What drives sovereign credit default swaps? The impact of US credit risk and international transmission channels on sovereign CDS

(1)

What drives sovereign credit default swaps?

The impact of US credit risk and international transmission channels

on sovereign CDS

Tim F. Steinhoff

(S2914042)

Master Thesis

Faculty of Economics and Business

University of Groningen

Supervisor: Dr. Anna Samarina

(2)

Abstract

This thesis analyzes whether the credit risk of the United States influences the credit risk of countries worldwide, indicated by the Credit Default Swap (CDS) market. Further, it investigates whether the international transmission channels, namely the bilateral trade and financial linkages as well as the exchange rate regime have an effect on those movements. Using a panel data framework with quarterly observations between 2004 and 2015, a sample of eight advanced countries is analyzed. The results suggest, that movements in the US CDS market are associated with movements in the CDS market of other countries. Furthermore, the findings show that these effects are more severe when the financial transmission channel is increasing. No evidence can be found whether the trade transmission channel or the exchange rate regime influence the effect that movements in the US CDS market have on the CDS spreads of other countries.

(3)

List of Figures

1 Stylized relationship of a CDS contract . . . 5

2 Model illustration . . . 12

3 Scatterplot between CDS spreads across the sample and US CDS spreads . . . 20

4 Overview of the average trade and financial transmission channels . . . 21

5 Marginal effects conditional on financial channel . . . 24

6 Marginal effects conditional on trade channel and ex. rate regime . . . 30

7 Bilateral trade relation to the US relative to a countries GDP . . . 38

8 Bilateral bank claims in the US relative to a countries GDP . . . 39

9 Overview CDS spreads . . . 40

List of Tables

1 Descriptive statistics - sovereign CDS spreads . . . 16

2 Descriptive statistics for independent variables . . . 18

3 Results model 1 and model 2 . . . 23

4 Sensitivity analysis - crises . . . 27

5 Sensitivity analysis - different maturities . . . 29

6 Sensitivity analysis - Ex. rate to US & dummy 2005/2006 . . . 31

7 Sensitivity analysis - Time FE & without crises period . . . 32

8 Variable description and sources . . . 36

(6)

1 Introduction

The recent decades of international economic development were marked by a constantly increas-ing globalization and financial liberalization. Most notably, the risincreas-ing technological progress, coming along with decreasing communication costs, caused major changes in the economic and financial order towards global productions and interwoven economic and financial structures.

In the context of the real economy, Baldwin (2006) describes this process as the great unbundling(s)of industries and recently also tasks within firms. However, not only the real economy is affected by these developments. The lower communication and processing costs of large amounts of data also enabled a faster development and a global integration of the financial system. In addition to that, new financial products were invented to further improve market efficiency and liquidity. These developments enabled market participants to buy or sell financial products around the world, bringing countries not only economically, but also financially closer (Lane and Milesi-Ferretti, 2008).

However, the increasing globalization and complexity of the world also bears risks. Due to the fact that countries are economically and financially closer, economic movements or shocks especially in large countries can be transmitted to other economies, which is also known as an increasing macroeconomic risk (Rajan, 2006; Savor and Wilson, 2013). A country that takes a special position in this development is the United States of America. Being not only the largest economy of the world, but also the center of international economics and finance, movements in the US economy can affect other countries around the globe (Hsiao et al., 2003; Rey, 2016).

As a consequence of these developments, financial market participants were seeking for ways that they can control and reduce this new type of risk. Supported by the fast technological progress and the resulting lower communication and data processing costs, financial products were developed in order to insure the portfolio of investors and to reduce risk (Silber, 1983).

(7)

exposures (Neal, 1996).

The increasing importance and demand for securitized products and derivatives is reflected in the high growth of the CDS market. The Bank for International Settlements (BIS) reports that the market size for sovereign CDS grew from a total of 0.17 trillion USD in 2004 up to 3.24 trillion USD in the first half of 2013 before it declined to 2.28 trillion USD in 2015 (BIS, 2015), which still is an enormous size.

In particular, in the context of international economics, sovereign CDS are an interesting indicator of how investors evaluate the credit conditions and the default probability of a country. This raises the questions of what factors drive the CDS market, what characteristics of a country affect the pricing of CDS and what role does the US play?

These questions are of great importance for both investors and the government. Investors seeking for risk insurances of their investments must keep in mind that economic movements in the US might play an important role of CDS pricing around the world. If so, this would also be important for decisions regarding the diversification of portfolios. For sovereigns and their policy makers it is important to further understand the determinants in CDS pricing. In particular, the fact of how investors see the credit risk of a country carries important information, because this risk perception will determine the price of new sovereign debt. In addition, also regulators need to understand the determinants of such derivatives in order to set up the right regulations. Therefore, the CDS market is subject to an increasing amount of research. One part of these studies examine in particular the role of international factors.

Studies like, for example, Longstaff et al. (2007) find that much of the variation across international sovereign CDS are explained by factors related to the stock market or volatility indicators of the US. This supports the argument that the US, as the economically largest and one of the most important countries in the world, seems to play a specific role also for market movements in other countries sovereign CDS.

(8)

Chinn (2004) in particular trade and financial linkages to large economies, like the US, have an international influence on stock and bond markets across different countries. Through these linkages economic movements in the US are transmitted to other countries and might therefore also determine the effect movements in the US CDS market have on the CDS market of other countries. Further, following for example Passari and Rey (2015), the choice of the exchange rate regime might also moderate how the US CDS market affects another country’s CDS spreads.

The questions raised in this thesis are therefore related to the determinants of CDS market movements around the world. Does the changing credit risk of the US, indicated by changing CDS spreads explain movements in other sovereign CDS markets? And do the bilateral trade and financial linkages to the US or the exchange rate regime of a country moderate this effect?

Aiming to find an answer to the questions raised above, this thesis will contribute to the existing literature and further investigate the determinants of CDS pricing. For this purpose, a panel data framework with a sample consisting out of eight economies and their bilateral trade as well as financial exposure to the US will be applied. Given the relative novelty of CDS products, the sample period ranges from 2004 to 2015 with quarterly observations. Further, by adding information about the chosen exchange rate regime, the question whether this decision moderates the impact US movements have on other countries CDS spreads, will also be investigated.

The results show that the US CDS market is generally correlated with CDS markets around the world. This supports the argument that movements in the US CDS market, as part of the center of international economics, also have an impact on the default probability of other countries. Further it finds, that this effect increases for countries with a higher financial transmission channel. However, no evidence can be found whether the trade transmission channel or the exchange rate regime also influence this effect.

(9)

2 Related Literature & Hypothesis

The following paragraph introduces the related literature. The first part focuses on the functioning scheme of CDS in general, before turning to the special role of sovereign CDS and their driving factors. The second part deals with the international transmission channels and how they transmit changing economic conditions and shocks from the US to other countries. At the end, based on the theoretical framework and related literature, the hypotheses of this study will be derived.

2.1 Credit Default Swaps and their meaning for the economy

In principle, CDS contracts allow investors and other market participants to insure and to trade credit risk, defined as the risk that a credit receiver cannot meet his obligations due to a default (Duffie and Singleton, 2012). In making the contract, two parties, the protection seller and the protection buyer, are preliminarily involved. Both are coming to an agreement to insure a credit event of a third party, the so called reference entity. To do so, the protection buyer agrees to pay a yearly or quarterly amount of money to the seller, also known as the CDS premium. In return for paying the premium, he receives the insurance against a default of the reference entity (Weistroffer et al., 2009). Figure 1 illustrates the stylized relationship between protection seller and protection buyer and indicates the flow of payments in respect to different events that can occur. The reference entity hereby is not involved in any flow of payments and therefore, no connection between either protection buyer or seller and the reference entity is drawn. Furthermore, the protection buyer does not normally need to have any connection to the reference entity when he will insure himself against a default event (ISDA, 2014).

