Determinants of the Size of the Sovereign. Credit Default Swap Market

Transcription

1 Determinants of the Size of the Sovereign Credit Default Swap Market January 17, 2015 Abstract We analyze the sovereign CDS market for 57 countries, using a novel dataset comprising weekly positions and turnover data. We document that CDS markets measured relative to a country s debt are larger for smaller countries, countries with a rating just above the investment-grade cutoff, and countries with weaker creditor rights. Analyzing changes in credit risk, we find that rating changes matter but only for negative rating events (downgrades and negative outlooks). In particular, weeks with downgrades and negative outlooks are associated with a significantly higher turnover in the sovereign CDS market even after controlling for changes in sovereign CDS spreads. We conclude that agencies ratings are a major determinant of the size of the sovereign credit default swap market. Keywords: Sovereign Credit Default Swaps, Sovereign CDS Trading Volume, DTCC Data, Sovereign Credit Ratings JEL-Classification: F34, G12 1

2 1 Introduction In recent years news about deteriorating creditworthiness of sovereigns have dominated the economic news. The costly stabilization programs that were necessary in the aftermath of the financial crisis and/or homemade problems have led to increasing debt levels. This in turn led to an increasing importance of and sometimes critical public focus on the credit default swap (CDS) market, an over-the-counter derivative market where default protection for sovereign and corporate bonds and loans can be bought. There is an extensive research looking at the impact of fundamental sovereign risk and changes in the sovereign debt rating on CDS spreads (Afonso et al. (2011); Dieckmann and Plank (2012); Hilscher and Nosbusch (2010); Ismailescu and Kazemi (2010); Longstaff et al. (2011); Pan and Singleton (2008)). However, despite this extensive literature on sovereign credit risk, surprisingly little is known about the anatomy of the CDS market in terms of volume and trading activity. This paper studies the determinants of the CDS market size and trading activity. In particular, we analyze how country fundamentals, credit ratings, credit rating changes, and macroeconomic policy affect the CDS market. We use a novel dataset provided by the Depository Trust & Clearing Corporation (DTCC) containing weekly observations for the net notional CDS amounts outstanding and CDS trading volume for 57 countries located all over the world. Our study is explorative in nature, i.e., the aim of our paper is to present for the first time determinants of the sovereign CDS market size and trading volume. Hereby, we lay the groundwork and provide guidance for future research in this area. Given the novelty of the data set and the limited prior literature on CDS volume 2

3 and trading activity, we start by thoroughly describing the data available. The two key measures available from the DTCC net notional CDS amounts outstanding and CDS trading volume allow for interesting inferences on the rationale of CDS trading. The CDS market consists of participants who are net buyers of protection (for example, a pension fund) and market participants who take the other side of the trade and are net sellers of protection (for example, banks). If one of the "pension funds" buys protection with a face value of 10 from one of the banks, this would be reflected in a trading volume of 10 and an increase in the net CDS outstanding of 10. If one of the "pension funds" unwinds its position by entering into an offsetting CDS long position, the trading volume would be reflected in a decrease in the net notionals outstanding. Finally, there can be trades within each group, for example the "banks" reallocating their CDS positions among each other, in which case the trading volume will not affect the net notionals outstanding. By simultaneously observing both the trading volume and the change in the net notional outstanding we are therefore able to draw conclusions about which of the CDS trading types described above is actually taking place in the market. The DTCC has started to disseminate its data on CDS trading volume in July This paper is, to our best knowledge, the first paper to analyze market size and trading volume in the sovereign CDS market. We begin by relating country fundamentals to the CDS market size and trading volume. In the following, we measure CDS market size as the net notional outstanding divided by a countries total debt and trading volume in percent of net notional outstanding. We document that CDS markets are larger for smaller countries, riskier countries, and counties with weaker creditor rights. In particular, countries with a 3

