Large Cap 65 60 125 Low Volatility 62 80 142 High Volatility 55 113 168 Low Start 53 33 86 High Start 87 187 274 Total Unique  [21] 139 211 350 Event study methodology The methodology used in this research follows the event study methodology described by De Jong and De Goeij (2011). In their handout they identify three steps in conducting an event study. This study will divide step three into two steps and hence a fourth step will be added. These four steps are: Identifying the event. Specifying a benchmark to calculate normal returns. Calculating abnormal returns for the event window. Analyzing abnormal returns by testing their significance. In the remainder of this section the above steps will be discussed separately in the order showed above. Results arising from the calculations of step three and four are presented and discussed in chapter 5. Step I: Indentifying the event The first step in conducting an event study concerns the identification of the specific event. It would not be wise to use the actual date of the rating change as the event because the rating change often is announced much earlier and hence market reactions take place long before the actual rating is changed. Therefore the date on which a rating change is announced is considered to be the event in this study. These events  [22]  are extracted from the Bloomberg-database where both short and long term credit ratings are taken into account. Step II: Specifying a benchmark The second step concerns the specification of a benchmark for estimating normal returns. Figure 1 below shows the time line around the event. The second and third step in conducting the event study will be discussed according to this figure. Figure 1: Time line used in event study This figure shows the windows for estimating normal (T1,T2) and abnormal (t1,t2) returns around the event (t=0) the event T1 T2 t1 (t=0) t2 Estimation Window Event Window For specifying normal returns, De Jong and De Goeij (2011) present three methods. First they introduce the mean-adjusted return model. This model uses the average stock price return over the estimation window as a benchmark return. In this study, events occur within days from each other and therefore the estimation window of one rating change could include the announcement of another revision which clearly contaminates the results. Another disadvantage of the model is that it does not take the market return into account, i.e. it does not control for the fact that stock returns might include abnormal performance due to the whole market being in an up or down state. For both reasons, the mean-adjusted return model will not be used in this research. The second method mentioned is the market -adjusted return model. In this model the return on the market is used as a benchmark and hence it clearly controls for the state of the market. But as this second method assumes the beta of each stock is equal to one, which clearly is incorrect, a third method is presented. This method, the market model residuals, is used in this research and defines normal returns as follows: (1) Where the normal returns are calculated by taking the corresponding market returns (Rmt) from the event window and by deriving Ordinary Least Squares (OLS) estimators and from following the regression coefficients: (2) Where estimators and are calculated by using the following formula: (3) (4) In equation (2) to (4), Rit and Rmt are returns from respectively the specific stock and the market index in the period between 90 to 30 days (i.e. the estimation period) before the announcement of a rating change. MSCI indices are used as a market index in this research. For Belgium this is the MSCI Belgium index, for the Netherlands the MSCI Netherlands index and for Luxembourg the MSCI Small Country index as the MSCI Luxembourg index is not available for the years 2002 to 2011. All returns on the above indices are obtained from Datastream and calculated using the following formula: (5) Step III: Calculating abnormal returns In the third step the abnormal performance of a specific event is estimated. The abnormal returns (ARit) from the day before until the day after the event (i.e. the event window from t1 to t2) are calculated by using the following formula: (6) Next the abnormal returns for a specific firm at a specific point in time are aggregated over the period covering the event window. The formula to calculate these cumulative abnormal returns (CARi) is defined as follows: (7) As De Jong and De Goeij (2011) state: In order to study stock price changes around events, each firms return data could be analysed separately. However, this is not very i nformative because a lot of stock price movements are caused by information unrelated to the event under study. The informativeness of the analysis is greatly improved by averaging the information over a number of firms (p.7). So in order to be able to test CARs on their