Sovereign credit risk and exchange rates Evidence from CDS quanto spreads

Transcription

1 Sovereign credit risk and exchange rates Evidence from CDS quanto spreads Patrick Augustin McGill University-Desautels Faculty of Management Mikhail Chernov Anderson School of Management, UCLA, NBER, and CEPR Dongho Song Johns Hopkins University 218 Conference on Derivatives and Volatility Chicago, IL November 9, 218

2 The Twin Ds and Quantos Sovereign Default and Devaluation go hand in hand Measurement presents difficulty P(Dev Def) = 84% (Reinhart, 22) P(Dev Def) = 48% (Na, Schmitt-Grohe, Uribe, Yue, 217) Long-lasting effect of default on currency? Krugman (1976): a change in expected depreciation rate Na, Schmitt-Grohe, Uribe, Yue (217): one-off drop A new development in derivatives markets allows for a view through financial lens Since 21, Sovereign CDS contracts are available in multiple currencies Quanto CDS spread reflects a premium for the interaction between the twin Ds Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 2 / 2

3 What we do Explain how Quanto CDS spreads are related to the twin Ds Eurozone quantos are particularly interesting: Reflect same FX & monetary policy, but term structures of quanto spreads are different Institutional setting makes them analysis-friendly 2. Develop a no-arbitrage affine term structure model 3. Identify credit risk (true distribution of credit events) and LGD 4. Measure probabilities of twin Ds (4% true; 75% risk-adjusted) risk premium for exposure to the twin Ds (up to.3% per week) Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 3 / 2

4 $1 $C $(1 L) $(1 + C ) Sovereign CDS in different currencies Swap transaction Swap $S t per e1 (S t e devaluation) transaction Time of default τ Loss given default L $1 e1/s ef /S $C ef /S $C e(1 + F )/S $(1 + C ) $1 e1/s ef /S $C ef /S $C e(1 + F )/S $(1 + C ) 2 Start, During the term, t Credit event, τ Maturity, T e denominated CDS ebond denominated Notional: CDS transaction $1= e1/s ec e /S el/s ec e /S Notional: $1= e1/s $1 ec e $C /S $(1 L) el/s $(1 ec+ e /S C ) $ denominated Swap CDS $ denominated transaction Notional: CDS $1 $C $ = ec $1 e1/s ef /S $C ef /S $ t $L = el/s τ $C $ = ec $ Notional: $1 /S $C e(1 + F )/S $(1 + C ) /S T $C $ = ec /S $ t $L = el/s τ $C $ = ec /S $ T Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 4 / 2

5 The economic role of Sovereign CDS This is not a redundant security One might think of replicating with bond and a currency swap... Start, Start, During the term, During the term, t Credit event, Credit event, τ Start, During the term, t Credit event, τ 2 2 Maturity, Maturity, T Maturity, T Bond Bond Bond transaction transaction transaction $1 $1 $1 $C $C $C $(1 L) $(1 $(1 L) $(1 + C ) $(1 L) $(1 + C ) Swap Swap Swap transaction transaction transaction $1 $1 e1/s e1/s ef /S ef /S $C $C ef /S ef /S $C $C e(1 )/S e(1 + F )/S $(1 $(1 + C ) $1 e1/s ef /S $C ef /S $C e(1 + F )/S $(1 + C ) Augustin, e denominated Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 5 / 2

6 The economic role of Sovereign CDS This is not a redundant security One might think of replicating with bond and a currency swap... Start, Start, During the term, During the term, t Credit event, Credit event, τ Start, During the term, t Credit event, τ 2 2 Maturity, Maturity, T Maturity, T Bond Bond Bond transaction transaction transaction $1 $1 $1 $C $C $C $(1 L) $(1 $(1 L) $(1 + C ) $(1 L) $(1 + C ) Swap Swap Swap transaction transaction transaction $1 $1 e1/s e1/s ef /S ef /S $C $C ef /S ef /S $C $C e(1 )/S e(1 + F )/S $(1 $(1 + C ) $1 e1/s ef /S $C ef /S $C e(1 + F )/S $(1 + C ) Augustin, e denominated Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 5 / 2