4 credit rating just above the investment-grade cutoff have larger CDS markets. 1 Following Porta et al. (1999), we use a country s legal system as a proxy for creditor rights. Porta et al. (1999) argue that a common law legal system, found mostly in Britain and its former colonies, is associated with superior government. Consistent with CDS market emerging if the demand to hedge sovereign credit risk is high, our results indicate that common law countries have significantly smaller CDS markets. Interestingly, while CDS markets are largest for countries with a credit rating just above the investment grade threshold, CDS trading activity highest in countries with a below investment grade rating. We further analyze whether the CDS market reacts to changes in credit ratings and macroeconomic policy events. We find that the CDS trading activity strongly increases in weeks with downgrades and in weeks with negative rating outlooks. These results are robust to controlling for sovereign CDS spread changes, suggesting that it is not the change in sovereign creditworthiness but indeed the (anticipated) rating action that is driving the results. Further, the market seems to anticipate negative events: CDS trading increases already in the week before a downgrade and in the week before a negative rating outlook. Comparable to studies that look at the CDS spread reactions to credit rating changes (Ismailescu and Kazemi (2010); Afonso et al. (2011)), we do not find a reaction to upgrades or positive rating outlooks. 2 Interestingly, while the trading activity is higher in weeks with rating changes, we do not find that the higher turnover is associated with a change in CDS market size in the respective weeks. This suggests that even negative rating events do not lead to a significant increase in the net 1 This is similar to the corporate CDS market (Oehmke and Zawadowski (2014)). 2 Finnerty et al. (2013) document a positive effect of upgrades on corporate CDS spreads using more extensive samples of credit rating events and CDS spreads than previous studies. However, downgrades have a much greater impact on CDS spreads than upgrades in terms of magnitude. 4

5 demand for CDS protection but rather to a reallocation of net demand or net supply between different market participants. For macroeconomic policy events we find an increase in trading activity in weeks before EU-Summits, however, only for EU-15 Members. We do not find abnormal CDS trading in weeks with LTRO announcements or implementations. We contribute to the literature in several ways. First, we add to the literature on the size and turnover in CDS markets. Prior studies have focussed on the corporate CDS market (Stulz (2010); Shachar (2012); Oehmke and Zawadowski (2014)). To the best of our knowledge, we are the first study that analyzes the determinants of the size and turnover in sovereign CDS markets. In particular, we present stylized facts on the determinants of the sovereign CDS market size and trading activity, showing that both are affected in different ways by country fundamentals and country credit risk. Second, our paper is related to the ongoing discussion about capital market reactions to sovereign credit rating revisions. Existing studies have analyzed the effect of sovereign ratings on bond yields and CDS spreads (Reisen and Von Maltzan (1999); Ismailescu and Kazemi (2010)). Reisen and Von Maltzan (1999) analyze government bond yield spreads and find a significant effect when a country was put on review for a downgrade. Ismailescu and Kazemi (2010) analyze sovereign CDS spreads and find that CDS markets anticipate negative rating events. We contribute to this literature by documenting a significant increase in CDS market trading activity prior to rating events. In particular, both downgrades and negative rating outlooks seem to be anticipated. While rating changes are associated with a higher trading activity the actual credit protection outstanding is not affected. More generally, our paper adds to the literature on the link between CDS and equity markets and the importance of rating changes for these 5

6 markets. While prior literature has analyzed these links for corporates (Norden and Weber (2004); Norden and Weber (2009); Löffler (2004)), we provide corresponding evidence for the size and trading activity in the sovereign CDS market. In particular, we provide evidence that agencies ratings are a major determinant of the market size and trading volume in the sovereign credit default swap market. The remainder of the paper proceeds as follows. Section 2 describes our data. Section 3 presents our results and Section 4 concludes. 2 Data We obtain our CDS data from the Depository Trust & Clearing Corporation (DTCC). The DTCC provides clearing, settlement, and information services for equities, corporate and municipal bonds, government and mortgage-backed securities, money market instruments, and over-the-counter derivatives. The DTCC receives their CDS data directly from dealers who are registered with the DTCC and then enters the data into a Trade Information Warehouse (TIW). According the DTCC the data covers trades from all major global derivatives dealers and more than 1700 buy-side firms and other market participants located in 52 countries making it the most comprehensive data source for information on CDS trades. 3 Prior to the release of the DTCC data surveys published by the Bank for International Settlements (BIS) were the main sources of CDS information. However, the BIS provides only aggregate market information. The main advantage of the DTCC data is that the DTCC reports CDS information on a single reference entity level. Available from the TIW are weekly data on notional 3 According to the DTCC, the TIW captures around 95% of globally traded CDS. 6

