• No results found

The impact of credit rating announcements on Credit Default Swap Notional Amounts

N/A
N/A
Protected

Academic year: 2021

Share "The impact of credit rating announcements on Credit Default Swap Notional Amounts"

Copied!
39
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

The impact of credit rating announcements

on

Credit Default Swap Notional Amounts

Master Thesis

June 2017

Abstract

This thesis tries to answer the question if there is a correlation between CDS net notionals and credit rating announcements. Using the cross-sectional regression with several control variables, I found that there is a significant increase in CDS net notionals in the event month, being the result of upcoming downward changes. The effect is opposite when the rating is changed to junk grade and no effect has been found in case of upgrades. As the next step, the event study is carried out to investigate the abnormal changes in the event window around the rating change. The results indicate that the increased demand for CDS net notionals as a result of credit rating deterioration is mostly concentrated in the pre-event months, but it can be observed as well in subsequent ones. The abnormal changes are most visible in case of downgrading either highest or lowest quality rankings, making it possible to anticipate the upcoming credit rating announcements. The effect of positive rating changes in the event study is not clear.

Michał Szkiłądź 11376546 MSc Finance Corporate Finance track

(2)

2

Statement of Originality

This document is written by Student Michał Szkiłądź who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Table of Contents

1. INTRODUCTION ... 4

2. LITERATURE REVIEW ... 6

2.1. GENERAL VIEWS ON CREDIT DEFAULT SWAPS ... 7

2.2. IMPACT OF CREDIT RATING ANNOUNCEMENTS AND PRICE DISCOVERY ISSUE ... 9

2.3. TOPIC RELEVANCE AND HYPOTHESES ... 10

3. DATA ... 11

3.1. SUMMARY STATISTICS ... 14

4. EMPIRICAL EVIDENCE ... 15

4.1. REGRESSION METHODOLOGY ... 16

4.2. REGRESSION RESULTS ... 19

4.3. EVENT STUDY METHODOLOGY ... 20

4.4. EVENT STUDY RESULTS ... 22

5. ROBUSTNESS CHECKS ... 24

6. CONCLUSIONS ... 26

7. REFERENCES ... 28

APPENDIX ... 38

(4)

4

1. Introduction

“Only when the tide goes out do you discover who’s been swimming naked”- Warren Buffett is known to have said once and these words have been cited hundred of times in relation to 2008 financial crisis. While during the tranquil periods, the usefulness and reliability of all the innovations, especially these related to financial markets is unquestionable, only by proving itself in unusual conditions when market perturbations occur, the product can give credible evidence on its long-term utility. Therefore, the initial sentence describes perfectly the essence of Credit Default Swaps (CDS) contracts whom this dissertation is devoted to.

These days, CDS are widely known mostly because of their contribution to the credit crunch. Their role was also widely condemned during Argentinian and Greek crises, when increased demand for “naked” (when buyer is not exposed to the credit risk on the underlying position, but enters into the contract for speculative motives) sovereign CDS contributed to significant increase in borrowing costs of these countries.

Credit Default Swaps are derivative instruments based on agreement between the two parties, ensuring that in case of credit event (such as loan default), the seller of the CDS will compensate the buyer of that instrument. The idea is, that in exchange for a periodic payments, CDS seller insures the buyer, so that the CDS contracts are used to transfer the credit risk between two counterparties. One of the most important issues regarding credits is the fact that the overall risk exposure is to the large extent affected by interest rate risk. Credit default swaps enable to separate these risks and contribute to diminishing the level of uncertainty in the outstanding positions. However, despite being criticized by many, CDS instruments are willingly used by speculators on the OTC market, where they are mostly traded, enabling the investors to reap higher benefits because of lower transaction costs. The contracts are terminated when a credit event occurs. CDS contracts present an innovative way of hedging against the credit risk as they can be regarded as the equivalents of bonds shortselling and reinvesting the proceeds at the riskless rate. Comparing to that solution, CDS contracts are much more attractive considering the fact that they eliminate

(5)

5

the risk associated with rolling over short positions. Similarly to Treasury notes and bond futures, the buyer of the protection usually can choose the delivery method (he may prefer bond to loan) to be delivered in case of credit event. There is a plenty of research proving that Credit Default Swap (CDS) market’s contribution to price discovery is higher comparing to stock as well as bond market what is described in more detail at the later stage of this paper. It has been also empirically covered that the CDS spreads could be seen as a good predictor of creditworthiness of the specific company. In fact CDS market does not only unveil the potential risks faster than other ones, but also reacts often before the official statements are disclosed. As the result of this connection, CDS spreads tend to fluctuate due to the public and private information before rating announcements. Despite the fact that most of the existing academic research is based on measuring the effects of CDS spread changes (Hull, 2004; Daniels and Jensen, 2005), there is a shortage of papers that comprehensively analyze changes in CDS net notional amounts1 as well as trading volume. The

first paper on this topic committed by Oehmke and Zawadowski (2016) indicates that both CDS positions and trading volume are larger when there have been multiple issuances of corresponding bonds, different in terms of their payoff conditions. In the second one written by Berg and Streitz (2016), authors analyze the impact of credit rating changes on CDS contracts but their research focuses on sovereigns. In their paper they document that downward rating changes as well as negative outlooks cause significant increase in CDS trading volume, but the amount of outstanding contracts captured by CDS net notionals remains invariable. The explanation provided by the authors behind this phenomenon is that the announcements lead to reallocation of the swaps rather than an increase or decrease in overall protection.

This thesis is to evaluate the changes in CDS net notional amount which are potentially the result of credit rating announcements. The area of the research is different from previous empirical

1 CDS net notionals amount represent the maximum possible net funds transfers between sellers

of protection and buyers of protection that could be required upon the occurrence of a credit event.

(6)

6

studies mostly due to the fact that it focuses on 5-year event study, checking the relation between CDS net notionals at reference entity level and credit rating changes. The aim of my research is to state if it is associated with increased or decreased demand for CDS contracts, which can be highly speculative, considering that after rating downgrade there is a significant increase in CDS spread as the credit risk rises. Differences in CDS net notionals are supposed to reflect the changes in risk related to the creditworthiness of the particular entity. The main hypothesis is to test whether the change in net notional amount of CDS is significantly correlated with credit rating announcement and if so, how fast the significant increase or decrease in number of issuances can be observed. As the analysis has descriptive characteristics to the large extent, the main goal of the thesis is to provide set of results and tests that would lay a foundation for a further future research regarding Credit Default Swap market.

2. Literature review

In this section relevant literature regarding Credit Default Swaps is discussed. The first paragraph presents an overall view on the CDS instruments as well as points out different views on pros and cons of this derivative in relation to the financial markets. The explicit trends regarding Credit Default Swaps have been pointed out to give a broader outlook on the case for a reader. In the second paragraph the summary of the empirical evidence showing impact of credit rating announcements on different markets is presented. This section discusses also the academic work regarding price discovery proving the importance of credit risk market. The last paragraph explains thesis relevance by pointing out the previous research regarding the impact of credit rating changes on CDS contracts. This section contains also the hypotheses tested on the pages of this dissertation. Additionally, the outline of the further research is discussed.

(7)

7

2.1. General views on Credit Default Swaps

CDS instruments were invented in mid-nineties and their prevalence has been rising up to the end of 2007, when the outstanding amount peaked at the level of $62.2 trillion (ISDA Market Survey). During that time CDS contracts have grown to one of the most significant and innovative financial instruments of the past decades. A lot of controversies around them arose mostly due to the fact that they can be bought without possessing the underlying security, which makes it possible for CDS market to outstrip the underlying debt market in terms of value. The good example of that is the portfolio held by Lehman Brothers, which as reported by the Financial Times, could have amounted to even $400 billion (FT.com, 2008). Moreover, CDS are criticized by many because of their role in facilitating the creation of synthetic securitized products such as mortgage backed securities (MBS) or collateralized debt obligations (CDOs), that enable the investors to get the credit exposure of a portfolio of fixed assets without actually owing them. Another aspect are the capital requirements that allowed the banks to hold less core equity capital by entering into CDS contracts. By laying off some portion of the risk, the banks were able to free up the capital that otherwise they would have been forced to hold because of higher levels of risk-weighted assets. Among all the negative opinions, the empirical evidence on the contribution of CDS contracts to the credit crunch is contradictory. The proponents of CDSs base their opinions mostly on the fact that not the derivatives were the problem themselves, but lack of the regulation and the fact that these instruments are traded solely over the counter. Stulz (2010) concludes that CDS neither caused the financial crisis, nor exacerbated it. In his paper, the author points out that over-the-counter credit default swaps market worked well during most of the financial crisis years. When the losses in underlying mortgage securities soared to the astonishing levels, it remained still liquid facilitating dealing with the defaults in the efficient way.