In general, protection seller and also the majority of buyers are banks and hedge funds. Moreover, insurance companies and pension funds are using credit derivatives to transfer their credit risk to institutions that have the ability and willingness to take it (Bomfim, 2005; Weistroffer et al., 2009).

(10)

Figure 1: Stylized relationship of a CDS contract

Protection buyer Protection Seller

Premium

Protection No credit event: No Payment

Credit Event: Payment

Reference entity

Source: Weistroffer et al. (2009).

seller receives the payments from the protection buyer and the contract will expire according to its maturity. However, in case of a default event, the CDS seller agrees to pay the buyer a compensation for their losses.

The second specification and an essential part of a CDS contract is the definition of the insured event. Other than the name CDS might suggests, not only a complete default of the reference entity will trigger the payment from the protection seller to the protection buyer, but also cases like, for example, restructuring of debt can be part of the insurance (Weistroffer et al., 2009; Gregory, 2012; ISDA, 2014).

Another important factor is the pricing of a CDS product which is in practice also called CDS spread or CDS premium1. Following Weistroffer et al. (2009) and Markit (2014), the premium a

CDS buyer needs to pay is determined by mainly two parts. First the default probability of the reference entity and second the estimated recovery rate after default:

CDS premium = default probability ∗ (1 − rate of recovery) (1) In general, CDS products give some important benefits to the buyer and further also to the economy as a whole. First of all, they allow investors to hedge themselves from losses

1_{The CDS premium is paid by the protection buyer to the protection seller and is denominated in basis points.}

(11)

in case a reference entity will face bankruptcy or any other event that is part of the contract (Smithson, 2003; Markit, 2014). Secondly, according to the reference entity, CDS contracts can be classified into two categories of corporate CDS or sovereign CDS. Both follow in the end the same principles and mechanisms, but sovereign CDS provide investors additionally the possibilities to hedge macroeconomic risk resulting from exposure to governments (Fontana and Scheicher, 2010).

Furthermore, CDS markets can also be seen as an indicator for a countries’ perceived credit risk (AIMA, 2011; Longstaff et al., 2007). This in turn does have important implication for a government. With a higher credit risk, the borrowing and re-financing costs of a governments debt might increase because investors will demand a higher compensation for holding this debt (Fender et al., 2011). Several empirical studies find this relationship between sovereign CDS and the corresponding government bond market (see e.g. Fontana and Scheicher (2010); Delatte et al. (2012)).

In conclusion, sovereign CDS give investors the ability to hedge themselves against the macroeconomic credit risk in their portfolio. At the same time the market is an indicator for countries that are concerned with raising funds, how investors evaluate the credit risk of a country. Due to this risk perception, the CDS market is subject to a growing amount of research, analyzing the factors that determine changes in CDS spreads. Given the importance for the economy as a whole and the possibility for investors to hedge themselves against macroeconomic risk with CDS, this thesis will focus on sovereign CDS.

2.2 Determinants of CDS spreads

Following Heinz and Sun (2014) the literature identifies three sets of factors explaining sovereign CDS spreads. These are macroeconomic fundamentals, liquidity and global factors.

(12)

Fontana and Scheicher, 2010).

Further, also the state of a country’s financial system is related to its CDS spreads. As argued by Dieckmann and Plank (2012), in case of a crisis and an ailing financial sector, countries with a large international financial system might have to offer support by taking more debt or initiating economic stimulus packages for the economy, leading to an increase of sovereign credit risk and hence higher CDS spreads. Dieckmann and Plank (2012) also show, that the reserves of foreign currencies are often seen as a countries’ ability to pay. Countries with high foreign reserves can liquidate these holdings to meet other debt obligations. This ability is seen as a positive indicator for the credit risk and consequently foreign exchange reserves are negatively correlated with CDS spreads.

Hull et al. (2004) and Arezki et al. (2011) show that CDS spreads are determined by the development of credit ratings. A low credit rating signals a larger default risk and hence the sovereign CDS spreads are higher for countries with lower ratings.

Fontana and Scheicher (2010) argue that liquidity is an important determinant of CDS spreads. For contracts with a high bid-ask spread, the CDS premium tend to increase (higher bid-ask spreads are indicators for less liquidity in the market (Lybek and Sarr, 2002)). Therefore, the ability to buy or sell CDS contracts determine the price of an insurance against the default of sovereigns (Palladini and Portes, 2011; Calice et al., 2013).

Other studies investigate how the sovereign CDS market is determined by global factors. Liu and Morley (2012) find that for France the exchange rate to the USD is driving CDS spreads. They conclude, that managing the exchange rate is important when a country is concerned about insuring its debt because a floating exchange rate will lead to lower CDS spreads. Longstaff et al. (2007) study the components that explain CDS spread changes. They provide evidence that spreads across the world are related to the U.S. stock market and other factors like global investment flows rather than to domestic measurements. They find, for example, that the S&P 500 index is negatively correlated with sovereign CDS spreads around the world. Related to that, Pan and Singleton (2008) show that sovereign CDS of Mexico, Turkey and Korea are driven by global credit exposures of investors as well as spillovers of U.S. real economic growth.

(13)

relationship between sovereign CDS spreads and the S&P 500 volatility index (VIX). A low VIX indicates less uncertainty and risk aversion (Rey, 2015). In case of a higher uncertainty the VIX will rise, investors demand more insurances and hence CDS spreads are rising.

Fender et al. (2011) state that sovereign credit risk, indicated by the CDS market, is generally driven by global factors rather than by own characteristics of a country’s economy. In addition, the authors provide evidence that sovereign CDS are driven by US monetary policy announcements and decisions.

To summarize, the literature related to the driving factors of sovereign CDS suggest that the credit risk of a country is driven by many diverse determinants. Macroeconomic fundamentals and also global factors play an important role in explaining changes in CDS spreads. Striking is, that many of the global factors are related to the US economy. The majority of studies hereby investigates, how sovereign CDS worldwide react directly to changes in the US economy. However, they do not take into account whether the credit risk of the US is associated with the credit risk of other countries and whether there are factors that determine this effect, like the international transmission channels.

2.3 The role of the US

Before turning to the international transmission channels, the following paragraph will shortly introduce, why the US plays an outstanding role in international economics. In context of this study, it is of interest, as the literature on the driving factors of sovereign CDS suggest that economic movements in the US are playing an important role in explaining CDS spreads around the world. In general, a large amount of literature investigates the role of the US in the global economy. The special interest to investigate the US arises not only through the fact that the US is the largest economy in the world (measured by nominal GDP in 2015 (IMF, 2015)) but also due to the interwoven structure of the world economy.

(14)

countries and studies such as Rey (2016) also mentions the outstanding role the US takes in the international financial system.

2.4 International transmission channels

The international transmission channels determine how economies are linked together. The liter-ature identifies two major channels that transmit changing economic conditions and respectively shocks from one country to another. These channels are the trade and the financial channel (Chui et al., 2004; Balakrishnan et al., 2011). The trade channel transmits shocks through trade linkages between two economies. A shock in country A, e.g. a change in US interest rates, will be transmitted to the output of country B (Forbes, 2002). To understand the mechanisms behind it, the literature further divides the trade channel into three different effects: The competitiveness effect, income effect and the cheap import effect.