7 amounts outstanding and weekly trading volume data. The most detailed data are provided for the top 1000 reference entities. 4 The DTCC has published weekly reports on outstanding notional CDS amounts since October 31, 2008 and started to also report weekly CDS trading volume in July The TIW distinguishes between the gross and the net amount outstanding. The difference between gross and net CDS amounts outstanding is driven by mechanisms that reduce the CDS volume while holding constant the actual risk positions. The two methods used are trade compression (TC) and central counterparty clearing (CCC). The mechanisms can best be illustrated via a simple example. Consider two market participants, A and B, that trade CDS on a given reference entity R. First A buys $10m credit protection from B. The gross notional amount outstanding is now given by $10m. The net notional amount outstanding is equivalent to the gross amount. Now assume that, while the initial contract is still active, B buys $10m credit protection from A. The gross notional amount outstanding is increased to $20m while the net notional amount outstanding is reduced to zero because both contracts cancel each other out in terms of credit protection. The process that reduces CDS contacts between two parties in order to avoid complex contract structures without actual credit protection is called TC. This service is provided by companies such as TriOptima, Markit or Creditex that developed special algorithms. Another mechanism to avoid complex contract structures is called CCC. Here all outstanding contracts are transferred to a central clearing entity that performs netting as described above. For clarification both mechanisms are again illustrated in Figure 1. 4 Top 1000 refers to the 1000 reference entities with the highest gross CDS amounts outstanding in the respective week. 7

8 [Figure 1 here] Besides gross and net notional amounts outstanding we further use data on weekly trading activity. The trading volume provided by the TIW includes only trades with actual risk transfer within the respective week like new contracts, terminations and assignments. The trading volume refers to the sum of the absolute amounts of the individual transactions by reference entity. 5 Transactions that do not constitute a change in the individual risk exposure, i.e. trade compression and central counterparty clearing, are not contained in the data. We restrict our sample to single-name CDS contracts on sovereign bonds that are among the top 1000 reference entities in terms of gross CDS amount outstanding in a given week. We analyze the time period from July 2010 to July 2012 because we only have information on both the CDS trading volume and the notional CDS amounts outstanding since July Our final sample includes weekly CDS information for 57 sovereign bonds. We match the TIW data to a series of other data sources. We use Thomson Reuters Datastream to collect data on CDS spreads and country level equity index returns. We focus on CDS spreads with a maturity of 5 years denominated in USD. We use MSCI country level indices to calculate weekly equity returns. We obtain information on credit rating changes and changes in rating outlooks from Bloomberg. We use the foreign currency long-term debt rating, since the majority of CDS contracts are long term and denominated in foreign currencies. We only consider the ratings by Moody s, Fitch and Standard and Poor s since these rating agencies account for the majority of 5 Since assignments involve both a cancellation of an existing and the establishment of a new contract only the "new lag" is included while the "old lag" is excluded to avoid double-counting. 8

9 the market share and cover all countries in our sample. We further obtain macro-level data from the International Monetary Fund (IWF) World Economic Outlook Database. 3 Results 3.1 Descriptive Statistics Due to the novelty of the dataset, we begin our discussion with an extensive descriptive analysis. Table 1 presents summary statistics for the overall panel. [Table 1 here] Overall, we have data on CDS positions for 57 countries for 106 weeks. The mean CDS Market Size (in mio USD) (net notional CDS amount outstanding) is given by $4,108 million while the median is $2,167 million. The CDS Trading Volume (in mio USD) in an average week is $695 million, the median is $214 million. However, conclusions on the actual market size and trading activity can better be drawn by normalizing the figures by the size of the reference entity. We therefore express CDS Market Size (in mio USD) in percent of the countries total debt and normalize the CDS Trading Volume (in mio USD) by the total net CDS amount outstanding 6 to obtain CDS Turnover (%). 7 The mean CDS Market Size (%) is only 2.2%, the median 1.7%. These figures suggest that the credit protection is small when compared to the size of the overall economy. The mean CDS Turnover (%) is 14%, the median 10%. The majority of the 6 We focus on the net CDS amount outstanding as a proxy for the market size and to normalize the CDS Trading Volume (in mio USD), however,allresultsreportedinthispaperremainqualitatively unaffected if we use the gross amount. 7 Similar to using market capitalization to scale trading activity in the equity market (Johnson (2008)). 9

10 countries in our sample have an investment grade rating (78%). We have 88 (98) weeks with downgrades (upgrades) by S&P, Moody s or Fitch and 26 (4) weeks with negative (positive) rating outlooks. We further present summary statistics for selected variables by country in Table 2. [Table 2 here] The country with the larges CDS Market Size (in mio USD) is Italy, followed by France. Interestingly, the CDS Market Size (in mio USD) of Italy is more than 20% larger than the CDS Market Size (in mio USD) of France. The 5 largest countries alone account for 40% of the worldwide CDS Market Size (in mio USD) covered by the DTCC data. Other large economies like the US and UK have surprisingly small CDS market sizes. For example, Italy has more than 5.5 times as much CDS outstanding as the US. The picture changes when CDS Market Size (in mio USD) is normalized by the size of the economy. The countries with the highest CDS Market Size (%) are Bulgaria, Panama, Latvia, and Iceland. Countries with high CDS volumes (both absolute and relative) also have above median CDS spreads suggesting a positive correlation between the credit risk and outstanding credit protection. Turning to the trading volume we observe a similar picture. The countries with high CDS Market Size (in mio USD) are also the countries with high CDS Trading Volume (in mio USD). The two largest countries in terms of average weekly CDS Trading Volume (in mio USD) are Spain and Italy. When looking at the CDS Turnover (%), Ukraine, Russia, Argentina, Turkey, and Spain are the countries with high volumes. Interestingly, these are rather large economies with below average debt levels but mostly above average CDS spreads. Table 2 also shows that the majority of the countries in 10