In fact, Credit Default Swaps can be seen as a risk management tool, because they create a possibility of diversification or hedging against the credit exposure. Greater risk distribution

(8)

8

provided by CDS contracts improves the ability of the financial markets to absorb the shocks, help the financial institutions to manage their risk exposures (usually specified by regulatory authorities), and gives the investors broader investment spectrum. The evidence on the positive effects of these instruments were empitically confirmed by Kiff (2009) who argues, that the time of distress, when the liquidity of the bonds is largely limited being reflected in higher bid ask spreads, the CDS market still remains highly liquid. However one of the most prevalent arguments against these swaps is that the risk is transferred from the parties that specialize in its management (such as banks) to the parties that have less experience and abilities of accessing it. This issue brings us to the question if Credit Default Swaps are the source of the moral hazard and therefore contribute to excessive risk taking from the banks’ side. The answer is dubious, as all the hedging activities come with expenses, so that the banks cannot shift risk without incurring additional costs.

Moreover, there is extensive amount of academic work discussing welfare implications of CDS from both corporations and regulators perspectives. The flexible nature of Credit Defaut Swaps enables the creation of new synthetic instruments which contribute to the emergence of bow-tie network as a result of large intermediation levels (D’Errico, Battiston, Peltonen, Scheicher; 2016). In over-the-counter markets, contrary to centrally organized ones where the quotes are available for investors, the trades are made bilaterally and information flow is opaque, so that the role of the dealers is quite significant as they facilitate the supply and demand to meet. The authors prove in their paper, that in this structure of strongly connected dealers, the probability of a loss due to materialization of counterparty risk is much higher than in the fragmented networks. As a result, fragile web of mutual exposures is created. If one company fails to meet its obligations, many others immediately start to face financial problems. Further research shows, that even 75% of market gross notional traded at OTC relates to excess (D'Errico, Roukny; 2017).

(9)

9

2.2. Impact of credit rating announcements and price discovery issue

There is a large amount of research documenting impact of credit rating announcements on both bonds and stock prices. Steiner and Heinke (2001) document that the abnormal eurobond returns are closely associated with the announcements of rating changes and watchlistings by Standard and Poor’s and Moody’s. The significant reactions on bonds price can be observed in case of both negative watchlistings and downgrades (downgrades into speculative grade are found to have even stronger effect), but the effect is not visible for positive outlook changes. Furthermore, Hull (2004) shows that there is a direct effect on credit risk market which can be observed by investigating the correlation of rating changes with CDS spreads. The paper documents, that the CDS spreads react not only at the moment of announcement, but the effects can be observed even beforehand in case of negative announcements. Therefore, the spreads provide helpful information in estimating the probability of negative credit rating changes. Surprisingly the effect of positive ones still remain not clear, as there are several opinions considering their impact on CDS spreads. The relation between the credit rating changes and the changes in number of corporate contracts outstanding is still unknown.

In addition the speculative aspect is also to be mentioned, hence betting on the creditworthiness increases not only the liquidity on the market but has also a positive impact on the pace of price discovery. In the past, when the market of credit derivatives did not exist, there was no benchmark to compare with when assessing the risk. Credit default swap market helps therefore to make the other markets more transparent and enhances their efficiency making them complete. Nevertheless the answer to the question which market leads the other is not simple. There are multiple contrary views on that topic, very often ambiguous, on one hand proving that the information tends to appear in the credit derivatives market before any effect can be observed in the corporate bond market (Longstaff 2003, Blanco et al., 2005; Zhu, 2006; Norden and Weber, 2009), and on the other hand that the stock market leads the CDS and bond markets (Norden and Weber, 2009). Forte and Peña (2009) find opposite results depending on the market stability. In their paper, the

(10)

10

authors present evidence that during the tranquil periods CDS market leads the others, while in the periods of the financial crisis stock market becomes the one which informational advantage prevails. Eventually, based on the extensive empirical analysis, most of the previous studies prove that credit market is characterized by faster price discovery. The above evidence obviously leads to the connection with the term called “insider trading” as lack of sufficient regulatory environment creates a possibility of above-average profits when making use of informational advantage. The outcome of my research can be the next argument in favour of previous theories, showing additional advantage (or lack of it) of Credit Default Swap market in terms of credit risk anticipation.

2.3. Topic relevance and hypotheses

In my research, I would like to investigate if CDS net notionals increase in the event window around credit rating change. Taking into consideration above empirical evidence, my hypotheses are:

H1: There is a significant increase in CDS net notionals in the months around the credit rating announcement for rating downgrade

H2: There is a significant decrease in CDS net notionals in the months around the credit rating announcement for rating upgrade

Additionally, I want to check if the effect for the downgrades to non-investment grade as well as the upgrades to the investment grade have significantly different effect from the rest of the sample. Therefore, the third hypothesis sounds as follows:

H3: The announcements have different impact on CDS notionals when the rating changes around the investment grade.

(11)

11

As it has been also proved, CDS market has signs of insider trading, especially for the firms having strong relationships with banks (Acharya and Johnson, 2007). In their paper, the authors demonstrate that during the days of negative credit information as well as for the firms that are more likely to undergo credit downgrade, information about the deterioration of the financial prospects flows first of all into the CDS market. Due to the faster incorporation of potential risks, comparing to the stock market, CDS spreads enable the speculators to predict upcoming changes in creditworthiness of the firm, which results in the change in their stock valuations. It is essential to state if the significant increase in CDS issuances can be observed in advance, what could have been similarly treated as a speculative tool. Therefore, the fourth hyphothesis is:

H4: The upcoming credit rating announcements can be anticipated in the months preceding the rating changes.

The effect can be also otherwise-the market merely reacts to the rating change and incorporates news into the prices with noticeable delay or that there is no correlation between CDS net notionals fluctuations and credit rating announcements as the effect can be spread proportionally over time. This area of this research is different from all the previous ones, because it is the only one to determine causality between the amount of CDS contracts that can be executed in case of credit event and rating announcements (both upcoming, and already released).

3. Data

The data regarding net notional amount of CDS has been obtained using Depository Trust & Clearing Corporation (DTCC) database. DTCC is a private company offering post-trade financial services that provides clearing and settlement services to the financial markets. The database presents weekly averages of CDS net notionals. As rating changes obtained using Wharton database are presented by month, monthly CDS net notionals averages were calculated to merge the data.

(12)

12

Data containing information regarding S&P credit rating changes was obtained using Compustat database. The ratings taken into consideration in this thesis are long-term local currency issuer ratings, which are widely acknowledged as an indicator of long-term financial stability of a given company. For the purpose of including the control variables in the regression, the balance sheet data has been obtained using Compustat database. All the firms using foreign currencies such as CAD in their accounting standards have been dropped out of the sample. The data regarding bonds comes from TRACE database. Similarly to the other inputs, the data was adjusted to represent monthly averages. In the robustness checks section, the analysts’ estimations regarding EPS in the given year have been obtained using IBES Academic.

Using primary SIC codes, the financial companies (6000-6300) have been excluded from the sample. While the presence of these institutions is pivotal from the market’s point of view, they usually act as dealers. Therefore, as their characteristics are different from the rest of the entities they would presumably cause interferences in the research. The final merged database consist of 410 US companies. Most of the companies come from the manufacturing sector (40,7%), Transportation, Communications, Electric, Gas and Sanitary services (20,5%) and Insurance and Real Estate sector (11%).