significance, these CARs are aggregated over the cross-section of events to get a cumulative average abnormal return (CAAR). The formula used to calculate this CAAR is defined as follows: (8) Where N is the number of firms that are included in the specific sample and for which the CAR could be calculated. There are also other ways to calculate this CAAR (e.g. by aggregating average abnormal returns over time), but for simplicity these will not be explained. Step IV: Testing abnormal performance The fourth and final step in conducting the event study concerns the testing of the abnormal performance around the event date. To test the significance of this abnormal performance the following null hypothesis is c onstructed: (9) This hypothesis is constructed to test whether the cumulative abnormal returns are significantly different from zero (i.e. an announcement effect is present). To check whether this actually is the case, the calculated CAAR needs to be tested on its significance and therefore the t-test is defined as follows: (10) Where s is the standard deviation of CAAR, calculated by using the following formula: (11) The t-test presented in equation (10) assumes that abnormal returns of the different events are uncorrelated. But as mentioned earlier, events tend to be clustered  [23]  and hence there is a potential for correlation between the cross section of abnormal returns. In that case, the variance of the average of N abnormal returns is no longer equal to 1/N times the variance of a single return, but larger (if the correlation is positive, as it typically is). As a consequence, the usual variance estimator underestimates the variance of the average abno rmal returns, the usual t-statistics are biased upwards, and the null hypothesis is rejected too often [De Jong and De Goeij (2011), p.13]. To correct for this cross-sectional dependence, Brown and Warner (1980) introduce the so called crude dependence adjustment-method (hereafter called CDA). This method calculates a new variance directly estimated from the time series of observations of average abnormal returns in the estimation period. The corresponding test statistic for this method is calculated by using the following formula: (12) Where T is equal to T2 T1 + 1, which is equal to the number of trading days within the estimation window, and is the newly obtained standard deviation calculated by using the following formula: (13) Where the average abnormal return (AARt) and the sample average AR* are calculated as follows: (14) (15) Results of equations (8), (10) and (12) will be presented and discussed in the next chapter. Empirical Results The tables in this chapter show Cumulative Average Abnormal Returns (CAAR), their significance levels and CDA-test results over the period from one day before till one day after the announcement of a rating change. Statistics added with a (*) are significant at the 5% level  [24]  . Conclusions that can be drawn from these results are presented in the next chapter. Announcement effects: Rating changes and the Financial Crisis This section covers the results regarding hypothesis I and II. Table 2 and 3 present these results and subsequent a discussion will be provided. Table 2: Results hypothesis I This table shows CAARs, T-Statistics and CDA Test-Results for the sample consisting of only upgrades and the one consisting of only downgrades extracted from the sample consisting of all rating changes made by Fitch, Moodys and SP in the period between 2002 and 2011 for Benelux-companies listed on respectively the BEL20-, AEX- and LuxX-index. CAAR (-1,+1) Upgrades Downgrades -0.0041 -0.0372 Significance Level Upgrades Downgrades -0.7233 (*) -3.2807 Crude Dependence Upgrades Downgrades -0.1800 (*) -2.0144 Results in table 2 present evidence of a highly significant abnormal return (i.e. -3.72%) around the announcement of a downgrade while there seems to be no evidence on this for upgrades. Even when correcting for possible cross-sectional correlation these results seem to hold. This evidence is consistent with earlier studies by Schweitzer et. al. (1992) for the US, Barron et. al. (1997) for the UK and Romero and Robles-Fernandez (2007) for Spain, who all conclude that rating change-announcement do convey new valuable information to the stock market and hence stock prices will abnormally be affected. Furthermore, the above evidence implies that the earlier explained asymmetry between up- and downgrades, presented by Dichev and Piotroski (2001), Vassalou and Xing (2005) and Halek and Eckles (2010), also seems to hold for Benelux-stocks. Table 3: Results hypothesis II This table shows CAARs, T-Statistics and CDA Test-Results for all up- and downgrades present in the sample with rating changes occurring before and the sample with rating changes occurring after the Financial Crisis extracted from the sample consisting of all rating changes made by Fitch, Moodys and