The competitiveness effect, first mentioned by Corsetti et al. (2000), works through the devaluation of a countries currency. The devaluation of a currency will lead to a reduction in the relative price of exporting goods resulting in a lower competitiveness of another country that produces similar goods. The income effect works through the effect a crisis or a shock will have on the income level of a country. If the income level declines, the demand for goods from abroad will consequently decrease, affecting the bilateral trade relations between two countries. The last channel, namely the cheap import effect arises also through currency devaluation. With a devaluation, the relative prices of exporting goods will be lower resulting in an improvement of another countries terms of trade (Forbes, 2002; Chui et al., 2004).

Several studies build on this framework to test whether these channels can be observed in reality. Forbes (2002) find that the competitveness effect is the most significant channel, indicating that countries with a high trade exposures to a crisis country are more vulnerable. Eichengreen et al. (1996) and Glick and Rose (1999) provide also evidence, that trade linkages between countries are more important to propagate shocks than macroeconomic similarities.

(15)

for example, through portfolio rebalancing or deleveraging effects (Chui et al., 2004). When an investor is exposed to a country in which the economic conditions are changing or a crisis occurs, he usually wants to decrease those risky investments (Caramazza et al., 2000). Consequently, economies with higher aggregated exposures to other countries are more likely to be affected to economic movements outside of their own jurisdictions (Balakrishnan et al., 2011). Several studies find evidence of this financial transmission channel (see e.g. Dornbusch and Park (2000); Van Rijckeghem and Weder (2001); Fratzscher (2003); Kaminsky and Reinhart (2003)).

More comprehensive studies like Forbes and Chinn (2004) investigate both, bilateral trade and financial linkages and why movements in large economies such as the US have an impact on the bond and stock markets of other countries. They find that more bilateral trade rather than financial linkages, indicated by bilateral bank-lending, are important determinants connecting economies.

However, having strong trade or financial transmission channels does not necessarily mean that an economy is completely dependent on another country. In particular, the exchange rate regime choice can influence how a country will be affected by economic movements in another country. Therefore, in international economics, the role of exchange rate regimes has been stressed out in recent years.

Again, the US hereby takes a special position because of its outstanding role as the biggest economy and the center in international economics and finance (Longstaff et al., 2007; Bruno and Shin, 2015; Rey, 2016). Following this, a fixed exchange rate regime will lead to direct spillovers from the center country (the US) to other economies, making them more vulnerable to economic movements from outside of their own sovereignty (Miniane and Rogers, 2007; Passari and Rey, 2015; Rey, 2016). With a falling real interest rate in the US, for example, deposits denominated in other currencies than the USD are becoming relatively more attractive than deposits denominated in USD. As a result, the USD will depreciate. This makes US goods in relation to foreign goods cheaper. When a currency is pegged to the USD those movements are consequently transmitted to other currencies and countries.

(16)

The fact, that the exchange rate plays an important role in transmitting economic condition directly to another country, they can also play a role in explaining movements in the CDS market as shown by Liu and Morley (2012).

To sum up, the literature about the international transmission channels show that the bilateral trade and financial linkages as well as exchange rate regime of economies can determine the transfer of changing conditions from one country to another. Given the fact that the international transmission channels are important determinants for an economy, they do have an impact on bond and stock markets. A fact that is of particular interest when economic conditions in large economies, like the US, are changing. The literature hereby intensively discusses the impact of these transmission channels on different markets of an economy. However, a market that has not been investigated in this context is the market of sovereign CDS.

2.5 Research Interest and Hypothesis

The previous sections focused either on the international transmission channels or on driving factors of sovereign CDS. However, there is little research combining both. Longstaff et al. (2007) examine international factors determining sovereign CDS, however they do not examine whether CDS spreads are associated with the credit risk of the US and further they also do not analyze whether the international transmission channels to the US and the exchange rate regime play a transmitting role. Forbes and Chinn (2004) investigate the effect of bilateral trade flows and bank lending on international stock and bond markets but they neither include the market for sovereign CDS nor the exchange rate regime in their analysis.

(17)

Figure 2: Model illustration

US CDS spreads CDS spreads other countries

Transmission channels: Trade channel Financial channel

Ex. rate regime

Source: own illustration

Supported by the fact that the US is one of the most important economies in the world, several studies show that economic movements in the US are associated with movements in markets around the world (Hsiao et al., 2003; Longstaff et al., 2007; Bruno and Shin, 2015; Rey, 2016). The first hypothesis therefore is:

Hypothesis 1: CDS spreads of the US affect the CDS market of other countries.

Further, as the literature regarding the international transmission channels in section 2.4 has shown, the trade and the financial transmission channels are important factors through which economic movements are transmitted from one country to another (Fratzscher, 2003; Forbes and Chinn, 2004). More bilateral trade exposure from a country to the US and therefore a higher trade channel, will make the trade partner more vulnerable to negative spillovers from changing conditions in the US. This, in turn, might result in a higher perceived credit risk of the sovereign, indicated by higher CDS spreads (Longstaff et al., 2007). Therefore, the second hypothesis is:

Hypothesis 2: CDS spreads of the US affect the CDS market of other countries more severe when the trade channel is larger.

(18)

changing conditions in the US (Caramazza et al., 2000; Fratzscher, 2003; Chui et al., 2004). The third hypothesis therefore is:

Hypothesis 3: CDS spreads of the US affect the CDS market of other countries more severe when the financial channel is larger.

Another channel that can play a role in how much countries are affected by changing US CDS spreads is the choice of the exchange rate regime. As argued in section 2.4, countries with a flexible exchange rate might have the possibility to protect themselves from spillovers (Forbes and Chinn, 2004; Passari and Rey, 2015; Rey, 2016). The fourth hypothesis therefore is:

Hypothesis 4: CDS spreads of the US affect the CDS market of other countries more severe when the exchange rate is fixed.

3 Methodology and Data

3.1 Methodology

In order to test the hypothesis derived in section 2.5 and given the panel structure of the data (as described in section 3.4 and section 3.6), a panel analysis is conducted. In general, panel data have the advantage to analyze data in both, the cross-section as well as time-series dimension. To analyze to what extend the US CDS market affects CDS spreads around the world and what role the international transmission channels play, this analysis will make use of interaction variables. Therefore, the following model is considered to test the first three hypothesis and will be further denoted as model 1:

CDSit = β0+ β1U St+ β2T rit+ β3F iit+ β4U St× T rit+ β5U St× F init

+β6Stit+ β7Resit+ β8SP 500t+ β10P Debtit+ β11P rCreit+ αi+ it

(2)

(19)

for the trade transmission channel, Fiita proxy for the financial transmission channel, Stit the

returns of a countries corresponding leading stock index, Resitthe foreign exchange reserves,

SP500tthe returns of the S&P 500 index, PDebtitthe public debt and PrCreitprivate bank credit.

αirefers to the unobserved individual effect (fixed effect) and itrefers to the error term2.

In order to address the fourth hypothesis, model 1 (equation 2) is extended with information regarding the exchange rate regime of a country. Therefore the following model is used and further denoted as model 2:

CDSit = β0+ β1U St+ β2T rit+ β3F iit+ β4Exit+ β5U St× T rit+ β6U St× F init

+β7U St× Exit+ β8Stit+ β9Resit+ β10SP 500t+ β11P Debtit

+β12P rCreit+ αi+ it

(3)

where Exitthe exchange rate regime

3.2 Estimation method and assumptions

Conducting a panel analysis requires some further specification and assumptions to cope with problems that can lead to inefficient estimates and inference statistics.

At first, to allow heterogeneity across individuals, the decision to rely on fixed effects (within estimates) or random effect (between estimates) has to be done. In order to do so, the models are estimated using random and fixed effects. The result are compared with the Hausman test to make a decision which specification should be used (Hill et al., 2011).