11 our sample are located in Europe. Not only do European countries account for the lion s share of total CDS Market Size (in mio USD) and CDS Trading Volume (in mio USD), the European countries also have the highest net credit protection in percent of the total debt. Taken together, the descriptive findings suggest that as expected there seems to be a link between credit quality and the CDS market size and CDS trading activity. The following section further explores the cross country differences in CDS Turnover (%) and CDS Market Size (%). 3.2 Cross-Country Determinants of CDS Market Size and Turnover We are interested in how average CDS Turnover (%) and CDS Market Size (%) differs for countries of different sizes and different credit qualities. The discussion above suggests that while countries with high CDS Turnover (%) are rather large economies, smaller countries overall have a higher CDS Market Size (%). This is further illustrated by Figure 2 showing that CDS Turnover (%) increases in country size while CDS Market Size (%) decreases in country size. [Figure 2 here] We run an OLS-regression with average CDS Market Size (%) and average CDS Turnover (%) as dependent variables. 8 The size of the economy is captured by the natural logarithm of the countries GDP. We further include the debt level as well as indicator variables for the different rating notches. 9 We use Rating BB as the omitted 8 Averages are calculated over our two year sample period. 9 We use the S&P rating, however, results are qualitatively unaffected if the Fitch or Moody s rating is used. 11

12 category. This allows for a direct test of a potential investment-grade threshold effect, i.e., the indicator variable Rating BBB captures the difference in CDS Market Size (%)/CDS Turnover (%) for countries just above the investment-grade threshold relative to countries just below the threshold. The regression results are presented in Table 3. [Table 3 here] Consistent with the descriptive section, the results show that larger economies have a higher average CDS trading activity but a smaller CDS Market Size (%). Countries with stronger credit protection, i.e., countries with common law legal systems, have smaller CDS markets. Similar to the corporate CDS market (Oehmke and Zawadowski (2014)), countries with a credit rating just above the investment grade cutoff tend to have a larger CDS Market Size (%) levels. Interestingly, while the level of CDS outstanding is higher for investment grade countries, we find that the average CDS Turnover (%) is much higher for countries with a rating below investment grade Time-Series Determinants of CDS Market Size and Turnover We intend to explore how changes in credit risk, equity market movements, CDS market movements, and macroeconomic policy events affect the CDS market. This is best illustrated by a simple example for one country. Figure 3 plots the US CDS spread, Trading Volume (in mio USD),andCDS Market Size (in mio USD) over our sample period. We see three peaks in the Trading Volume (in mio USD) corresponding to three major events relevant for the US debt market. 11 In April 2011 Standard and 10 All results are robust to controlling for the average CDS spread and the average squared CDS spread. 11 Note that Trading Volume (in mio USD) is per definition a gross item and can therefore be quite high when compared to CDS Market Size (in mio USD) which is a net item. 12

13 Poor s announced a negative outlook on the US credit Rating. The actual downgrade happened in early August. We see a further peak in trading activity around the time the US hit the debt ceiling in May While the trading volume shows a clear reaction to rating events, net credit protection outstanding does not. We do not see a large increase in net protection for the negative outlook, but we do see a large increase in net credit protection around the time of the actual downgrade. How can these facts be interpreted? Figure 4 provides a visualization of four generic types of trades in the CDS market. The starting point is a situation with an aggregate net notional outstanding of 100. For example, this could be a situation where pensions funds have bought a combined credit protection of 100 from banks who act as suppliers of CDS protection. If one of the pension funds buys protection with a face value of 10 from one of the banks, this would be reflected in a trading volume of 10 and an increase in the net CDS outstanding of 10, as indicated in Case 1 in Figure 4. If one of the pension funds unwinds its position by entering into an offsetting CDS long position, the trading volume would be reflected in a decrease in the net notionals outstanding, as indicated in Case 2 in Figure 4. If CDS positions are reallocated within a group, for example one bank selling CDS protection to another bank then the trading volume will not affect the net notionals outstanding (Case 3 and Case 4 in Figure 4). By simultaneously observing both the trading volume and the change in the net notional outstanding we are therefore able to draw conclusions about which of the CDS trading types described above is actually taking place in the market. In light of this discussion, we see that in the US example described above the trading volume associated with the negative outlook led to a reallocation of CDS contracts, 12 A cap set by Congress on the amount of debt the federal government can legally borrow. 13