[Figure 1 here]

The research period covered is 01.01.2010-31.03.2015. The length of the period is assumed to be sufficient to carry out the experiment. The period of the considered sample cannot be longer mostly due to the time of financial crisis which could have been a threat to the following thesis as it could bias the outcome.

(13)

13

It is clear that during the period 2010-2015 average amount of CDS net notionals per firm dropped significantly in a given sample (average level calculated per month). The result is in line with the recent research published by International Swaps and Derivatives Association (ISDA)2, proving

the decline in CDS gross notionals. Credit Default Swap market expanded rapidly during the period of loose monetary policy and credit expansion between 2002 and 2007 and its growth slowed down as the crisis relented. While, according to ISDA, total amount outstanding on single name CDSs based on corporate and sovereign borrowers amounted to $15.3 trillion, by June 2015 this number contracted about 61% to the level of just $6 trillion. The main reason behind the increased amount of the average protection in the beginning of the period can be the financial crisis. As its aftereffects are spread over time, the number of contracts is declining gradually. The reason for a drop can be also potentially found in the series of changes to the regulatory framework, which to the large extent limited the possibility of profitable trading for specific types of derivatives. In the aftermath of the crisis the Basel III was introduced, significantly increasing the costs of Credit Default Swaps. In the countries of the European Union, the ban on naked sovereign CDS was imposed. Both in US under Dodd-Frank Wall Street Reform and Consumer Protection Act, and European Market Infrastructure Regulation in the European Union, the clearing of certain types of CDSs was enforced. The main purpose of the above regulations was to increase control over CDS over-the-counter market, being identified by many among the main grounds for the financial crisis because of numerous controversies surrounding these instruments. Additional factor on explaining diminishing amount of CDS net notionals can be dwindling default rate on corporate debt that significantly reduces the demand for hedging against credit risk.

2 Culp, van der Merwe, Staerkle, “Single-name Credit Default Swaps: A Review of the

(14)

14

3.1. Summary statistics

The initial sample consists of 24,645 of non-event observations, 257 upgrades and 207 downgrades covering the period from January 2010 to March 2015. The number of 207 downgrades includes 23 rating changes to non-investment grade, which are supposed to have a stronger effect on the outstanding amount of CDS net notionals. Adequately 13 of the upgrades are the ones when the firm becomes investment grade.

Below tables present summary statistics for average yearly levels of CDS net notionals comparing non-event observations with a group of downgraded and upgraded companies. For each group the mean, standard deviation and median are presented. The differences in the level of CDS net notionals are mostly visible even when using means at the yearly level. In case of credit rating downgrades, the amount of CDS net notionals is higher and statistically significant in most cases. The descriptive statistics are along with intuition presenting increased demand for CDS contracts in terms of deterioration of the prospects being the effect of the negative credit rating announcements. When upgrades, the results are mixed however the only significant value shows the downward trend in case of rating improvement. The difference between the means of the two groups is tested with a T-test.

[Table 1 and 2 here]

The ratings in the sample are mostly concentrated around BBB (including positive “+” and negative “-“ outlooks). Ratings below “BBB-“ are considered as non-investment grade. The distinction between these groups may be essential on the later stage of empirical evidence, primarily due to the fact that many institutional investors underlie to the policy of limiting their investments only to the investment grade ones.

(15)

15

Table 4 summarizes statistics for the control variables used in the regressions. It is clear that higher rating groups are associated with higher Altman Z-score (lower probability of default), higher dividends and profitability ratios and lower leverage ratio. The boundary groups (AA and above, B, CCC and below) are characterized by the increased stock trading volume, while the average par value volume of bonds being traded in a given month is visibly larger within the investment grade groups. In line with the intuition, bond yield in the safer rating groups is much lower, ranging in means from 2.36% in terms of AA and above group to 10.16% for CCC and below group. As the thesis concerns the factors affecting changes in the levels of CDS net notionals, the monthly differences of the above variables have been presented to check their impact on the dependent variable. When looking for the changes in leverage ratio, it is clear that BB group is characterized by the largest standard deviation of this measure. The changes in profitability measures, changes in trading volumes on stock and bond markets as well as bond yields do not show any trend when dividing the sample into rating subgroups.

[Table 4 here]

4. Empirical evidence

The empirical evidence section is structured as follows. I start the empirical evidence section with the cross-sectional regression to explain the changes in CDS net notionals with credit rating change dummies including set of variables to control for financial stability of the firm. The list of the control variables along with data sources and definitions has been presented in the appendix. Since to my knowledge this paper is the first one to assess the effect of the credit rating changes on the amounts of single entity CDS contracts outstanding, I pick up several control variables that on the basis of the relevant literature positions, I assume to be valuable proxies of the financial stability of the firms included in the sample. The set of the control variables helps to mitigate the

(16)

16

omitted variable bias, what has a noticeable impact on the increased R-squared value. As the next step I use the event study to examine the strength of the impact in the event window around credit rating announcement.

4.1. Regression methodology

The explained variable is the change in CDS net notional amount calculated as a percentage change between the subsequent months in the sample. Following the existing literature, I implement firm-specific explanatory variables. The first one is Altman Z-score, which acts as a well acknowledged benchmark measuring probability of default. The ratio is considered to amount below 1.81 if the company is in distress and the likelihood of the collapse is increased (the firm is expected to bankrupt in two years). Alternatively, if the value of the ratio exceeds 2.99, the company can be regarded as healthy in terms of its financial stability. The expected effect of Altman Z-score on CDS net notional changes is negative, since higher score of the benchmark requires less credit risk insurance. Another aspect that should matter is the firm size, however the direction of the impact is ambiguous. On one hand large firms are associated with higher number of counterparties and therefore of the transactions. On the other hand, smaller firms are more susceptible to market fluctuations and therefore are associated with higher risk. Therefore, I introduce two variables controlling for the firm size: Log (total assets) and Log (tangible assets). The logarithms of the measures were used to take into consideration that the ralationship can be not inherently linear. Following Zhang (2009), I include following additional controls: dividend payout ratio, change in leverage ratio, change in return on equity (ROE) and change in return on sales (ROS). Higher dividend payout ratio is supposed to result in higher demand for CDS contracts. The reasoning behind it is that it entails the reduction in asset value, which makes the company more vulnerable to the prosperity fluctuations. The leverage ratio can be treated as a proxy for the company’s indebtedness and its increased level is associated with larger probability of insolvency. Similarly to Altman Z-score, the higher values of profitability measures reduce the risk of bankruptcy, hence

(17)

17

the predicted effect of these variables is negative. Finally, as according to the literature, stock market, bond market and credit risk market are interconnected, I add monthly stock trading volume as well as bond trading volume to reflect that relationship. Stok trading volume usually changes when the new information about company is released. Credit risk market should follow a similar pattern. Last control variable that I add to my regression is bond yield. According to Hull (2004) the relation between CDS spreads and bond yields should be approximately as follows:

! = # − %

, where s denotes CDS spread, y is yield on par yield bond issued by reference entity, and r means yield on par yield riskless bond. If the investor believes that the spread does not reflect the true value indicated by the entity’s bond yield, he can speculate on the quality of the credit. As the author presents in his paper, if the spread is greater than the right side of the equation, the investors can profit from buying treasury bond, shorting corporate bond and selling CDS contract. On the other hand, if the spread is lower, they will be more prone to buy a corporate bond, buy CDS contract and shortsell the treasury bond. The swap spread reflects the demand for the instrument.

The set of the dummy variables has been included to spot the effect of positive and negative rating changes. As previous studies considering CDS spreads have shown, while the impact of upgrade announcements is not clear, downgrades should significantly influence the spreads of these contracts and I expect the relation regarding CDS net notionals to have a similar effect.