Another issue is that the standard errors of the estimations could be biased due to serial correlation. Serial correlation can be detected, when the value of a variable is dependent on the value from one time period before. In the frames of this study this would mean, that for example the CDS spreads of country i at time t+1 are dependent on the spreads of time t (Wooldridge, 2010). In order to address this issue the models are tested for autocorrelation following Wooldridge (2010) and Drukker (2003).

Moreover, another problem can arise when the error terms itdo not meet the assumption

that var(it)=σ2. Is this the case, the inference statistics could be biased and not reflect the true

2_{The error term is assumed to have a zero mean E(}

(20)

value, which is also known as heteroskedasticity (Wooldridge, 2010). Again, to address this issue the models are tested following Greene (2003).

Lastly, an important assumption of this study is, that movements in US CDS spreads is exogenous from movements in those of other countries. This assumption is supported by the fact, that the US is one of the most important economies in the world and movements in the US economy are transmitted to other countries (see e.g. Ng (2000); Forbes and Chinn (2004); Longstaff et al. (2007)). Further it is less likely that the credit risk of the US is driven by the credit risk of other counties than vice versa. Nevertheless, a time where this assumption could not hold is during the global financial crisis and the European debt crisis. To control whether both periods influence the estimations, in section 5, dummy variables covering each crisis periods are added to the estimations.

3.3 Sample selection

The sample of this study includes eight countries over the period from quarter one 2004 to quarter three 2015 (47 quarters). All countries are advanced (see table 1 for an overview). The initial sample consists of all G20 states. In order to analyze both, the trade and financial transmission channel, only countries that are reporting to the Bank for International Settlements (BIS) are considered, which left the sample size of eight countries. Due to the fact that not all bilateral bank data are available for the required time period and some CDS are also not reported in every quarter, the panel is unbalanced.

3.4 Dependent variables

(21)

Table 1: Descriptive statistics - sovereign CDS spreads

Mean

SD

Min

Median

Max

N

Australia

39.63

29.15

2.63

33.50

131.23

47 Brazil

198.93

122.96

72.80

163.00

691.70

47 France

55.19

54.28

1.50

42.56

214.86

40 Germany

17.87

17.72

1.50

12.10

76.02

46 Italy

110.56

110.68

5.60

86.64

399.56

46 Japan

31.76

25.28

2.40

26.86

91.76

47 Turkey

216.77

84.43

116.14

196.45

536.30

45 UK

48.87

29.52

8.90

44.61

123.75

31 USA

23.63

17.39

1.40

17.57

62.00

47

Notes:Table 1 reports summary statistics for 5 Year sovereign CDS spreads measured at the end of each quarter. The unit is in basis points and the period ranges from the first quarter of 2004 until the third quarter of 2015.

The CDS spreads, as well as all other variables are corrected for outliers. Outliers can influence the results, when for example the data are reported or retrieved with an incorrect notation, such as the decimal separator at the wrong place. To avoid biased results due to outliers, the data are corrected using an algorithm developed by Billor et al. (2000). The advantage of using this algorithm is that outliers are efficiently detected and all variables are treated in the same way.

(22)

3.5 Independent variables

In order to test the hypothesis stated in section 2.5, the CDS spreads of the United States are retrieved from Datastream similar to the data from section 3.4 and reported in table 1. The summary statistics for all other independent and control variables are reported in table 2. Additionally, an overview of all variables, the expected signs as well as their sources are provided in table 8 in the appendix.

Further, proxies for the trade and financial transmission channels between a country and the US are required. For the former one, the literature suggests to use the bilateral trade relation from a certain country (countryi) to the US divided by the GDP of countryi. These trade linkages

are expressed as the exports from countryi to the US plus the imports of countryi from the US

(Forbes, 2002; Forbes and Chinn, 2004; Balakrishnan et al., 2011). The data are retrieved from Datastream3. The variable can formally expressed as:

T rade channeli =

Exports to USi + Imports from USi

GDPi

(4) where i stands for the individual country.

Similar to the trade channel, the literature uses the bilateral claims of banks divided by GDP for a proxy of the financial transmission channel. The data are obtained from the Bank of International Settlements (BIS) and represent the total claims of banks from countryi in the US

plus the total claims from banks in the US in countryi(Van Rijckeghem and Weder, 2001; Forbes

and Chinn, 2004; Ehrmann and Fratzscher, 2009). The variable can be formally written as:

F inancial channeli =

Bank claims in USi + Bank claims from USi

GDPi

(5)

where i stands for the individual country.

In order to test hypothesis 4, the exchange rate regime classification provided by Reinhart and Rogoff (2004) is used. The authors classify in their coarse classification the exchange rate regime of countries on a 6-point scale, where 1 represents a pegged and 4 a freely floating exchange

3_{The data are collected as imports of the US from country}

i and exports of the US to countryi since their

(23)

Table 2: Descriptive statistics for independent variables

Mean SD Min Median Max N

Trade channel 3.17 0.88 1.49 3.00 5.58 376 Financial channel 0.08 0.07 0.01 0.07 0.32 338 Reserves 0.28 0.28 0.04 0.12 1.19 376 Stock 0.02 0.10 -0.29 0.03 0.43 368 Public debt 92.42 50.65 18.98 80.59 230.70 376 Private credit 209.75 78.62 77.30 209.00 393.20 376 SP500 0.02 0.08 -0.22 0.02 0.14 368

Exchange rate regime 2.54 1.26 1.00 3.00 4.00 376 Notes:Table 2 reports summary statistics for all independent variables. The sample period is from the first quarter of 2004 to the third quarter of 2015. See section 3.5 and section 3.6 for a more detailed explanations. The variables of public debt and private credit are only available on a yearly basis and therefore imputed using a quadratic imputation to obtain quarterly observations.

rate4_{. Due to the fact that this analysis focuses on quarterly data, however the classifications}

are provided either on a yearly or monthly basis, the value at the end of each quarter is used from the monthly data. Due to the fact that the data are only available until 2010 and in order to cope the whole sample period the data are extended until 2015. For the Eurozone countries, the exchange rate regime did not change since 2010. The data for all other countries are extended using information mainly from the Annual Report on Exchange Arrangements and Exchange Restrictionsprovided by the International Monetary Fund (see IMF (2012, 2013, 2014); CESifo (2015)).

3.6 Control variables

Longstaff et al. (2007) mentions that the amount of variables that could influence the credit risk of a country is large. As discussed in section 2.1, local and global factors might influence the credit risk of a country. Therefore, for both factors, control variables are included and explained below. Mostly all data are retrieved from Datastream. Only the private credit is taken from the BIS consolidated banking statistics. Summary statistics are reported in table 2.

The first variable to account for the state of the economy is the return of the leading stock index of each country (further denoted as Stock). In line with Longstaff et al. (2007) the data are retrieved in the local currency to avoid biases due to exchange rate movements. As a second

4_{scale 5 and 6 represents freely falling exchange rates or those with missing data. However, this is not an issue}

(24)

variable the foreign exchange reserves in relation to a countries GDP are included (Reserves). Both variables are commonly used in similar studies (see e.g. Collin Dufresne et al. (2001); Longstaff et al. (2007); Fontana and Scheicher (2010); Liu and Morley (2012)).

Also the debt-to-GDP ratio might influence the CDS pricing of sovereign CDS. Therefore, the public debt as percentage of GDP (Public debt) is included as a control variable (Hilscher and Nosbusch, 2010; Fontana and Scheicher, 2010)5_{. Similar to sovereign debt, also the private}

sector can take credit which is also related to movements in sovereign CDS spreads (Dieckmann and Plank, 2012). To account for that, the private credit as percent of GDP is included (Private credit).