7 Preliminary look at the data German 5-year CDS premiums denominated in USD Cross-correlation German 5-year Quanto spreads Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 6 / 2

8 Interepretation of a quanto spread Consider a contract trading on all-points-upfront basis C $ t,t = L E t [e r(τ T t) I (τ T )] C e t,t = L E t [e r(τ T t) I (τ T ) S τ T /S t ] The relative quanto spread is C $ t,t C e t,t C $ t,t = E t [ 1 S τ T S t ] [ covt e r(τ T t) I (τ T ) Et e r(τ T t) I (τ T ), S τ T S t ] e devaluation corresponds to an increase in the first term Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 7 / 2

9 Term structures of quanto spreads y 3y 5y 7y 1y 15y Germany Belgium France Ireland Italy Spain Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 8 / 2

10 A no-arbitrage alone term structure model Joint model for CDS and FX: multiple building blocks credit risk factors (1 global and 2 regional factors: core and periphery ) Contagion [omitted in this presentation] Depreciation rate Interest/forward rate factors Key assumption: FX jumps only at credit events s t+1 = s + δs x t+1 + ( v + δ v v t ) 1/2 ε s,t+1 z s,t+1 Implications: same probability of devaluation vs different probabilities of default can identify actual occurrence of credit events via FX time series can identify Loss given default L, [35% 5%] Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 9 / 2

11 Setup I: state The state x t (under risk-adjusted probability) x t+1 = µ x + Φx x t + Σx,t ε x,t+1 consists of three different sub-vectors x t = (ut, v t, wt ), where u t is used for modeling interest and forward rates [not covered today] v t is used for modeling variance of the exchange rate w t is used for modeling default intensities Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 1 / 2

12 Setup II: credit risk The default hazard rate of each country k is where H k t P (τ k = t τ k t; F t ) = 1 e h k t, h k t is the default (credit event) intensity = h k + δw k w t Credit factor w t consists of 1 global & 2 regional factors: core and periphery each ht k is a function of one of the regional factors overidentifying restriction: Germany loads on the global credit factor, w 1t only Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 11 / 2

13 Setup III: exchange rate The (log) depreciation rate: s t+1 = s + δ s x t+1 + ( v + δ v v t ) 1/2 ε s,t+1 z s,t+1, Jump z s,t+1 is connected to sovereign defaults via Jump intensity λt = ht k k Conditionally on the number of jumps, the jump magnitude is z t+1 j t+1 Γ (j t+1, θ ) One-sided jump magnitude reflects EUR depreciation upon any country s default The setup allows to speak to the Krugman vs Na et al. discussion Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 12 / 2

14 Setup IV: Loss given default and default risk premiums 1. Identification of Loss given default USD CDS premiums depend on L h k t FX forward rates and EUR CDS premiums depend, via FX, on λ t = k h k t Can infer L, [35% 5%] 2. Identification of default risk premium Jumps in FX and credit events depend on λ t Identification of credit events: extreme changes in quanto spreads Can infer default risk premium, λ t /λ t Credit risk premium consistent with previous corporate studies ( 2) Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 13 / 2

15 Credit events Greek CDS premium Greek quanto spread Large changes in quanto spreads 1 Austria Belgium Cyprus Estonia 1 Finland France Germany Ireland Italy Latvia Lithuania Netherlands Portugal Slovakia Slovenia Spain Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 14 / 2

16 2 Time series of default intensities and credit risk premium 1.5 Germany Belgium France Ireland Italy Spain Germany Belgium France Ireland Italy Spain Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 15 / 2