Additionally, I introduce two dummies indicating if the rating of a given company has been decreased to non-investment grade or increased to the investment grade. By doing this, I want to separate the groups, which are expected to have considerably different characteristics because of the following premises. First of all, the regulatory environment as well as companies inner policies can be the source of the opposite effect. The reason behind this is that many financial institutions are limited in terms of investing in junk grade instruments. According to International Organization Of Securities Commissions (The Credit Default Swap Market Report; 2012), the fraction of the

(18)

18

transactions between dealers and non-financial institutions is marginal, what proves that the financial institutions are the ones that the OTC market is mostly dependent on. Therefore, such a change should significantly reduce the demand for the contacts. Secondly, as the CDS becomes more and more risky the spreads are getting higher, which can theoretically lead to the situation when the market participants are no longer interested in such an insurance, because it gets too expensive. These arguments make up a solid basis to make an assumption that these separated groups of upgrades and downgrades can have completely different effect on the changes in CDS net notionals. Apart from that, to control for the downward trend in average level of CDS net notional over the sample period (described in Data section), the year dummy is introduced. The companies are also grouped by their field of activity using primary SIC numbers to control for different characteristics. The last binary explanatory variable is a rating dummy. Sample has been divided here into subsets of firms rated at: AA and above, A, BBB, BB, B, CCC and below. The purpose of this operation is to take into account the fact that depending on the rating quality, the changes in CDS net notionals will be supposedly diverse. In all of the cases the standard errors are clustered within the firm to avoid autocorrelation.

Finally, taking into account the above mentioned, the general equation used in the regression model looks as follows:

Change in net notional CDS value =

a

+ &

1 Upgrade dummy +

&

2 Upgrade to investment grade dummy +

&

3 Downgrade +

&

4 Downgrade to non-investment grade dummy +

&

5 Altman Z-score +

&

6 Log (total assets) +

&

7 Log (tangible assets) +

&

8 Dividend payout

+ &

9 ∆Leverage ratio +

&

10 ∆ROE +

&

11 ∆ROS

+ &

12∆Stock trading volume (monthly) +

&

13 ∆Bond trading volume (monthly) +

&

14∆Bond yield (monthly) +

&

15 Time effects dummy +

&

16 Sector dummy +

&

17 Rating dummy

+ (

(19)

19

4.2. Regression results

Firstly I regress CDS net notional changes on the group of key binary variables including upgrade and downgrade dummies as well as their equivalents controlling for the effect around the investment grade threshold. The impact of the latter ones on CDS net notionals changes is expected to be different from the general trend due to the reasons pointed out in the previous chapters of this paper. In the final regression, I find no impact of the positive rating changes on the monthly differences in CDS net notionals, however the effect of negative announcements is noticeable.

Column (1) presents the initial results indicating that downward rating changes are associated with 1.34% increase in CDS net notionals (at 5% significance level). Changes of the credit rating in both ways around the investment grade lead to the reduction accounting for 4.5% (upgrade) and 2.4% (downgrade). The second column presents the same regression but using firm fixed effects instead of industry fixed effects. These results have much less explanatory power, diminishing R-squared from 8.9% to only 1.2%, so that this configuration is no longer used at the further regressions. As the results are puzzling, I add the control variables, starting with Altman Z-score, Log (total assets) and Log (tangible assets). The direction of Altman Z-score is in line with expectations, showing that the financial health of the company has a negative impact on the CDS changes, causing the increase in the supply of the credit risk insurances. The effect of the Log (total assets) is positive indicating that larger firm size causes higher demand for CDS contracts. The reason for that may be the that larger firms are involved in the higher number of the transactions so there is a need to secure the cash flows. The Log (tangible assets) is insignificant. Next I include controls for the probability of default and dividends. The effect of the both variables is insignificant but the increase in R-squared can be observed. As the next step I add to my regression the explanatory variables reflecting the profitability measures: ROE and ROS. Return on equity turns out to be insignificant, but the negative effect of the return on sales variable is consistent with prior predictions. In the last regression in Column(6) I add to the regression several explanatory variables capturing changes in

(20)

20

stock and bond markets trading volume what results in increasing the level of R-squared to 17.5%. Despite the fact that the effect of changes in bond trading volume and yield have highly significant effect on the changes in CDS net notionals, their impact is economically insignificant. In all of the cases, the positive effect of downward changes on CDS net notionals can be observed, accounting for 1.4% increase in the amount of outstanding contracts in the final regression. Last results show that there is no effect caused by the upward rating changes.

[Table 5 here]

4.3. Event study methodology

By analyzing the abnormal changes in the event window around the rating change I expect to find a confirmation of the previous results. Additionally, the below setup will help to state if the changes are higher in various months around the announcements. That is closely connected with the fourth hypothesis that the credit rating changes can be anticipated on the basis of abnormal changes in the amount of CDS net notionals in the preceding months. The methodology is based on mean adjusted model. As proved by Brown and Warner (1980, 1985), the results obtained using this model yield the results being often similar to those of more sophisticated models. To test how the CDS net notionals change around the credit rating announcement I use several different event windows: [-3,-2], [-2,-1], [-1,0], [0,1], [1,2], [2,3]. This configuration is supposed to capture all of the relevant changes. I assume it also to be short enough to be not biased by exogenous factors that can make the obtained results noisy.

Similarly to the cross-sectional regression, the changes in CDS net notionals are calculated as follows, depending on the event window:

(21)

21

The normal (expected) CDS net notional change is estimated from non-event period changes. To avoid the direct effect of assumed relation between CDS notionals and credit rating announcements, for the purpose of the calculation of the expected change I use four consecutive months starting from the seventh month before the rating change ( i.e. “-7”, “-6”, “-5”, “-4” month prior to the event). On the basis of these non-event observations, the normal mean and variance for every event are calculated the following way:

C

n

=

+*

∑C

nt

-

./

=

(+1*)*

∑(C

nt

- C

n

)

2

For each month in the event period, the abnormal changes are calculated (separately for downgraded and upgraded companies) by subtracting event month change from the expected changes calculated priorly using non-event months. The result is standarized by dividing by standard deviation.

34

nt

=(C

nt

- C

n

)/ -

n

As the next thing, standarized abnormal changes are put together in the portfolios, representing average standarized abnormal changes in a given month in the event window.

34

t

=

∑56978

where N is the number of standarised abnormal changes in given month in the event window.

Under the Central Limit Theorem, the portfolio change is normally distributed having mean of “0” and its variance is 1/Nt. The null hypothesis is that portfolio abnormal changes equal zero in

(22)

22

Z=

(568;1:) <

To allow better understanding of the methodology applied, the below image that presents the idea graphically has been provided:

4.4. Event study results

The event study is structured as follows: the sample is divided into rating subsets to control for different effects between particular groups and the average abnormal changes are tested throughout the whole period. The increased amount of CDS net notionals around the event can be mostly found for downgrades, while the results for upgrades are mixed. Nevertheless, its is clear that credit rating announcement contribute to the perturbations in CDS net notionals for both downgrades and upgrades, even if the direction for the latter is not clear.

Table 6 represents the firms divided into rating subgroups. This configuration gives clear evidence that downward rating changes are associated with increased demand for CDS contracts. Surprisingly, the effect is mostly driven by highest quality ratings represented by AA and above subgroup in my sample as well as low quality non-investment grade group B. Furthermore, the incremental differences in the highest quality rating group can be observed even three months

(23)

23

before and after negative rating change and the values are both statistically and economically significant. Such a strong effect can be interpreted as a downgrade is expected in the group of top quality ratings. As the spreads are not high for instruments generally acknowledged as safe, the swaps are bought to balance off the increased risk portion associated with negative announcement. The spreads are very low at that point, so that the purchase of the insurance comes at slight cost. Therefore, high upward movement can be noticed accounting for 12.69% increase in CDS net notionals in the event window [-3,3]. This pattern can be clearly observed also in B group, representing the lowest quality rating subset in the sample, where most of the upward movement accounting to 7.17% can be observed in the pre-event months [-3,0]. The explanation behind the high rise here is that when company is already graded as junk, any subsequent rating deterioration comes with greater possibility of default. As a consequence, market participants rush to purchase the insurance as the execution of the loan in case of credit event is uncertain. Nevertheless, although the average abnormal changes in the whole event window [-3,3] are equal to 12.65%, they have no statistical significance in the period directly surrounding the event [-1,2]. According to Berg and Streitz (2016), the negative announcements result in growth in the turnover of sovereign CDS contracts. If the effect in case of single entity swaps is alike, the changes are possibly noised by some market turmoil.