Further, percentage changes of the S&P500 index (SP500) is included since US real economic growth can also play an important role explaining CDS spreads across the world (Longstaff et al., 2007).

3.7 Correlation among independent variables

To test whether there is multicollinearity among the independent variables, a correlation matrix after Pearson is estimated and reported in table 9 in the appendix. Highly and significantly correlated are only the variables Private credit and Public debt (correlation of 0.80). Denis and Mihov (2003) discuss this issue more intensively. To avoid multicollinearity, the variables are included separately.

4 Estimation results

The following section will discuss the results of the models introduced in section 3. First, some stylized facts will point out the relation between the sovereign CDS of the US and the CDS spreads of the sample countries. Further, some stylized facts about the trade and financial transmission channels will be introduced followed by the results of model 1 and model 2.

5_{The data are retrieved with an annual frequency. In order to use them for the quarterly analysis the data are}

(25)

4.1 Stylized facts

The scatter plot in figure 3 gives a first impression of the relationship between sovereign CDS spreads of the US and the CDS spreads of the sample countries. Indicated by the linear fitted regression line with a positive slope, the plot suggests that there is a positive relationship between US sovereign CDS and the CDS spreads across the sample.

In order to get an overview how the CDS spreads react over time figure 9 in the appendix reports a more comprehensive overview of CDS spreads with respect to each sample country, the average CDS spreads of the sample as well as the US CDS spreads.

Figure 3: Scatterplot between CDS spreads across the sample and US CDS spreads

0

200

400

600

800

Sample sovereign CDS (in bsp.)

0 20 40 60

US sovereign CDS (in bsp.)

Notes:Figure 3 reports the scatter plot between CDS spreads of the sample countries (y-axis) and CDS spreads of the US (x-axis). The dashed line shows the linear fitted values. The sample period is from 2004 to 2015. All values are in basis points and corrected for outliers.

(26)

crisis. Due to the fact, that both channels can vary across the sample countries, a detailed summary for all bilateral trade and financial relations between the sample countries and the US is reported in figure 7 and figure 8 in the appendix.

Figure 4: Overview of the average trade and financial transmission channels

2.5 3 3.5 4 Tr. channel (% of GDP) 2004q1 2007q1 2010q1 2013q1 2016q1 date .05 .07 .09 Fin. channel (% of GDP) _2004q1 _2007q1 _2010q1 _2013q1 _2016q1 date

Notes: Figure 4 reports an overview of the average trade and financial transmission channels. Both channels are expressed as percent of GDP.

(27)

4.2 Estimation results

Table 3 presents the estimated results. To avoid multicollinearity, public debt and private credit are added separately.

The results of model 1 in column 1 and 2 of table 3 show that in both specifications the coefficient of the US CDS spreads and the coefficient of the trade channel are positive and statistically significant on the 5% and 1% level. In contrast to that, the coefficient of the financial channel is statistically not different from zero. Further, the interaction between the US CDS spreads and the trade channel is not significant. However, the interaction between US CDS spreads and the financial channel is positive and significant on the 10% and 5% level.

The results of model 2 in column 3 and 4 of table 3 show that the coefficients of the US CDS spreads are positive and significant on the 5% level in both specification. Also the coefficient of the trade channel is significant on the 1% level. The coefficient of the financial channel and the exchange rate regime are statistically not different from zero. The interaction term between US CDS spreads and the trade channel as well as the exchange rate regime are both not significant. However, the interaction terms between US CDS spreads and the financial channel are both significant on the 10% level.

The control variables in both models have their expected sign. An increase of the S&P 500, for example, should be negatively associated with CDS spreads around the world. This is the case as indicated by the negative and on the 1% level significant coefficient in all specifications. Therefore, similar to the findings of related studies (e.g. Fontana and Scheicher (2010)) economic movements in the US have, holding all else equal, an impact on other sovereign CDS around the world. Further, the coefficient for the public debt remains insignificant, while the private credit is positive and significant on the 1% level. Consequently also a one unit increase in private credit will lead, holding all else equal and on average, to an increase in sovereign CDS spreads.

(28)

Table 3: Results model 1 and model 2 Model 1 Model 2 (1) (2) (3) (4) US CDS 1.839** 1.865** 1.672** 1.875** (0.755) (0.734) (0.823) (0.809) Trade channel 44.06*** 44.44*** 43.87*** 44.48*** (8.304) (7.874) (8.627) (7.965) Financial channel 22.73 -131.8 -23.84 -132.6 (263.6) (258.7) (283.6) (279.2)

Ex. rate regime 1.870 15.25

(18.30) (17.04) US CDS × Trade channel -0.358 -0.428 -0.330 -0.407

(0.265) (0.263) (0.268) (0.264) US CDS × Financial channel 6.586* 6.964** 6.683* 6.030*

(3.500) (3.422) (3.699) (3.643)

US CDS × Ex. rate regime 0.204 0.0813

(0.444) (0.431) Stock 131.2*** 134.7*** 126.2*** 125.3*** (34.58) (34.15) (35.22) (34.52) Reserves -173.1*** -206.0*** -166.3*** -186.9*** (44.19) (44.46) (46.66) (46.33) S&P 500 -174.7*** -200.4*** -165.5*** -184.8*** (47.30) (46.86) (48.50) (47.90) Public debt 0.162 0.258 (0.279) (0.333) Private credit 0.653*** 0.828*** (0.229) (0.254) Constant -46.85 -142.7*** -54.03 -191.6*** (32.83) (47.21) (41.50) (59.31)

Country FE Yes Yes Yes Yes

Observations 309 309 309 309

Number of country 8 8 8 8

Adj. R-squared 0.243 0.263 0.240 0.265

Notes: Table 3 reports the results of model 1 and model 2. To avoid multicollinearity public debt and private credit are included separately as explained in section 3.7. The sample period is from quarter one 2004 to quarter three 2015. The underlying CDS contracts have a maturity of five years. ***;**;* denote significance at the 1%,5% or 10% level, respectively. Clustered standard errors are reported in parenthesis.

(29)

insurance costs of the US will increase by one basis point, the CDS seller would evaluate the credit risk of other sovereigns differently and demand on average 187.50 USD more to insure the 1 million USD. Therefore, the findings support hypothesis 1.

Next to the US CDS spreads, also a one unit increase of the trade channel will lead (holding all else equal) to an average increase in other countries CDS spreads of 44.48 basis points and therefore increases the insurance costs of the fictive example by 4,448 USD.

In order to address hypothesis 2, 3 and 4, the interaction variables are of main interest. Considering the same model as above (table 3 column 4), only the financial transmission channel is positive and significant on the 10% level. To illustrate how US CDS spreads influence movements in other countries CDS spreads depending on the size of the financial channel, the marginal effects are reported in figure 5.

Figure 5: Marginal effects conditional on financial channel

0

1

2

3

4

Effect of US on other CDS spreads

.01 .03 .05 .07 .09 .11 .13 .15 .17 .19 .21 .23 .25 .27 .29 .31 Financial channel

Notes:Figure 5 reports the marginal effect of US CDS spreads on the CDS spreads of other countries holding the financial channel constant at different levels. The vertical lines indicate the 95% confidence interval. The effects are significant when the confidence intervals are above or below zero.

(30)

on other countries CDS spreads holding the financial channel constant above 0.03% of GDP6_.

Further, conditional on an increasing financial channel, indicated by the positive slope, the impact changes in US CDS spreads have on the CDS spreads of other countries increases.