14 while the downgrade itself resulted in trading volume, which expanded the net demand and supply of credit protection in the market for US sovereign CDS. Figure 5 provides a more general univariate analysis for weeks with versus weeks without downgrades for the entire sample. The left-hand picture show the net notionals outstanding CDS Market Size (%) and the right-hand picture shows the trading volume CDS Turnover (%). We find a higher trading activity in weeks with downgrades but no severe differece in CDS Market Size (%). [Figure 3-5 here] We now turn to a multivariate analysis and run a Newey-West regression with three weeks lag using CDS Turnover (%) and CDS Market Size (%) as dependent variables. 13 We include CDS spread as well as the return of the countries equity index in the respective week as exogenous variables. We further include squared equity market returns and squared spread changes to account for a possible non-linear relationship between market movements and trading activity. We include dummies that equal one if a rating event took place within the respective week and zero otherwise. We consider both actual rating changes as well as changes in credit rating outlooks because the market reaction may already take place in the week of the outlook change. We also include dummy variables that equal one if a rating event took place in the following week to account for the possibility that the market anticipates these events. Further included are indicator variables for weeks with EU Summits, LTRO announcements, and LTRO implementations. To account for a possible seasonal variation in CDS 13 We multiply CDS Market Size (%) by 100 to avoid very small coefficients. 14

15 market activity we include month of the year effects. Also included are country fixed effects. Regression results are reported in Table 4 and Table 5. [Table 4-5 here] Interestingly, the results of model (4) show that increases in the CDS spread are associated with a reduction in CDS Market Size (%). However, the relationship is nonlinear as the coefficient for CDS Spread 2 is significantly positive, implying that the reaction for large absolute changes in CDS spreads is more severe. Results for equity returns show that an increase in equity return is associated with an increase in CDS Market Size (%). The relationship between CDS Turnover (%) and CDS spreads is also U-shaped: large absolute changes in CDS spreads are associated with a higher CDS Turnover (%). The same is true for large absolute changes in CDS spreads. Turning to the credit rating changes, we find a strong increase in trading activity in weeks with downgrades and negative rating outlooks. The effects are also economically large with a CDS Turnover (%) increase of 40% (30%) relative to the mean in weeks with downgrades (negative rating outlooks). Negative events seem to be anticipated by the market as the CDS Turnover (%) already increases in the week before a downgrade or a negative outlook. We do not find a reaction to upgrades of positive rating outlooks. Interestingly, while we find a strong impact of credit rating changes on CDS Turnover (%), wedonotfindachangeincds Market Size (%) in the respective weeks. This suggests that even negative rating events do not lead to a significant increase in the net demand for CDS protection but rather to a reallocation of net demand or net supply between different market participants. The strong impact of credit rating changes on CDS trading volume is robust to controlling for changes in CDS spreads and squared 15

16 changes in CDS spreads as well as changes in the aggregate equity returns in the respective country and squared changes (see column (4) in Table 5). Thus, agencies ratings do not only matter because they are a proxy for the creditworthiness of the sovereign. We conclude that agencies ratings are a major determinant of trading in the sovereign CDS market above and beyond the creditworthiness of the sovereign. Looking at ECB policy we find that the CDS market trading activity increases before EU-Summits. This effect, however, is solely driven by an increased trading activity for the EU-15 Members. 14 CDS Market Size (%) increases in weeks with LTRO announcements and decreases around EU Summits. While these results are interesting, we refrain from drawing strong conclusions from these findings as ECB interventions and EU Summits are infrequent events during our sample period. 4 Conclusion This paper investigates the determinants of the sovereign credit default swap (CDS) trading activity and CDS market size. Our main results for the cross sectional differences in average CDS Market Size (%) and average CDS Turnover (%) are as follows: (1) Smaller countries, countries with weaker creditor rights, and countries that have a credit rating just above the investment grade cutoff, have significantly higher CDS Market Size (%) levels. (2) Larger countries, and countries with below investment grade ratings have significantly higher average CDS Turnover (%). Analyzing how equity returns affect CDS Market Size (%), we find a U-shaped relationship. Similar results can be found for the relationship between CDS Turnover 14 These results are available upon request. 16