SP in the period between 2002 and 2011 for Benelux-companies listed on respectively the BEL20-, AEX- and LuxX-index. CAAR (-1,+1) Upgrades Downgrades Before Crisis 0.0036 -0.0446 After Crisis -0.0178 -0.0319 Significance Level Upgrades Downgrades Before Crisis 0.7645 (*) -1.9438 After Crisis -1.3273 (*) -3.0369 Crude Dependence Upgrades Downgrades Before Crisis 0.1179 -1.5623 After Crisis -0.3629 -1.2742 Table 3 presents results regarding announcement effects for both up- and downgrades present in the sample consisting of rating changes occurring before and the sample with rating changes occurring after the Financial Crisis. While upgrades do not yield any significant results both before and after the crisis, downgrades are associated with significant announcement effects of respectively -4.46% and -3.19%. This significance seems to disappear when correcting for cross-sectional dependence. As the above findings indicate, the abnormal stock price performance around the announcement of a downgrade differs considerably when comparing the two samples. After the Financial Crisis, investors seem to react less to the announcement of a downgrade which can be inferred from the announcement effect being reduced by 1.27%. These findings are in line with earlier presented evidence by Calderoni et. al. (2009) althoug h it indicates that Benelux-investors still react considerably. Announcement effects: The informational content of rating changes This section discusses results regarding the hypothesis III to V which are constructed to test the informational content of rating changes. Subsequent to every table a discussion of the results is provided. Table 4: Results hypothesis III This table shows CAARs, T-Statistics and CDA Test-Results for all up- and downgrades present in the sample with rating changes that experience pre-announcement abnormal stock returns in line with the direction of the rating change and the sample without these pre-announcement returns extracted from the sample consisting of all rating changes made by Fitch, Moodys and SP in the period between 2002 and 2011 for Benelux-companies listed on respectively the BEL20-, AEX- and LuxX-index. CAAR (-1,+1) Upgrades Downgrades In Line -0.0044 -0.0086 Not in Line -0.0240 -0.0744 Significance Level Upgrades Downgrades In Line -0.5222 -1.5982 Not in Line -1.2162 (*) -2.7061 Crude Dependence Upgrades Downgrades In Line -0.1250 -0.3400 Not in Line -0.6955 (*) -2.3372 Table 4 presents results for up- and downgrades present in the sample with and the sample without rating changes that experience pre-announcement abnormal stock returns in line with the direction of the rating change. The results indicate that both up- and downgrades with these pre-announcement returns experience minor and insignificant abnormal returns in a three-day-window around the announcement of a rating change. When these pre-announcement abnormal returns cannot be observed, the abnormal returns become greater and more significant for both up- and downgrades. While downgrades experience a highly significant (even when controlling for cross-sectional dependence) announcement return of -7.44%, the abnormal return around upgrades seems to remain insignificant. The above evidence contradicts findings by Goh and Ederington (1999) who conclude that announcement effects are greater with negative pre-announ cement abnormal returns in case of downgrades. While they conclude that the market reacts more heavily to these downgrades because of a more credible signal, this study presents evidence of a significant reaction only to downgrades that come as a surprise which is consistent with evidence presented by Gropp and Richards (2001). Both Goh and Ederington (1999) and this study do not find any significant evidence regarding upgrades. Table 5: Results hypothesis IV This table shows CAARs, T-Statistics and CDA Test-Results for all up- and downgrades present in the sample with concurrent news disclosed in the period from two days before till two days after the announcement of a rating change the sample without these concurrent news-disclosures extracted from the sample consisting of all rating changes made by Fitch, Moodys and SP in the period between 2002 and 2011 for Benelux-companies listed on respectively the BEL20-, AEX- and LuxX-index. CAAR (-1,+1) Upgrades Downgrades With News -0.0074 -0.0717 Without News -0.0018 0.0004 Significance Level Upgrades Downgrades With News -0.5800 (*) -3.4739 Without News -0.4905 -0.0680 Crude Dependence Upgrades Downgrades With News -0.1510 (*) -2.7079 Without News -0.0791 -0.0131 Table 5 provides evidence of a cumulative average abnormal return of -7.17% around downgrades for announcements which are surrounded by the disclosure of other concurrent news. Because the sample without these disclosures does not yield any significant returns, this result implies that investors in the Benelux do not view a standalone