The economic interpretation illustrates that holding the financial channel constant at, for example, 0.07% of GDP, a one basis point increase in US CDS spreads will lead, on average, to an increase of around one basis point in the sample countries CDS spreads. Further, when the size of the financial channel is increasing to, for example, 0.25% of GDP, a one basis point increase in US CDS spreads will lead to an average increase in the CDS spreads of other countries of two basis point. Consequently, the impact of US CDS spreads on the CDS spreads of other countries is increasing when the financial transmission channel is larger. Therefore, the results support hypothesis 3.

Due to the fact, that the interaction between US CDS spreads and the trade channel as well as the exchange rate regime remain insignificant, this thesis does not find evidence for hypothesis 2 and hypothesis 4.

To sum up, both models in table 3 show that the US CDS spreads have a positive impact on sovereign CDS spreads across the sample. Looking at how the international transmission channels moderate this effect, in particular a higher financial channel increases the effect of how changes in the US credit risk is transmitted to the CDS market of other economies.

4.3 Econometric tests

As pointed out in section 3.2, the models above are tested for several factors in order to have efficient estimates and to avoid biased results. First of all, to allow heterogeneity across indi-viduals and therefore to make use of the main advantage of panel data, the decision to rely on fixed or random effect estimations is important. Therefore, the Hausman test is applied and rejects the null hypothesis to rely on random effect estimations for both, model 1 and model 2. Consequently fixed effect estimations are used. This allows to control for omitted variables that are constant over time but vary across countries (country fixed effects).

Secondly, the standard errors of the estimations could be biased due to serial correlation. Therefore, the models are tested for order one autocorrelation following Wooldridge (2010) and

6_{The vertical lines in figure indicate the 95% confidence intervals. The marginal effects are significant when the}

(31)

Drukker (2003). This test is in particular suitable for the models in this paper, because it is not influenced by the unbalanced panel structure. The results suggest that there is serial correlation.

Thirdly, the models are tested for the presence of heteroskedasticity using a modified Wald test following Greene (2003). The test rejects the null hypothesis of homoskedastic errors and consequently heteroskedasticity is present.

In order to address the issues of autocorrelation as well as heteroskedasticity, clustered standard errors are chosen to obtain the right inferences.

5 Sensitivity analysis

To analyze whether the findings in section 4 are sensitive to modifications, several robustness checks are conducted. One factor that could influence the results are financial or economic crises. During the sample period, two crises were present at different points in time. This was the global financial crisis (starting in quarter three 2007 until quarter four 2010) and the European sovereign debt crisis (starting in quarter one 2010 until quarter three 2015 - the end of the sample).

(32)

Table 4: Sensitivity analysis - crises Model 1 Model 2 (1) (2) (3) (4) US CDS 1.891** 1.828** 2.073** 1.876** (0.741) (0.735) (0.814) (0.810) Trade channel 43.85*** 44.65*** 39.85*** 44.47*** (8.164) (7.876) (8.362) (7.979) Financial channel -109.9 -178.7 37.47 -128.0 (270.8) (262.9) (294.6) (283.3)

Ex. rate regime 26.91 15.78

(18.23) (17.84) US CDS × Trade channel -0.431 -0.380 -0.414 -0.412

(0.264) (0.267) (0.263) (0.269) US CDS × Financial channel 6.880** 6.651* 4.652 6.042*

(3.441) (3.436) (3.714) (3.651)

US CDS × Ex. rate regime 0.169 0.0914

(0.433) (0.443) Stock 134.2*** 131.9*** 113.6*** 125.3*** (34.25) (34.26) (35.04) (34.59) Reserves -207.5*** -204.3*** -183.4*** -186.3*** (44.84) (44.49) (46.21) (46.78) S&P 500 -202.0*** -189.8*** -184.3*** -185.3*** (47.26) (48.02) (47.73) (48.30) Private credit 0.627** 0.749*** 0.754*** 0.823*** (0.249) (0.248) (0.257) (0.259) Crisis -2.267 -18.25* (8.155) (10.39)

Sov. debt crisis -8.458 1.113

(8.371) (10.89)

Constant -135.9** -159.3*** -175.6*** -191.5***

(53.23) (49.98) (59.80) (59.43)

Adj. R-squared 0.260 0.263 0.271 0.263

Notes:Table 4 reports the results for model 1 and model 2 where either a dummy variable indicating the global financial crisis or the European sovereign debt crisis is added to the estimation. The underlying CDS contracts have a maturity of five years. The sample period is from quarter one 2004 to quarter three 2015. ***;**;* denote significance at the 1%,5% or 10% level, respectively. Clustered standard errors are reported in parenthesis.

(33)

different maturities, model 1 and model 2 are re-estimated. The results are reported in table 5 and suggest that changes in US CDS spreads, holding all else equal, also have a positive impact on the CDS market of contracts with different maturities. Further, the interaction term of the trade transmission channel is negative and significant (on the 1% level) in all specifications and the interaction term of the financial transmission channel is insignificant. Lastly, the interaction term between US CDS contracts with a maturity of one year and the exchange rate regime is significant on the 1% level. To further analyze the relationship between 1 year CDS spreads of the US and the trade channel as well as the exchange rate regime, the marginal effects are reported in figure 6.

The results show, that the impact of US CDS spreads on other countries CDS spreads decreases holding the trade channel constant at different values (figure 6, panel a). However, the impact is only significant when the impact remains positive. Contrary to that, holding the exchange rate regime constant at different values, the impact of US CDS spreads on other countries CDS spreads is increasing (figure 6, panel b). Therefore, the effect of US CDS spreads is higher for countries with a floating exchange rate regime. Both findings are contrary to hypothesis 2 and hypothesis 4.

Additionally, the exchange rate regime of a country is substituted with the exchange rate of a country to the USD and reported in column 1 and 2 of table 6. The argument behind it is, that a country with a fixed exchange rate towards the USD is more affected by movements in the US than a country with a floating exchange rate. However, the results in table 6 (column 1 and 2) show that the coefficients of the exchange rate to the US remain insignificant. Only the interaction terms between US CDS spreads and the exchange rate regime are positive and significant, indicating that the impact of US CDS increases when the exchange rate is floating.

(34)

Table 5: Sensitivity analysis - different maturities Model 1 Model 2 CDS 1Y CDS 3Y CDS 1Y CDS 3Y US CDS 1Y 4.181*** 3.138*** (0.669) (0.635) US CDS 3Y 3.782*** 3.356*** (0.798) (0.839) Trade channel 10.72** 37.53*** 16.67*** 38.73*** (4.376) (6.584) (4.135) (6.607) Financial channel 171.3 62.53 -71.46 -41.57 (131.7) (204.0) (125.4) (212.1)

Ex. rate regime -15.88** 0.901

(7.331) (12.65) US CDS 1Y × Trade channel -0.918*** -0.826*** (0.244) (0.223) US CDS 3Y × Trade channel -0.788*** -0.745*** (0.289) (0.287) US CDS 1Y × Financial channel -2.134 -0.245 (2.632) (2.455) US CDS 3Y × Financial channel 2.163 2.458 (3.313) (3.407)

US CDS 1Y × Ex. rate regime 2.237***

(0.333)

US CDS 3Y × Ex. rate regime 0.727*

(0.421) Stock 93.44*** 112.6*** 76.22*** 100.9*** (19.49) (27.68) (18.07) (27.77) Reserves 12.43 -86.93** 17.73 -74.57** (24.43) (35.18) (23.48) (36.50) S&P 500 -167.7*** -168.7*** -115.8*** -146.3*** (26.90) (38.36) (25.55) (38.82) Private credit 0.0188 0.269 0.112 0.414** (0.124) (0.178) (0.129) (0.199) Constant -32.99 -117.5*** -50.89* -147.6*** (26.35) (38.07) (29.04) (46.41)

Adj. R-squared 0.314 0.273 0.426 0.287

(35)

Figure 6: Marginal effects conditional on trade channel and ex. rate regime

−2

0

2

4

1.5 1.75 2 2.25 2.5 2.75 3 3.25 3.5 3.75 4 4.25 4.5 4.75 5 5.25 5.5 Trade channel

(a) Marginal effects trade channel

2 4 6 8 10 12

1 1.5 2 2.5 3 3.5 4

Ex. rate regime

(b) Marginal effects ex. rate regime

Notes: Figure 6 reports the marginal effects of US CDS spreads on the CDS spreads of other countries holding the trade channel (panel A) and the exchange rate regime (panel B) constant at different values. The CDS contracts have a maturity of one year. The vertical lines indicate the 95% confidence interval. The effects are significant when the confidence intervals are above or below zero.