Interestingly, in case of BBB rating, the average abnormal changes are clearly flattened comparing to the other groups and changes have no economic significance. The direction during the pre-announcement months is opposite comparing to prior results and the rise in post pre-announcement months is low, what is in line with the outcome of the regression suggesting drop in the demand for CDS contracts around the investment grade threshold. Hence as priorly said, all the downgrades from non-investment grade are placed in BBB group, reduced demand can be presumably interpreted as the impact of institutional investors taking the long position to neutralize the insurance that they have priorly bought, now being forced by the regulatory issues to sell the underlying instrument because it became junk grade.

(24)

24

The positive and statistically significant average abnormal changes can be found for BB group, where they are also concentrated in the pre-event months. The effect for lowest quality ratings, represented in my sample by CCC and below subset is omitted due to the insufficient number of the observations. In any case, apart from BBB group, the upcoming downward change can be anticipated by measuring average abnormal changes in the pre-event months. The forecast can be made with highest probability for either highest or lowest quality rankings.

Similarly to previous results, the abnormal changes around the event being a result of positive rating announcements do not show any clear trend as the directions are contrary. The negative effect of upgrades after credit rating announcement can be spotted in A and BB groups and the positive is visible when looking at BBB and the B groups, but their economic significance is much lower comparing to downgrade changes. As the results are inconsistent it makes it impossible to infer causality. The results for AA and above group as well as CCC and below were omitted due to the insufficient number of upgrades.

[Table 6 here]

5. Robustness checks

The possible issues affecting the robustness of my results are concentrated around 3 areas. First of all, the research period overlaps the years in which the aftermath of the crisis can still be visible. The unusual conditions being the result of the increased uncertainty and the willingness to secure the operations have probably still an effect on demand for CDS contracts in 2010, 2011 and 2012. In the Column(1) the sample was restricted to the years starting from 2013 to reduce the bias. Secondly, some of the companies in the sample underwent multiple rating changes over the research period. In such cases, especially when downgraded or upgraded several times in the specific year, I expect the average abnormal changes in CDS net notionals to be spread over the

(25)

25

time. Therefore, as the frequent rating changes make the investors more prone to insure against the credit risk, the overall effect can be flattened. Column(2) presents the results after excluding such observations from the sample.

The last issue regards the investors’ expectations and the anticipation of the credit rating changes. As the credit rating can be seen as a financial stability indicator of the firm, which is often defined as the ability to generate profit, I add additional control variable to take into account the disagreement in forecasted EPS. The measure is expressed as a standard deviation of the analysts’ forecasts of EPS standarized by the actual net profit (loss). Disagreement about the financial prospects is closely related to the probability of default. Then the intuition here is that when there is much uncertainty around the financial results of the company, traders will be more prone to secure their operations. The sample in the last two columns was restricted to high disagreement firms (Column 3) and low disagreement ones (Column 4). Although the abnormal changes in case of downgrade announcements remain significant in both groups, the effect of uncertainty around the financial results is not clear. High disagreement group is characterized by 3.5% increase in CDS swaps at 5% significance level, but the changes for low disagreement group are even higher accounting too 4% rise at 10% significance level. Because of the drop in the significance level the outcome is not sufficient to state the causality in the effect of the disagreement in analysts’ recommendations. However, the final outcome proves that there is a significant increase in CDS net notionals being the effect of a downward rating changes. The effect visible in every configuration is even stronger comparing to the results of the main regression on the whole sample, ranging from 3.4% to 4.7% rise in CDS net notionals.

(26)

26

6. Conclusions

This thesis is to evaluate the impact of credit rating announcements on the CDS net notionals outstanding. The research has been made using 410 unique US companies in the period from January 2010 to March 2015. Firstly, the sample is tested using cross-sectional regression using monthly changes in CDS net notionals as dependent variable, set of dummies indicating credit rating changes and several control variables helping to reduce omitted variable bias by capturing firm specific characteristics such as leverage ratio, profitability measures, size and the volume of underlying instruments traded on the different markets (stock and bond). The results clearly prove that downgrades lead to 1.4% increase in CDS net notional amount outstanding in the event month and the effect is significant in all the regression configurations. The opposite impact can be seen if the firm becomes junk grade and no significant effect has been seen in case of upgrades.

Secondly, the relation is evidenced using the event study. The experiment based solely on measuring the average abnormal changes in the event window around the rating change shows statistically significant increase in CDS net notional changes that is concentrated mainly in pre-event months. The impact is most pronounced in case of boundary ratings, grouped in the table in AA and above and B subsets. For both of the groups, the average abnormal changes in [-3,0] period account for over 7% increase in CDS net notionals enabling the investors the anticipation of the announcements. That gives a valid confirmation of the previous work, documenting the speculative characteristics of the credit risk market (Acharya and Johnson, 2007). The results indicate that the market participants potentially possess informational advantage, allowing them to clearly anticipate the upcoming rating changes and therefore profit from difference in spreads. The effect for BBB group containing changes to non-investment grade is largely limited.

This paper proves the relation between the credit rating announcements and the number of CDS net positions outstanding and contributes to better understanding of the mechanisms affecting the credit risk market. Similarly to previous studies, I find the evidence of negative changes, having significant impact on the measures of derivative market. This study is also the first one, that

(27)

27

documented the relation between credit rating announcements and changes in single entity corporate CDS contracts making it a solid background for a further investigation on the specificity of the derivatives market.

(28)

28

7. References

1. Acharya, V., and T. Johnson. “Insider trading in credit derivatives.” Journal of Financial Economics, 84 (2007), 110-141.

2. Berg, Tobias and Streitz, Daniel, “Determinants of the Size of the Sovereign Credit Default Swap Market”, The Journal of Fixed Income, Winter 2016, Vol. 25, No. 3: pp. 58-73 3. Bessembinder, H., Kahle, K., Maxwell, W., Xu, D., 2009. Measuring abnormal bond

performance. Review of Financial Studies 22 (10), 4219–4258.

4. Blanco, F., Brennan, S., and Marsh, I.W., 2005, “An Empirical Analysis of the Dynamic Relationship between Investment Grade Bonds and Credit Default Swaps”, Journal of Finance, 60, 2255-2281.

5. Brown, S.J., Warner, J.B., 1985. Using daily stock returns: The case of event studies. Journal of Financial Economics 14 (1), 3–31.

6. Culp, van der Merwe, Staerkle, “Single-name Credit Default Swaps: A Review of the Empirical Academic Literature”; ISDA; 2016

7. Daniels, Kenneth N. And Shin Jensen, Malene, The Effect of Credit Ratings on Credit Default Swap Spreads and Credit Spreads. Journal of Fixed Income, December 2005. 8. D’Errico, M., Battiston, S., Peltonen, T., and Scheicher, M. (2016). How does risk flow in

the Credit Default Swap market? European Systemic Risk Board Working Paper Series No 33.

9. D'Errico, Marco and Roukny, Tarik, Compressing Over-the-Counter Markets (May 3, 2017). European Systemic Risk Board Working Paper Series No 44.

10. Finnerty, John D. ; Miller, Cameron D. ; Chen, Ren-Raw,” The impact of credit rating announcements on credit default swap spreads” Journal of Banking & Finance, 2013, vol. 37, issue 6, pages 2011-2030

11. Forte, S., and Peña, J.I., 2009, “Credit Spreads: An Empirical Analysis on the Informational Content of Stocks, Bonds and CDS”, Journal of Banking and Finance, 33, 2013-2025. 12. Forte, Santiago and Lovreta, Lidija, Credit Risk Discovery in the Stock and CDS Markets:

Who Leads, When, and Why (December 17, 2009).

13. Hull, J., Predescu, M., White, A., 2004. The relationship between credit default swap spreads, bond yields, and credit rating announcements. Journal of Banking and Finance 28 (11), 2789– 2811.