17 Cross section of default intensities Factor loadings (selected countries 8/17) Country, k Global, δw1 k Core, δw2 k Periphery, δw3 k Spain Greece Austria Slovakia Germany Lithuania France Finland Fixed effects panel regressions of default intensities on debt/gdp explain 44% of cross-country variation imply 1% debt/gdp leads to an 8 bps in CDS premium Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 16 / 2

18 The Twin Ds The (log) depreciation rate is empirically: s t+1 = s + δs6 v t+1 + ( v + δ v v t ) 1/2 ε s,t+1 z s,t+1, Back to the Krugman vs Na et al. discussion: Conditional mean does not depend on credit factors Short-term effect Estimated δ s6 <, so $ appreciates gradually even without defaults Define devaluation as a three sigma move downwards P(Devaluation Default) = 4.14% (over 1 year it is.3%) P (Devaluation Default) = 75.43% (over 1 year it is.91%) This suggests the occurrence of devaluations during particularly bad states of the economy that tend to coincide with high marginal utility of representative agent Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 17 / 2

19 Expectations of the depreciation rate, S T /S t without default w 3m 1y 3y 5y 7y 9y 11y 13y 15y with default w 3m 1y 3y 5y 7y 9y 11y 13y 15y Difference reflects extra compensation for currency loss in case of default can be up to 3 bps per week at long horizons Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 18 / 2

20 Components of relative Quanto CDS C e t,t C $ t,t = E t [ ] [ ] Sτ T + covt e r(τ T t) I (τ T ) S t Et e r(τ T t) I (τ T ), S τ T S t Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 19 / 2

21 Conclusion New financial instruments lead to new evidence on the Twin Ds We develop a no-arbitrage model that captures joint dynamics of sovereign CDS quanto spreads in the 17 Eurozone countries the USD/EUR exchange rate Identification of credit risk premium and LGD Cross-country variation in default intensities linked to fiscal policy Twin Ds probability of devaluation conditional on default is low risk premium is large much larger than the credit risk premium economically important as coincide with high marginal utility states Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 2 / 2

22 Thank You! Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 21 / 2

23 Affine Term Structure Models of Sovereign Credit Spreads Study TS Model Default Intensity FX Factors Focus Components Components FX Quanto Homoscedastic Heteroscedastic Contagion Homoscedastic Heteroscedastic Extr. Evts Monfort and Renne (214) Duffie et al. (23) Hoerdahl and Tristani (212) Pan and Singleton (28) Longstaff et al. (211) Doshi et al. (217) Zhang (28) Ang and Longstaff (213) Aït-Sahalia et al. (214) + Benzoni et al. (215) Carr and Wu (27) Ehlers and Schönbucher (24) Brigo et al. (216) Monfort et al. (217) The present study Country Common Region Joint Est. Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 22 / 2

24 German 5-year CDS premiums Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 23 / 2

25 Descriptive Statistics of CDS Quanto Spreads Markit, weekly (Wednesday) Spreads, CR restructuring clause Average 5-y Quanto spreads range from 5bps - 9bps Term structure of Quanto spreads range from -29bps - 29bps Maturity 1y 3y 5y 7y 1y 15y 3y 1y-1y Country Obs Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD AT BE CY EE FI FR DE GR IE IT LV LT NL PT SK SI ES Total 5, Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 24 / 2

26 Liquidity stats USD CDS contracts are more liquid than the EUR ones: average difference between quoting dealers:.6 to 2.7 Gross notionals are the largest for France (38), Germany (34), and Spain (48) [percentages of Italy] Sovereign CDS represent about 13% of the aggregate CDS market (including corporate single name and index) in terms of gross notionals Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 25 / 2

27 Descriptive Statistics of CDS Liquidity USD CDS more liquid than EUR CDS: average diff. betw. quoting dealers: Italy largest market in gross notionals, followed by France, Germany, and Spain Sov. single name CDS represent 18% of market in 216 (BIS, gross notionals) EUR USD USD Notionals N Mean Min p5 Max Mean Min p5 Max Gross Gross% Net Net% Austria Belgium Cyprus Estonia Finland France Germany Greece Ireland Italy Latvia Lithuania Netherlands Portugal Slovakia Slovenia Spain All 5, Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 26 / 2