To control for omitted variables that are constant across countries but vary over time, time dummies are added to the estimations (time fixed effects). The results are reported in table 7 column 1 and 2 and show that the coefficients of US CDS spreads and the trade transmission channel remain positive and statistically significant. However, the interaction term between US CDS spreads and the financial transmission channel remains insignificant. The year dummies are tested for their joint significant and the null hypothesis is rejected. Therefore the coefficients have an impact on the estimations7_.

Lastly, to completely avoid the global financial and sovereign debt crisis, only the time period until quarter three 2007 is considered. The results are reported in column 3 and 4 of table 7 and show, that before the crisis period, the US CDS spreads did not have an influence on other countries CDS. However, the trade transmission channel is positive and significant and the financial transmission channel are both positive and significant. The interaction between US CDS spreads and the trade channel changes its sign from negative to positive and also the interaction coefficient of the financial transmission channel changes its coefficient from positive to negative (all coefficients are significant on the 5% and 1% level). Contrary to the findings above, when only a crisis dummy is included, the results in table 7 suggest that the crises have an

7_{The results are not reported in the main analysis because adding time dummy variables cause also a substantial}

(36)

influence on the fact of how movements in the US CDS market affect other CDS market around the world. Further, the interaction with the exchange rate regime remains insignificant.

Table 6: Sensitivity analysis - Ex. rate to US & dummy 2005/2006

Model ex. rate Dummy 2005/2006

(1) (2) (3) (4) US CDS 2.222*** 2.222*** 1.922*** 1.929** (0.758) (0.758) (0.700) (0.771) Trade channel 53.62*** 53.62*** 38.68*** 38.68*** (8.789) (8.789) (7.585) (7.658) Financial channel -251.0 -251.0 124.1 124.6 (262.2) (262.2) (251.2) (270.0)

Ex. rate regime 16.50

(16.23) Ex. rate to USD -0.162 -0.162

(0.940) (0.940)

US CDS × Trade channel -0.397 -0.397 -0.253 -0.228 (0.274) (0.274) (0.253) (0.254) US CDS × Financial channel 7.843** 7.843** 3.080 2.028

(3.407) (3.407) (3.341) (3.544)

US CDS × Ex. rate regime 0.0949

(0.411) US CDS × Ex. rate to USD 0.0157** 0.0157**

(0.00791) (0.00791) Stock 135.4*** 135.4*** 108.0*** 97.43*** (33.90) (33.90) (32.94) (33.26) Reserves -183.5*** -183.5*** -173.7*** -152.5*** (48.23) (48.23) (42.83) (44.56) S&P 500 -190.3*** -190.3*** -177.0*** -159.6*** (46.71) (46.71) (44.91) (45.84) Private credit 0.533** 0.533** 0.673*** 0.864*** (0.236) (0.236) (0.219) (0.242) Dummy 2005/2006 61.09*** 61.78*** (11.18) (11.15) Constant -146.4*** -146.4*** -171.0*** -224.6*** (48.00) (48.00) (45.34) (56.80)

Adj. R-squared 0.276 0.276 0.329 0.334

(37)

Table 7: Sensitivity analysis - Time FE & without crises period

Model time FE Without crises period

(1) (2) (3) (4) US CDS 1.910** 2.141** -12.15 -2.205 (0.827) (0.906) (8.370) (10.68) Trade channel 38.21*** 34.72*** 75.83*** 44.98** (7.962) (8.450) (18.17) (20.71) Financial channel 39.18 205.2 2,247*** 6,211*** (266.0) (316.5) (718.0) (1,589)

Ex. rate regime 23.04 285.6***

(20.23) (105.6)

US CDS × Trade channel -0.150 -0.226 5.730** 6.788*** (0.262) (0.266) (2.211) (2.229) US CDS × Financial channel 2.100 1.382 -185.0*** -361.9***

(3.453) (3.682) (52.03) (82.27)

US CDS × Ex. rate regime 0.0518 -11.23

(0.460) (11.77) Stock 100.5*** 93.38*** 100.8 82.13 (34.23) (34.51) (76.52) (74.16) Reserves -168.0*** -155.9*** -118.8 -117.9 (45.51) (46.13) (84.52) (82.61) S&P 500 -136.9*** -135.0*** -100.3 -105.5 (50.41) (50.58) (240.5) (233.7) Private credit 1.006*** 0.919*** 3.175*** 3.559*** (0.340) (0.346) (1.030) (1.011) Constant -131.8* -131.3* -756.9*** -1,087*** (71.25) (71.32) (205.6) (234.8)

Time FE Yes Yes No No

Adj. R-squared 0.357 0.358 0.449 0.489

Notes:Table 7 reports the results with time fixed effects (year FE) in column 1 and 2. In column 3 and 4 the results excluding the crises period from quarter 3 onwards is reported. The sample period in column 1 and 2 is from quarter one 2004 to quarter three 2015. The underlying CDS contracts have a maturity of five years. ***;**;* denote significance at the 1%,5% or 10% level, respectively. Clustered standard errors are reported in parenthesis.

(38)

6 Discussion and Limitations

The results provide evidence that changes in US credit risk influences the credit risk of other countries (see section 4.2 for more details). Further, in line with related studies (for example, Van Rijckeghem and Weder (2001); Kaminsky and Reinhart (2003)), the findings suggest that in particular the financial transmission channel plays a role how severe changes in US CDS spreads are transmitted to the CDS spreads of other countries. In contrast to that, no evidence can be found whether the trade transmission channel or the exchange rate regime also have an influence.

The reason for this results can be manifold. Looking at the functioning scheme of CDS contracts (see section 2.1), seller and buyer of CDS products are primarily banks and other financial institutions. Given the fact that the financial transmission channel reflects the bilateral bank claims (see section 3.5), those information might be easier priced into CDS spreads. Therefore, different to stock and bond markets (as shown by Forbes and Chinn (2004)), on the CDS market more the financial linkages rather than the trade transmission channel play a role how changes in US credit risk are transmitted to the CDS market of countries.

Interestingly, excluding the crises period and only investigating the time before the financial crisis leads to different results. As reported in table 7, the coefficient of the interaction between US CDS spreads and the trade channel is significant and positive while the interaction term with the financial channel changes its sign from positive to negative. This suggests that the results are not only sensitive to the crises periods in the sample, but also that the crises changed the factors that influence how severe changes in the US CDS market affect the credit risk of other countries.