14. Kiff J., Elliott J., Kazarian E., Scarlota J., Spackman C. (2009), Credit derivatives: Systemic Risks and Policy Options, IMF Working Paper

(29)

29

of corporate finance ; Vol. 1. - Amsterdam [u.a.] : Elsevier North-Holland, ISBN 978-0-444-50898-0. - 2007, p. 3-36

16. Longstaff, F.A., Mithal, S. and Neis, E., 2003, “The Credit-Default Swap Market: Is Credit Protection Priced Correctly?”, Working Paper.

17. Longstaff, F.A., Mithal, S. and Neis, E., 2005, “Corporate Yield Spreads: Default Risk or Liquidity? New Evidence from the Credit-Default Swap Market”, Journal of Finance, 60, 2213-2253.

18. Micu, Marianand Remolona, Eli M. and Wooldridge, PhilipD. The Price Impact of Rating Announcements: Which Announcements Matter? (June 2006). BIS Working Paper No. 207.

19. Norden, L. and Weber, M., 2004, “Informational efficiency of credit default swap and stock markets: The impact of credit rating announcements”, Journal of Banking & Finance, 28, 2813-2943.

20. Norden, L., and Weber, M., 2009, “The Co-movement of Credit Default Swap, Bond and Stock Markets: An Empirical Analysis”, European Financial Management, 15, 529-562. 21. Oehmke, Martin and Zawadowski, Adam, The Anatomy of the CDS Market (April 25,

2016). Review of Financial Studies

22. Shivakumar, L., Urcan, O., Vasvari, F.P., Zhang, L., 2011. The debt market relevance of management earnings forecasts: Evidence from before and during the credit crisis. Review of Accounting Studies 16 (3), 464–486.

23. Steiner, M and V Heinke (2001): “Event study concerning international bond price effects of credit rating actions”, International Journal of Finance and Economics, vol 6, pp 139– 57.

24. Stulz, René M., “Credit Default Swaps and the Credit Crisis”, Journal of Economic Perspectives—Volume 24, Number 1—Winter 2010—Pages 73–92

25. Zhu, H., 2006, “An Empirical Comparison of Credit Spreads between the Bond Market and the Credit Default Swap Market”, Journal of Financial Services Research, 29, 211-235.

(30)

30

Figure 1: Companies by sector (primary SIC)

2 9 45 167 34 3 28 30 84 8 0 20 40 60 80 100 120 140 160 180 Agriculture, Forestry

and Fishing Construction Insurance and Real Estate Manufacturing Mining Nonclassifiable Retail Trade Services Communications, Transportation, Electric, Gas and Sanitary service

(31)

31 ) 400000000 800000000 1200000000 A vg. CDS net not ional 2010m1 2011m1 2012m1 2013m1 2014m1 2015m1 Time(by month)

Figure 2: Average firm level of CDS net notionals over the period from January 2010 to March 2015 (USD EQ)

(32)

32 Table N N Year number of non event observation s

Mean St. Dev Median

number of downward

credit changes

Mean St. Dev Median difference (%) 2010 4801 1.01E+09 6.87E+08 8.51E+08 39 1.36E+09 9.82E+08 1.11E+09 34.88%***

2011 4761 9.64E+08 6.83E+08 7.95E+08 43 1.08E+09 7.04E+08 1.03E+09 11.66%

2012 4721 8.83E+08 6.48E+08 7.44E+08 54 1.08E+09 7.26E+08 8.12E+08 21.85%**

2013 4645 7.69E+08 6.12E+08 6.55E+08 32 1.05E+09 1.18E+09 7.89E+08 36.54%**

2014 4586 6.52E+08 5.69E+08 5.37E+08 25 5.62E+08 5.98E+08 3.89E+08 -13.74%

2015 1131 5.66E+08 5.08E+08 4.72E+08 14 6.55E+08 4.71E+08 6.26E+08 15.65%

Total 24645 207

Average CDS net notionals (firm level)

Downgrade Non event Table N N Year number of non event observation s

Mean St. Dev Median

number of upward

credit changes

Mean St. Dev Median difference (%) 2010 4801 1.01E+09 6.87E+08 8.51E+08 64 7.84E+08 5.8E+08 5.68E+08 -22.26%*** 2011 4761 9.64E+08 6.83E+08 7.95E+08 55 9.45E+08 5.87E+08 8.53E+08 -1.96%

2012 4721 8.83E+08 6.48E+08 7.44E+08 31 8.98E+08 6.14E+08 7.02E+08 1.78%

2013 4645 7.69E+08 6.12E+08 6.55E+08 58 6.94E+08 4E+08 6.36E+08 -9.74%

2014 4586 6.52E+08 5.69E+08 5.37E+08 42 6.73E+08 4.03E+08 6.07E+08 3.26%

2015 1131 5.66E+08 5.08E+08 4.72E+08 7 4.39E+08 2.83E+08 3.77E+08 -22.49%

Total 24645 257

Average CDS net notionals (firm level)

Upgrade Non event

Table 1: Average levels of CDS net notionals

This table compares the average level of CDS net notionals in the companies that underlaid a downgrade rating change to non-event firms. The data is presented by year and covers the period from January 2010 to March 2015.

Table 2: Average levels of CDS net notionals

This table compares the average level of CDS net notionals in the companies that underlaid a upgrade rating change to non-event firms. The data is presented by year and covers the period from January 2010 to March 2015.

(33)

33

% of the sample Rating N % of the

sample

Number of upgrades

Number of downgrades

AAA 64 0.3% 0 0 investment grade

AA+ 39 0.2% 0 1 investment grade

AA 571 2.3% 3 1 investment grade

AA- 723 2.9% 6 3 investment grade

A+ 1465 5.8% 11 4 investment grade A 2453 9.8% 19 10 investment grade A- 3097 12.3% 26 15 investment grade BBB+ 3933 15.7% 32 23 investment grade BBB 4457 17.8% 36 29 investment grade BBB- 2819 11.2% 11 29 investment grade

BB+ 1046 4.2% 23 18 NON investment grade

BB 1048 4.2% 24 16 NON investment grade

BB- 1271 5.1% 22 19 NON investment grade

B+ 889 3.5% 22 13 NON investment grade

B 601 2.4% 9 9 NON investment grade

B- 432 1.7% 6 7 NON investment grade

CCC+ 147 0.6% 6 4 NON investment grade

CCC 38 0.2% 0 4 NON investment grade

CCC- 14 0.1% 1 1 NON investment grade

CC 2 0.0% 0 1 NON investment grade

0.0% C 0 0.0% 0 0 NON investment grade

D 0 0.0% 0 0 NON investment grade

Total 25109 100.0% 257 207

Table 3: Rating frequency matched with monthly averages of CDS net notionals in the sample period from

January 1, 2010 to March 31 2015

Table 1: Rating frequency matched with monthly averages

of CDS net notionals in the sample period from January 1, 2010 to March 31 2015

(34)

34

(1) (2) (3) (4) (5) (6)

N mean std p10 p50 p90

Altman Z-score 14,660 2.670 1.528 0.94 2.53 7.01

Log (total assets) 33,906 10.086 1.288 8.47 10.00 11.81 Log (tangible assets) 30,004 -1.619 1.255 -3.21 -1.32 -0.43

Dividend payout 16,528 2.32% 2.43% 0.00% 2.01% 4.50%

∆ Leverage ratio 28,254 0.33% 13.59% -1.94% 0.00% 1.29%

∆ ROE 29,040 -18.24% 2754% -3.84% 0.00% 1.85%

∆ ROS 33,186 1.13% 136% 0.00% 0.00% 0.00%

∆ Stock trading volume (monthly) 25,971 56.36% 3576% -32.79% -1.38% 52.28% ∆ Bond trading volume (monthly) 17,830 24.64% 402% -40.35% -1.18% 68.29% ∆ Bond yield (monthly) 17,460 3.19% 205% -11.81% -1.05% 13.26%

(1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6)

N mean std p10 p50 p90 N mean std p10 p50 p90 N mean std p10 p50 p90

Altman Z-score 769 3.994 1.247 2.63 3.76 6.44 3,777 3.515 1.644 1.51 3.39 5.71 6,528 2.480 1.232 0.98 2.37 4.13