17 (%) and equity returns, implying that larger absolute changes in the equity market are associated with an increase in CDS Market Size (%) and a higher CDS trading activity. We also find a U-shaped relationship between changes in CDS spreads and CDS Turnover (%). We further look at the CDS market response to credit rating changes. We show that both weeks with downgrades and weeks with negative outlooks are associated with a significantly higher CDS Turnover (%). Interestingly, negative credit rating events seem to be anticipated by the market as the trading activity already increases in the week before the actual event. We do not find a CDS Turnover (%) reaction for positive rating events and also no change in CDS Market Size (%) around negative credit rating events. This simultaneous analysis of trading activity and the market size allows us to draw conclusion about the nature of CDS trading. In particular it suggests that negative rating events do not lead to a significant increase in the net demand for CDS protection but rather to a reallocation of net demand between different market participants. Overall, we conclude that agencies ratings are a major determinant of the sovereign CDS market size and trading activity in the sovereign CDS market. 17

18 References Afonso, A., D. Furceri, and P. Gomes (2011). Credit ratings and the euro area sovereign debt crisis. Working Paper. Dieckmann, S. and T. Plank (2012). Default risk of advanced economies: An empirical analysis of credit default swaps during the financial crisis. Review of Finance 16 (4), Finnerty, J. D., C. D. Miller, and R.-R. Chen (2013). The impact of credit rating announcements on credit default swap spreads. Journal of Banking & Finance 37 (6), Hilscher, J. and Y. Nosbusch (2010). Determinants of sovereign risk: Macroeconomic fundamentals and the pricing of sovereign debt. Review of Finance 14 (2), Ismailescu, I. and H. Kazemi (2010). The reaction of emerging market credit default swap spreads to sovereign credit rating changes. Journal of Banking & Finance 34 (12), Johnson, T. C. (2008). Volume, liquidity, and liquidity risk. Journal of Financial Economics 87 (2), Löffler, G. (2004). Ratings versus market-based measures of default risk in portfolio governance. Journal of Banking & Finance 28 (11), Longstaff, F. A., J. Pan, L. H. Pedersen, and K. J. Singleton (2011). How sovereign is sovereign credit risk? American Economic Journal: Macroeconomics 3, Norden, L. and M. Weber (2004). Informational efficiency of credit default swap and 18

19 stock markets: The impact of credit rating announcements. Journal of Banking & Finance 28 (11), Norden, L. and M. Weber (2009). The co-movement of credit default swap, bond and stock markets: an empirical analysis. European financial management 15 (3), Oehmke, M. and A. Zawadowski (2014). The anatomy of the CDS market. Working Paper. Pan, J. and K. J. Singleton (2008). Default and recovery implicit in the term structure of sovereign CDS spreads. Journal of Finance 63 (5), Porta, R. L., F. L. de Silanes, A. Shleifer, and R. Vishny (1999). The quality of government. Journal of Law, Economics, and Organization 15 (1), Reisen, H. and J. Von Maltzan (1999). Boom and bust and sovereign ratings. International Finance 2 (2), Shachar, O. (2012). Exposing the exposed: Intermediation capacity in the credit default swap market. Working Paper, NYU Stern. Stulz, R. (2010). Credit default swaps and the credit crisis. Journal of Economic Perspectives 24 (1), Treisman, D. (2000). The causes of corruption: A cross-national study. Journal of Public Economics 76 (3),

20 A Appendix A.1 Variable Definitions Variable Name Source Definition CDS Market Size (in mio USD) DTCC The net notional CDS amount outstanding in week t in million USD. CDS Trading Volume (in mio USD) DTCC The CDS risk transfer activity in week t in million USD. CDS Market Size (%) DTCC/World Economic Outlook The net notional CDS amount outstanding in week t in percent of the countries total Database debt. CDS Turnover (%) DTCC The CDS Trading Volume in week t in percent of the net notional CDS amount outstanding in week t. GDP World Economic Outlook Database The countries GDP in billion USD. Total Debt World Economic Outlook Database The countries total debt in billion USD. Equity Return Datastream The return of the MSCI Country Index in week t in percent. CDS Spread Datastream The 5 year CDS spread in week t in basis points. Rating AA (or higher) Bloomberg A dummy that equals one if the country was rated AA- or better in week t by S&P and zero otherwise. Continued on next page 20