Another interesting finding is that the driving factors of CDS contracts with a shorter maturity are different from those with a maturity of five years. Looking at the marginal effects reported in table 6 (panel A), the impact that changes of US CDS spreads have on CDS markets world wide is decreasing conditional on an increasing trade channel. A possible explanation for these findings could be that investors see the bilateral trade relations in the short run as a positive factor for the economy. Consequently, the bilateral trade relations reduce the impact movements in the US CDS market have on the CDS spreads of other countries.

(39)

term is positive and significant. One reason that these findings are contrary to hypothesis four and also contrary to related studies such as Liu and Morley (2012), might simply be due to the fact that the sample is not large enough and the majority of the sample countries have a floating exchange rate, indicated by the high median of the exchange rate regime variable (see table 8). Next to the reasons mentioned above, also the underlying market and therefore the source of CDS movements around the world might be a reason that this study could only find evidence for two of four hypothesis. By interacting the international transmission channel with CDS spreads of the US, it is investigated whether the channels have an influence on the severity market movements in the US CDS market have on the CDS market of other countries. However, as shown in table 3 and in line with Longstaff et al. (2007), also the S&P 500 seems to have a large impact on the CDS market of the countries across the sample. Therefore, it is not only the credit risk of the US but also the state of the US economy that is linked with changing CDS spreads of other countries. Consequently, a question that is left open for further research is, whether the international transmission channels have an effect on the severity of how the real economic movements in the US affect the credit risk of other countries.

In addition to that, further research could extend the study and investigate a larger sample. Not only the US can act as a center country, but also other countries might have large local influences. Therefore, a more comprehensive study could analyze the bilateral relations, for example, in Asia, with China or Japan as the center country and investigate their impacts on sovereign CDS markets.

As any empirical study, this thesis has some limitations. First of all, the results in this study could be biased due to omitted variables that are not considered in the specification but have an influence on the results. Further, the setup of the analysis does not allow to distinguish between advanced and emerging economies. As shown by Balakrishnan et al. (2011), emerging economies might be more dependent on other large economies than the advanced countries in the sample. Moreover, the sample period is rather short and includes too many crises episodes that drive the results.

(40)

consequence is a correlation of the estimates with the error term and a so called endogeneity problem. To account for that, future research could make use of instrumental variables or other techniques that allow to control for it. However, finding the right instruments is rather difficult.

7 Conclusion

Credit derivatives like sovereign CDS contracts give investors the possibility to insure their portfolio against macroeconomic credit risk and are also an indicator of a countries credit condition. Consequently, the sovereign CDS market and its driving factors are of growing interest among researchers. This thesis adds to the literature by investigating whether the credit risk of the US is driving the credit risk of other economies. Further, it aims to analyze whether the international transmission channels have an effect on how severe changing US CDS spreads are transmitted to CDS markets of other countries.

The results suggest, that the credit risk of the US is associated with the credit risk of countries around the world. Further, this thesis finds, that a larger financial transmission channel intensifies this effect, while a larger trade transmission channel or the exchange rate regime do not have an effect how severe movements in the US CDS market are transmitted to CDS market of other countries. The direct impact US CDS spreads have on other countries CDS spreads is not sensitive to modifications. However, the influence of US CDS spreads conditional on the financial transmission channel is sensitive to crisis related specifications such as including time fixed effects or excluding the crises periods.

(41)

Appendix A

Variable description

Table 8: Variable description and sources

Variable Definition Expected

Sign Source Trade channel Trade channel proxy:

+

Datastream

Financial channel Financial channel proxy:

+

BIS

Private credit Total credit to the non-financial sector as %

of GDP +

BIS Public Debt Public debt as % of GDP

+

Datastream

Exchangerate regime

Coarse exchange rate classification on a 6-point scale, where 1 represents a pegged and 4 a freely floating exchange rate.

-Reinhart and Rogoff (2004)

Stock Returns of the leading stock indices per

country

-Datastream Reserves Foreign exchange reserves/GDP - Datastream

SP500 Returns S&P 500 - Datastream

(42)

Appendix B

Correlations

Table 9: Correlation matrix

1

2

3

4

5

6

7

8

9 1 US CDS

1 2 Trade channel

-0.17

1 3 Financial channel 0.12

0.4

1 4 Reserves

0.05 0.18 -0.26

1 5 Stock

-0.17 0.03 -0.06 0.03

1 6 Public debt

0.14 0.21 0.37 0.47 -0.01

1 7 Private credit

0.13 0.32 0.56 0.18 -0.05 0.80

1 8 SP500

-0.14 -0.02 -0.01 0.01 0.54 0.04 0.04

1 9 Ex. rate regime

-0.03 0.11 -0.12 0.62 0.04 -0.02 0.000

0

1

(43)

Appendix C

Stylized facts

Figure 7: Bilateral trade relation to the US relative to a countries GDP

1 2 3 4 5 6

Bil. trade to US/GDP

2004q1 2006q1 2008q1 2010q1 2012q1 2014q1 Time

Australia Brazil

France Germany

Italy Japan

Turkey United Kingdom

(44)

Figure 8: Bilateral bank claims in the US relative to a countries GDP

0

.1

.2

.3

Bil. bank claims in US/GDP

2004q1 2006q1 2008q1 2010q1 2012q1 2014q1 Time

Australia Brazil

France Germany

Italy Japan

Turkey United Kingdom

(45)

Figure 9: Overview CDS spreads 0 100 200 2004q1 2007q1 2010q1 2013q1 2016q1 date

Australia

0 300 600 2004q1 2007q1 2010q1 2013q1 2016q1 date

Brazil

0 100 200 2004q1 2007q1 2010q1 2013q1 2016q1 date

France

0 100 200 2004q1 2007q1 2010q1 2013q1 2016q1 date

Germany

0 200 400 2004q1 2007q1 2010q1 2013q1 2016q1 date

Italy

0 100 200 2004q1 2007q1 2010q1 2013q1 2016q1 date

Japan

0 200 400 2004q1 2007q1 2010q1 2013q1 2016q1 date

Turkey

0 100 200 2004q1 2007q1 2010q1 2013q1 2016q1 date

United Kingdom

Country

Average

US

(46)

References

AIMA (2011). The European Sovereign CDS Market. Alternative Investment Management Association Research Note.

Arezki, R., B. Candelon, and A. N. R. Sy (2011). Sovereign rating news and financial markets spillovers: Evidence from the European debt crisis. IMF working papers, 1–27.

Balakrishnan, R., S. Danninger, S. Elekdag, and I. Tytell (2011). The transmission of financial stress from advanced to emerging economies. Emerging Markets Finance and Trade 47(sup2), 40–68.

Baldwin, R. (2006). Globalisation: the great unbundling (s). Economic Council of Finland 20(3), 5–47.

Billor, N., A. S. Hadi, and P. F. Velleman (2000). BACON: blocked adaptive computationally efficient outlier nominators. Computational Statistics & Data Analysis 34(3), 279–298. BIS (2015). OTC, credit default swaps, by sector of reference. Bank for International Settlements. Bomfim, A. N. (2005). Understanding credit derivatives and related instruments. Academic

Press.

Bruno, V. and H. S. Shin (2015). Capital flows and the risk-taking channel of monetary policy. Journal of Monetary Economics 71, 119–132.

Calice, G., J. Chen, and J. Williams (2013). Liquidity spillovers in sovereign bond and CDS markets: An analysis of the Eurozone sovereign debt crisis. Journal of Economic Behavior & Organization 85, 122–143.

Caramazza, F., L. A. Ricci, and R. Salgado (2000). Trade and financial contagion in currency crises. IMF working papers.

What drives sovereign credit default swaps? The impact of US credit risk and international transmission channels on sovereign CDS