Log (total assets) 1,928 11.856 0.989 10.45 11.75 13.09 8,900 10.915 1.128 9.60 10.72 12.50 15,327 9.907 0.988 8.68 9.89 11.09

Log (tangible assets) 1,736 -1.800 1.070 -3.46 -1.56 -0.62 7,453 -1.808 1.296 -3.26 -1.55 -0.46 13,500 -1.398 1.076 -2.97 -1.09 -0.36

Dividend payout 805 2.89% 0.98% 2.05% 2.90% 4.05% 4,019 2.41% 1.47% 0.98% 2.13% 4.11% 7,380 2.66% 2.41% 0.33% 2.38% 5.00%

∆ Leverage ratio 1,631 0.23% 4.10% -2.26% 0.00% 2.22% 7,000 0.42% 7.21% -2.04% 0.00% 1.49% 12,721 0.32% 9.75% -1.82% 0.00% 1.41%

∆ ROE 1,556 0.32% 12.97% -2.92% 0.00% 0.80% 7,381 1.06% 91.7% -3.70% 0.00% 1.43% 12,909 -39.36% 4110.9% -3.28% 0.00% 1.57%

∆ ROS 1,891 -0.28% 8.91% 0.00% 0.00% 0.00% 8,735 1.32% 133.3% 0.00% 0.00% 0.00% 15,012 -0.20% 67.9% 0.00% 0.00% 0.00%

∆ Stock trading volume (monthly) 1,715 5.97% 41.22% -31.38% -1.74% 46.32% 6,818 174.01% 6954.5% -30.54% -1.55% 46.42% 11,562 18.56% 441.8% -32.81% -1.11% 54.10% ∆ Bond trading volume (monthly) 1,010 5.45% 36.19% -29.28% 0.52% 42.02% 5,391 24.00% 370.3% -37.95% -1.64% 60.04% 7,446 24.83% 201.7% -43.17% -1.63% 79.61% ∆ Bond yield (monthly) 980 3.85% 86.08% -11.84% -1.23% 13.45% 5,226 7.75% 370.7% -12.20% -1.19% 12.38% 7,369 1.17% 22.4% -11.48% -0.90% 13.82%

(1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6)

N mean std p10 p50 p90 N mean std p10 p50 p90 N mean std p10 p50 p90

Altman Z-score 2,460 2.008 1.351 0.73 1.76 3.70 1,058 1.492 1.183 -0.22 1.41 3.08 68 1.429 0.515 0.76 1.61 2.16

Log (total assets) 4,789 9.140 0.984 7.76 9.32 10.23 2,686 8.950 0.889 7.82 8.91 10.31 276 8.449 0.881 7.38 8.68 9.49

Log (tangible assets) 4,498 -1.430 1.078 -3.03 -1.06 0.47 2,541 -2.225 1.726 -5.28 -1.52 -0.59 276 -3.623 2.060 -6.37 -3.37 -0.79

Dividend payout 2,761 1.96% 3.26% 0.00% 1.05% 5.11% 1,424 0.82% 2.69% 0.00% 0.00% 2.25% 139 0.43% 1.64% 0.00% 0.00% 0.43%

∆ Leverage ratio 4,401 0.50% 28.45% -2.45% 0.00% 0.93% 2,321 -0.20% 4.64% -1.54% 0.00% 0.55% 180 1.33% 4.81% 0.00% 0.00% 4.83%

∆ ROE 4,408 -8.19% 587.11% -4.71% 0.00% 2.75% 2,538 2.56% 428.95% -8.31% 0.00% 4.31% 248 -0.68% 53.20% -19.22% 0.00% 12.51%

∆ ROS 4,652 0.21% 102.29% 0.00% 0.00% 0.00% 2,635 10.80% 360.20% 0.00% 0.00% 0.00% 261 0.85% 67.01% 0.00% 0.00% 0.00%

∆ Stock trading volume (monthly) 3669 8.98% 58.79% -34.68% -1.75% 55.14% 1948 8.42% 49.40% -37.30% -0.76% 61.44% 259 12.17% 54.84% -39.85% 0.74% 81.38% ∆ Bond trading volume (monthly) 2,445 36.45% 857.31% -40.42% -0.47% 71.35% 1,379 18.46% 153.97% -38.85% -0.28% 60.29% 159 30.82% 174.98% -45.12% -0.08% 87.91% ∆ Bond yield (monthly) 2374 1.02% 20.93% -11.50% -0.97% 13.52% 1352 0.30% 16.32% -11.95% -1.47% 12.27% 159 0.30% 17.63% -17.52% -1.33% 20.15%

VARIABLES

A BBB

AA and above VARIABLES

Table 4: Regression summary statistics

BB B CCC and below

Whole sample VARIABLES

This table provides summary statistics for data: number of observations, mean, standard deviation, 10th, 50th and 90th percentile. The description of the variables and data sources can be found in the separate table in the appendix. The data is presented by year and covers the period from January 2010 to March 2015.

(35)

35

Table 5: Changes in net notional CDS amounts

(1) (2) (3) (4) (5) (6)

net CDS change net CDS change net CDS change net CDS change net CDS change net CDS change

Upgrade (dummy) -0.00128 -0.00103 0.00321 0.00326 0.00298 0.00185

(-0.35) (-0.28) (0.70) (0.67) (0.60) (0.29)

Upgrade to investment grade (dummy) -0.0454** -0.0485** -0.0533** -0.0487* -0.0482* -0.0141

(-2.02) (-2.12) (-2.29) (-1.88) (-1.85) (-1.00)

Downgrade (dummy) 0.0134** 0.0123** 0.0196** 0.0222** 0.0226** 0.0141**

(2.28) (2.07) (2.00) (2.23) (2.13) (2.10)

Downgrade to non investment grade (dummy) -0.0241* -0.0220* -0.0378** -0.0400** -0.0407** -0.0310*

(-1.92) (-1.74) (-2.04) (-2.18) (-2.17) (-1.67)

Altman Z-score -0.000786* -0.000811* -0.000858* -0.00104**

(-1.76) (-1.67) (-1.65) (-2.08)

Log (total assets) 0.00196*** 0.00203*** 0.00212*** 0.00174**

(3.06) -3.07 (3.14) (2.35)

Log (tangible assets) 0.000256 0.000201 0.0000733 -0.0000311

(0.29) (0.23) (0.08) (-0.04) Dividend payout -0.0246 -0.0277 -0.0127 (-1.13) (-0.99) (-0.39) ∆ Leverage ratio -0.00710 -0.00565 -0.00831 (-1.39) (-1.21) (-0.89) ∆ ROE 0.00000450 0.0000119 (0.69) (0.83) ∆ ROS -0.00176** -0.00361 (-2.35) (-1.39)

∆ Stock trading volume (monthly) 0.0000402

(0.82)

∆ Bond trading volume (monthly) -0.000157***

(-3.89)

∆ Bond yield (monthly) 0.000514***

(11.51)

Rating controls Yes Yes Yes Yes Yes Yes

Time fixed effects Yes Yes Yes Yes Yes Yes

Industry fixed effects (primary SIC) Yes No Yes Yes Yes Yes

Firm fixed effects No Yes No No No No

Number of Firms 410 410 251 250 247 222

Number of Observations 34,585 34,585 16,575 16,067 15,389 11,990

R-squared 8.9% 1.2% 11.50% 15.5% 16.4% 17.5%

This table presents the results of a regression of changes in net CDS positions on proxies for financial stability, rating change dummies and controls. All regressions include time and rating dummies. The rating controls are dummies for: AA and above, A, BBB, BB,B, CCC and below is ommited. Column (2) includes firm fixed effects, all other columns include industry dummies based on primary SIC codes. The sample period is January 2010 to March 2015. T-stats are given in parentheses based on standard errors clustered by firm. ***, **, and * denote significance at the 1%, 5% and 10%

(36)