28 Preliminary look at the data German 5-year CDS premiums denominated in USD Cross-correlation German 5-year Quanto spreads Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 27 / 2

29 Contagion Contagion: a credit event affects probability of subsequent events Different from co-moving probabilities of events Nicoletta Kotsianas (FT, 4/21): The fact the [quanto CDS spread] has widened out for Germany a sovereign with very little default risk illustrates that the market is pricing in increased contagion issues for the currency and the fears are seeping into all credit related products that settle in euros, noted the CDS traders. John Cochrane (WSJ, 12/21): The bailout is being justified on grounds of containing contagion. This is nonsense. The notion is that news of an Irish restructuring would scare investors in Spanish bonds... But haven t investors in Spanish bonds already noticed that there s a bit of a problem? Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 28 / 2

30 Modeling contagion Credit events P arrive at a rate d t [prob(p = ) = e d t ] How does P affect probability of a credit event in the next period? d t+1 Exp(P) The larger is P, the higher is the event probability in the next period. By convention, P = d t+1 = This is the ARG process of Monfort, Pegoraro, Renne, and Roussellet (214) Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 29 / 2

31 Credit events Greek CDS premium Greek quanto spread Large changes in quanto spreads 1 Austria Belgium Cyprus Estonia 1 Finland France Germany Ireland Italy Latvia Lithuania Netherlands Portugal Slovakia Slovenia Spain Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 3 / 2

32 Implementation Data summary: , weekly frequency Quanto spreads for each of the 6 countries: 1, 3, 5, 7, 1, and 15 years OIS yield curve:.5, 1, 3, 5, 7, 1, and 15 years Time series of the USD/EUR depreciation rate Time series of credit events Estimation method: Bayesian MCMC with particle filtering Overidentifying restrictions: Germany loads on the global credit factor, w 1t only; other countries load on w 1t and on either w 2t (core), or w 3t (periphery) Contagion: weak statistical support Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 31 / 2

33 Quanto spreads: Core Germany Belgium France Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 32 / 2

34 Quanto spreads: Periphery Ireland Italy Spain Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 33 / 2

35 Credit factors Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 34 / 2

36 Cross section of default intensities, loadings Core, δ k w2 Periphery, δ k w3 Country, k Global, δw1 k Spain Italy Cyprus Greece Slovenia Ireland Austria Slovakia Germany Netherlands Lithuania Latvia Portugal Belgium Estonia France Finland Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 35 / 2

37 Expectations of the depreciation rate without default w 3m 1y 3y 5y 7y 9y 11y 13y 15y with default w 3m 1y 3y 5y 7y 9y 11y 13y 15y expected excess return difference w 3m 1y 3y 5y 7y 9y 11y 13y 15y Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 36 / 2

38 Relative quanto spread and depreciation rate: Core Germany y 3y 5y 7y 1y 15y Belgium y 3y 5y 7y 1y 15y France y 3y 5y 7y 1y 15y Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 37 / 2

39 Relative quanto spread and depreciation rate: Periphery Ireland y 3y 5y 7y 1y 15y Italy y 3y 5y 7y 1y 15y Spain y 3y 5y 7y 1y 15y Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 38 / 2

40 Time series of OIS interest rates - bootstrapped ZC rates 4 3 6m 1y 3y 5y 7y 1y 15y Back to Implementation Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 39 / 2

41 Model Comparison With contagion Without contagion 5% 5% 95% 5% 5% 95% ln p(y Θ) BIC Back to Implementation Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 4 / 2

42 Model Fit Bond yield Quanto spread Germany Belgium France Ireland Italy Portugal Avg across countries 1y y y y y y Avg across maturity Back to Results Augustin, Chernov & Song Sovereign credit risk and exchange rates: Evidence from CDS quanto spreads 41 / 2