21 - continued from previous page - Variable Name Source Definition Rating A Bloomberg A dummy that equals one if the country was rated A+, A, or A- in week t by S&P and zero otherwise. Rating BBB Bloomberg A dummy that equals one if the country was rated BBB+, BBB, or BBB- in week t by S&P and zero otherwise. Rating BB Bloomberg A dummy that equals one if the country was rated BB+, BB, or BB- in week t by S&P and zero otherwise. Rating B (or lower) Bloomberg A dummy that equals one if the country was rated B+ or lower in week t by S&P and zero otherwise. Upgrade Bloomberg A dummy that equals one if the country was upgraded by S&P, Moody s or Fitch in week t and zero otherwise. Downgrade Bloomberg A dummy that equals one if the country was downgraded by S&P, Moody s or Fitch in week t and zero otherwise. Positive Outlook Bloomberg A dummy that equals one if the country received a positive outlook by S&P, Moody s or Fitch in week t and zero otherwise. Negative Outlook Bloomberg A dummy that equals one if the country received a negative outlook by S&P, Moody s or Fitch in week t and zero otherwise. Continued on next page 21

22 - continued from previous page - Variable Name Source Definition LTRO Announcement ECB homepage A dummy that equals one if the EU announced the implementation of a longer-term refinancing operation in week t and zero otherwise. LTRO Implementation ECB homepage A dummy that equals one if the EU implemented of a longer-term refinancing operation in week t and zero otherwise. EU Summit ECB homepage A dummy that equals one if a EU summit took place in week t and zero otherwise. Common Law Treisman (2000) A dummy that equals one if the country has a common law legal system and zero otherwise. 22

23 A.2 Figures and Tables 23

24 Figure 1: Trade Compression and Central Counterparty Clearing This figure provides illustrative examples for Trade Compression and Central Counterparty Clearing. 24

25 Figure 2: Country Size, CDS Market Size, and CDS Turnover This figure shows the average CDS Turnover (%) and CDS Market Size (%) for different quantiles of ln(gdp). 25

26 Figure 3: US CDS Trading Volume, Market Size, and CDS Spread This figure shows the weekly CDS spread for contracts denominated in USD with a maturity of 5 years. Further included are CDS Trading Volume and CDS Market Size (both in million USD). 26

27 Figure 4: Illustration of the effect of CDS Turnover on the CDS Market Size This figure provides an illustration of the effect of CDS Turnover on the CDS Market Size 27

28 Figure 5: Downgrades, CDS Market Size, and CDS Turnover This figure shows the trading activity and changes in CDS Market Size (%) in weeks with vs. weeks without credit rating downgrades. 28

29 Table 1: Summary Statistics This table provides summary statistics for the main variables used in our analysis. Variables are defined in Appendix A.1. N Mean sd p25 p50 p75 Panel A: CDS Market/Equity Market Characteristics CDS Market Size (in mio USD) 5,974 4, , , , CDS Trading Volume (in mio USD) 5, , CDS Market Size (%) 5, CDS Turnover (%) 5, CDS Spread (bp) 6, Equity Return (%) 6, Panel B: Country Characteristics Total Debt/GDP 6, Panel C: Credit Ratings Rating AA (or higher) 6, Rating A 6, Rating BBB 6, Rating BB 6, Rating B (or lower) 6, Upgrade 6, Downgrade 6, Positive Outlook 6, Negative Outlook 6,

30 Table 2: Descriptive Statistics by Country The table contains descriptives by country for selected variables. Variables are defined in Appendix A.1. Country CDS Market Size (m$) CDS Trading Volume (m$) CDS Market Size (%) CDS Turnover (%) Total Debt/GDP CDS Spread (bp) Argentina 2, Australia 3, Austria 6, Belgium 6, , Bolivia 2, Brazil 16, , Bulgaria Chile China 7, , Colombia 1, Croatia Czech Republic Denmark 2, Egypt Continued on next page 30

31 - continued from previous page - Country CDS Market Size (m$) CDS Trading Volume (m$) CDS Market Size (%) CDS Turnover (%) Total Debt/GDP CDS Spread (bp) Estonia Finland 2, France 19, , Germany 17, , Greece 5, ,642.0 Hungary 3, Iceland Indonesia 2, Ireland 4, Israel 1, Italy 23, , Japan 8, , Kazakhstan Korea 4, Latvia Lebanon Continued on next page 31

32 - continued from previous page - Country CDS Market Size (m$) CDS Trading Volume (m$) CDS Market Size (%) CDS Turnover (%) Total Debt/GDP CDS Spread (bp) Lithuania Malaysia 1, Mexico 8, , Netherlands 3, New Zealand Norway 1, Panama Peru 1, Philippines 2, Poland 2, Portugal 6, , Qatar Romania 1, Russia 4, , Saudi Arabia Slovakia Continued on next page 32