36

AAH N AAH t-Stat AAH N AAH t-Stat

[-3,-2] 8 4.87% 9.08*** [-3,-2] 3 7.49% [-2,-1] 8 2.09% 3.25*** [-2,-1] 3 [-1,0] 8 0.40% 1.93* [-1,0] 3 [0,1] 8 2.89% 4.38*** [0,1] 3 4.85% [1,2] 7 1.53% 2.58*** [1,2] 3 [2,3] 7 0.36% 3.5*** [2,3] 3 [-3,-2] 41 -1.46% -1.58 [-3,-2] 34 -1.0% -1.94* 0.56% [-2,-1] 41 0.003% 1.94* [-3,0] -2.30% [-2,-1] 34 -0.7% -0.76 [-1,0] 41 2.04% 3.65*** [-1,0] 34 -0.7% -5.7*** [0,1] 40 -0.47% -0.75 [0,1] 33 -2.7% -8.05*** 1.51% [1,2] 39 2.13% 1.52 [3,0] -4.83% [1,2] 33 0.0% 1.07 [2,3] 38 -0.14% 1.11 [2,3] 33 -2.2% -5.25*** [-3,-2] 59 -0.16% -0.08 [-3,-2] 87 1.1% 8.6*** -0.12% [-2,-1] 59 -0.05% -1.64 [-3,0] 3.41% [-2,-1] 87 1.2% 3.05*** [-1,0] 59 0.09% -0.25 [-1,0] 88 1.1% 3.29*** [0,1] 56 0.19% 0.41 [0,1] 87 0.9% 3.82*** 0.79% [1,2] 50 -0.18% 0.89 [3,0] 3.81% [1,2] 87 1.1% 3.74*** [2,3] 49 0.79% 3.73*** [2,3] 85 1.8% 11.21*** [-3,-2] 29 0.34% 1.71* [-3,-2] 43 0.0% -4.47*** 3.43% [-2,-1] 29 1.76% 3.05*** [-3,0] -1.25% [-2,-1] 43 -0.6% -0.81 [-1,0] 29 1.30% 3.97*** [-1,0] 42 -0.6% -6.29*** [0,1] 28 0.002% 1.77* [0,1] 40 -1.5% -2.03** 0.76% [1,2] 23 1.39% -6.01*** [3,0] -1.56% [1,2] 39 -0.3% -10.64*** [2,3] 23 -0.62% -0.63 [2,3] 39 0.2% 3.26*** [-3,-2] 7 2.49% 1.75* [-3,-2] 38 -0.5% 0.65 7.17% [-2,-1] 7 2.49% 2.3** [-3,0] 0.30% [-2,-1] 38 0.1% -2.98*** [-1,0] 6 2.03% -0.85 [-1,0] 38 0.8% -10.16*** [0,1] 7 0.24% 0.03 [0,1] 37 0.6% 9.98*** 5.11% [1,2] 6 0.89% 0.59 [3,0] 0.33% [1,2] 37 0.0% 16.73*** [2,3] 5 3.93% 2.41** [2,3] 37 -0.3% 7*** [-3,-2] 4 [-3,-2] 4 [-2,-1] 4 [-2,-1] 4 [-1,0] 4 [-1,0] 4 [0,1] 4 [0,1] 4 [1,2] 3 [1,2] 4 [2,3] 3 [2,3] 4 [3,0] [-3,0] [3,0] B Upgrade Downgrade [3,0] CCC and below

Table 6: Changes in net notional CDS amounts

This table presents the abnormal changes in CDS net notionals in the event window around the rating change. T-stats used to test the significance of the abnormal returns are given in parentheses. ***, **, and * denote significance at the 1%, 5% and 10% level, respectively.

AA and above

A

BBB

BB

omitted due to the insufficient number of the observations omitted due to the

insufficient number of the observations

omitted due to the insufficient number of the observations [-3,0] [3,0] [-3,0] [3,0] [-3,0] [-3,0]

(37)

37

Table 7: Changes in CDS net notional amounts

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

net CDS change net CDS change net CDS change net CDS change

Upgrade (dummy) -0.00515 -0.0105 -0.00499 -0.0361

(-0.50) (-0.91) (-0.36) (-1.29)

Upgrade to investment grade (dummy) -0.0174 -0.00947 (omitted) (omitted)

(-1.09) (-0.59) (.) (.)

Downgrade (dummy) 0.0471*** 0.0336** 0.0351** 0.0402*

(3.52) (3.20) (2.17) (1.66)

Downgrade to non investment grade (dummy) -0.0366 0.00739 -0.0250 (omitted)

(-1.17) (0.34) (-1.29) (.)

Altman Z-score -0.000701 -0.000401 0.00460 0.00198

(-0.85) (-0.48) (1.59) (1.15)

Log (total assets) 0.000221 0.00100 0.00216 -0.00432

(0.16) (0.72) (0.67) (-1.62)

Log (tangible assets) 0.000411 -0.000322 -0.00589* -0.00251

(0.28) (-0.22) (-1.72) (-0.86) Dividend payout 0.0203 0.0136 -0.0245 0.136 (0.31) (0.20) (-0.18) (1.28) ∆ Leverage ratio 0.0102 0.0102 0.0263 0.0129 (1.18) (1.22) (1.37) (0.86) ∆ ROE -0.000226 -0.000223 -0.00785*** -0.000438*** (-1.30) (-1.31) (-6.46) (-10.58) ∆ ROS -0.00308 -0.00523 -0.00611* -0.0236** (-1.16) (-1.11) (-1.87) (-2.15)

∆ Stock trading volume (monthly) -0.00888*** -0.00942*** -0.00475 -0.0113*

(-3.73) (-4.11) (-0.79) (-1.90)

∆ Bond trading volume (monthly) -0.000162*** -0.000167*** 0.00137 -0.0102***

(-4.53) (-4.49) (1.58) (-2.73)

∆ Bond yield (monthly) -0.00105 -0.00324 0.00500 -0.0227

(-0.26) (-0.80) (0.46) (-0.97)

Rating controls Yes Yes Yes Yes

Time fixed effects Yes Yes Yes Yes

Industry fixed effects (primary SIC) Yes Yes Yes Yes

Firm fixed effects No No No No

Number of Firms 211 201 97 123

Number of Observations 4,850 4,630 752 877

R-squared 21.2% 22.27% 10.47% 14.90%

This table presents the results of a regression of changes in net CDS positions on proxies for financial stability, rating change dummies and controls. All regressions include time and rating dummies. The rating controls are dummies for: AA and above, A, BBB, BB,B, CCC and below is ommited. All the columns include industry dummies based on primary SIC codes. The sample period is January 2010 to March 2015. To control for the egzogenous changes, the sample in Column (1) presents the results only in post crisis years (2013-2015). Column (2) was restricted to the firms that underwent only single rating

change during the research period. Third columns presents the regression for the sample being additionally restricted to only companies with high disagreement about earnings forecast. The measure is expressed as the 75th percentile of Forecast disagreement defined as standard deviation of forecasted EPS

standarized by actual net profit (loss). Last column shows the results for the companies with low disagreement expressed as 25th percentile of Forecast

disagreement variable. T-stats are given in parentheses based on standard errors clustered by firm. ***, **, and * denote significance at the 1%, 5% and 10%

Referenties

GERELATEERDE DOCUMENTEN

Based on these findings, guidelines are drawn up for the three main parties involved in the CDS market. The first involved party, the ISDA, is the party that has the most influence

The benefit of the community envisioning has already show as it raised the criterion of social conformism (sense of community); Social participation on the web needs to be

Although most of the research efforts have been performed to analyse the effect of degradation mechanisms, very limited research has been carried out on the countermeasures

Its focus on design and creativity, as well as the diversity of the students, requires an approach in informatics courses different from classical computer science programmes..

With the strategic use of historical data and future projections, this study derived quantitative insights into the relative impacts of human activities and climate change

Taking into account the insight that scientific objects are phenomenotechnical realizations, an epistemology inspired by Bachelard allows to connect (i) the practical actions

This paper considers the reciprocal implementation of three planning approaches, new urbanism, new ruralism and green urbanism, to reflect on multifunctionality and draws

The construction of walls in semimodern and modern buildings, on the other hand, using commercial materials such as glass wool, polystyrene, and cement, results in considerably