downgrade-announcement as a credible pricing relevant signal. Even when controlling for the correlation between the cross section of stock returns, the results remain significant which can be inferred from a test statistic of -2.7079. Because the variable News is not created by whether the disclosure has positive or negative effects on the companies stock price, one cannot conclude that rating changes invigorated by a similar pricing relevant news-announcement is considered a more credible signal by the market. Again, for upgrades no evidence is found. The above findings a re very similar to the ones presented by Linciano (2004) who states that rating changes per se do not abnormally influence the price of a companies stock. Other studies by Hand et. al. (1992) and Galil and Soffer (2011) confirm this result but also conclude that announcement effects around upgrades tend to become more positive when no concurrent news disclosures can be observed around the announcement of rating change. This cannot be concluded in this research as can be derived from table 5. Table 6: Results hypothesis V(1) This table shows CAARs, T-Statistics and CDA Test-Results for all up- and downgrades present in the sample with small and the sample with large capitalized firms extracted from the sample consisting of all rating changes made by Fitch, Moodys and SP in the period between 2002 and 2011 for Benelux-companies listed on respectively the BEL20-, AEX- and LuxX-index. CAAR (-1,+1) Upgrades Downgrades Small Cap -0.0118 -0.0478 Large Cap -0.0023 0.0007 Significance Level Upgrades Downgrades Small Cap -1.1435 (*) -3.2074 Large Cap -0.5876 0.0584 Crude Dependence Upgrades Downgrades Small Cap -0.2325 (*) -2.0525 Large Cap -0.1385 0.0251 Table 6 presents results regarding a sample consisting of firms which are small and a sample of firms which are large capitalized where each sample is divided into up- and downgrades. From the table it can be concluded that only small capitalized firms experience significant announcement effects of -4.78% around downgrades while for large firms results are even slightly positive but greatly insignificant at the 95% confidence level. Executing the CDA-test reduces the test statistic considerably but still shows a significance level of -2.0525. Upgrade-results are counter-intuitive  [25]  but again insignificant, hence no conclusions can be drawn from these results. The above results claim that small firms experience greater abnormal stock price performance around the announcement of a downgrade. This conclusion is rather logical as small firms tend to be less followed and hence the market is in greater n eed of information about these stocks. When CRAs provide new information to the market through their rating activities, market participants will act accordingly and hence stock prices will be affected. The evidence above is in line with earlier presented literature by Dichev and Piotroski (2001), Vassalou and Xing (2005) and Elayan et. al. (2003), outside the fact that the latter study does find evidence of small firms being associated with significant abnormal stock price performances around the announcement of upgrades. Table 7: Results hypothesis V(2) This table shows CAARs, T-Statistics and CDA Test-Results for all up- and downgrades present in the sample with firms that experience low and the sample with firms that experience high levels of stock price volatility extracted from the sample consisting of all rating changes made by Fitch, Moodys and SP in the period between 2002 and 2011 for Benelux-companies listed on respectively the BEL20-, AEX- and LuxX-index. CAAR (-1,+1) Upgrades Downgrades Low Volatility 0.0026 0.0001 High Volatility -0.0176 -0.0581 Significance Level Upgrades Downgrades Low Volatility 0.9799 0.0138 High Volatility -1.4101 (*) -3.1357 Crude Dependence Upgrades Downgrades Low Volatility 0.2219 0.0066 High Volatility -0.2967 (*) -1.8427 Table 7 shows that only firms which experience high levels of annual stock price volatility face significant abnormal stock price performances around downgrades amounting to -5.81%. Even when correcting for potential cross sectional dependence this result remains significant at the 95% confidence level. Low volatility firms seem to experience no abnormal return around the announcement of a downgrade, but as the significance level is rather close to zero nothing can be concluded here. Upgrades tend to yield somewhat same results, relatively speaking (i.e. small effects for low volatility firms and greater effects for firms which experience high levels of volatility). But as can be derived from the table these results are again insignificant. Hsueh and Liu (1992) provide somewhat same evidence in their study by stating that rating changes only yield significant announcement effects when market condi tions are less certain and hence the market is in greater need of information. But, as this stu