33 - continued from previous page - Country CDS Market Size (m$) CDS Trading Volume (m$) CDS Market Size (%) CDS Turnover (%) Total Debt/GDP CDS Spread (bp) Slovenia South Africa 2, Spain 16, , Sweden 2, Thailand 1, Tunesia Turkey 5, , Ukraine 1, United Kingdom 11, , United States Of America 4, Vietnam

34 Table 3: CDS Market Size, CDS Trading Volume, and Fundamentals. This table reports results of OLS regressions relating CDS Market Size (%) and CDS Turnover (%) to country characteristics. Variables are defined in Appendix A.1. Standarderrorsarereportedinparentheses.*,**,***Indicatestatisticalsignificanceatthe10%,5%,1%level. (1) (2) (3) (4) Market Size (%) Market Size (%) Trading Volume (%) Trading Volume (%) ln(gdp) (0.16) (0.17) (0.62) (0.70) Total Debt/GDP (0.57) (0.58) (2.20) (2.32) Common Law (0.54) (2.18) Rating AA (or higher) (0.67) (0.71) (2.56) (2.86) Rating A (0.73) (0.73) (2.80) (2.94) Rating BBB (0.64) (0.66) (2.46) (2.67) Rating BB (omitted) (omitted) (omitted) (omitted) Rating B (or lower) (0.96) (0.94) (3.67) (3.80) Observations R

35 Table 4: Determinants of the CDS Market Size This table reports the results of a Newey-West regression with 3 weeks lag relating CDS Market Size (%*100) to country characteristics, rating changes, and macroeconomic policy events. Variables are defined in Appendix A.1. Standarderrorsarereportedinparentheses.*,**,***Indicatestatisticalsignificanceatthe 10%,5%,1% level. (1) (2) (3) (4) CDS Market Size (%*100) CDS Market Size (%*100) CDS Market Size (%*100) CDS Market Size (%*100) CDS Spread (0.0589) (0.0599) CDS Spread (0.0006) (0.0006) Equity Return (0.0427) (0.0475) Equity Return (0.0048) (0.0053) Upgrade (2.9016) (2.8644) Upgrade t (1.7891) (1.8393) Downgrade (0.6332) (0.6624) Downgrade t (0.8855) (0.8925) Positive Outlook ( ) ( ) Positive Outlookt (5.8637) (5.8406) Negative Outlook (1.4403) (1.5281) Negative Outlook t (1.4031) (1.4740) LTRO Announcement (0.7964) (0.8230) LTRO Announcementt (0.6990) (0.9072) LTRO Implementation (1.6483) (1.6753) LTRO Implementation t (1.1104) (1.1278) EU Summit EU Summitt 1 (1.1524) (1.2038) (0.8675) (0.8727) Observations 5,915 5,915 5,915 5,915 R Country Fixed Effects Yes Yes Yes Yes Month-of-Year Effects Yes Yes Yes Yes 35

36 Table 5: Determinants of the CDS Trading Volume This table reports the results of a Newey-West regression with 3 weeks lag relating the Trading Volume (%) to country characteristics, rating changes, and macroeconomic policy events. Variables are defined in Appendix A.1. Standarderrorsarereportedinparentheses.*,**,***Indicatestatisticalsignificanceatthe 10%,5%,1% level. (1) (2) (3) (4) Trading Volume (%) Trading Volume (%) Trading Volume (%) Trading Volume (%) CDS Spread (0.1381) (0.1458) CDS Spread (0.0013) (0.0013) Equity Return (0.0545) (0.0598) Equity Return (0.0089) (0.0098) Upgrade (3.3069) (6.3342) Upgrade t (1.4618) (1.4746) Downgrade (1.8495) (1.8693) t 1 Downgrade (1.5522) (1.5859) Positive Outlook (3.7671) (3.3819) Positive Outlookt ( ) (5.2007) Negative Outlook (2.5582) (2.5404) t 1 Negative Outlook (1.8476) (2.0232) LTRO Announcement (0.9991) (0.9826) LTRO Announcementt (1.1114) (1.3451) LTRO Implementation (1.2841) (1.2706) LTRO Implementation t (1.5830) (1.6037) EU Summit (1.1391) (1.1660) EU Summitt (0.9900) (1.0067) Obs. 5,917 5,973 5,973 5,917 R Country Fixed Effects Yes Yes Yes Yes Month-of-Year Effects Yes Yes Yes Yes 36