Upload
vutu
View
224
Download
0
Embed Size (px)
Citation preview
Measuring and Dynamically Hedging Counterparty Credit Exposure and Risk
PRESENTED TO: Financial Engineering SeminarDepartment of Industrial Engineering and Operations Research.Columbia University
BY: Evan Picoult, Managing DirectorRisk ArchitectureCitigroupNew York, New York
DATE: Monday, January 24th, 2005
PLACE: New York City
o COUNTERPARTY RISK - Counterparty of a forward or derivative defaults prior to finalsettlement of cash flows and the contract (portfolio) has a positive economic value.
BASIC DEFINITIONS
o LENDING RISK - Borrower defaults - an accrual (non MTM) perspective
- Or the loan portfolio’s economic value also decreases because of:- decrease in the credit quality of the obligor and/or- increase in general market spreads.
o ISSUER RISK (Specific Risk) - Issuer of security defaults.
- Or the security’s market value also decreases because of:- decrease in the credit quality of issuer and/or- increase in general market spreads.
o SETTLEMENT RISK - In an exchange: you pay but do not receive what you are owed.
TYPE OF CREDIT RISK CAUSE OF ECONOMIC LOSS
- Or the economic value of derivatives with counterpartyalso decreases because of:
- decrease in credit quality of counterparty and/or- increase in general market spreads
How to measure this?
a.k.a. Pre-settlement risk
Types of Credit Risk
Page 2E 2Evan Picoult, Citigroup January, 2005
Page 3E 3Evan Picoult, Citigroup January, 2005
Contents
• PORTFOLIO SIMULATION OF A COUNTERPARTY’S EXPOSURE PROFILE.
• ECONOMIC CAPITAL FOR LOAN CREDIT RISK– Default only perspective.– Loss of economic value perspective.
• ECONOMIC CAPITAL FOR COUNTERPARTY RISK - DEFAULT ONLY:– Full coherent simulation of potential exposure and default.– Approximation using incoherent simulation with expected positive exposure profile
scaled up by factor α .
• ECONOMIC CAPITAL FOR COUNTERPARTY RISK - ECONOMIC LOSS:– Defining the Credit Value Adjustment (CVA) for credit risk of counterparty’s portfolio. – Simulating default, recovery and changes in CVA over time.
• DYNAMICALLY HEDGING COUNTERPARTY RISK
BASIC QUESTION: WHAT ARE THE CONSEQUENCES OF CREDIT EXPOSURE DEPENDING ON THE POTENTIAL FUTURE STATE OF MARKET RATES?
MEASURINGCOUNTERPARTY CREDIT EXPOSURE
FORMS OF CREDIT RISK
Page 4E 4Evan Picoult, Citigroup January, 2005
Potential Exposure Of A Single Transaction
EXPOSURE PROFILE OF SINGLE TRANSACTION AND MARKET RATE SCENARIOS
Example 1: Forward FX, We buy GBP and sell US$ for settlement in two years at 1.5000 US$/GBP.
Random path of forward FX rate for a fixed settlement date, over life of forward transaction in scenario 1.
Profile of market value of forward FX transaction over its life, for scenario 1.
Exposure Profile of transaction for scenario 1.
We only have exposure when the contract has a positive value to us.
Random Scenario 1 for Forward FX Rate
1.250
1.375
1.500
1.625
1.750
0 3 6 9 12 15 18 21 24Time (months)
Forw
ard
Exc
hang
e R
ate
Forward FX Replacement Cost for Scenario 1
-20%
-15%
-10%
-5%
0%
5%
10%
15%
20%
0 3 6 9 12 15 18 21 24Time (Months)
Rep
lace
men
t C
ost
(% N
otio
nal)
Forward FX Exposure Under Scenario 1
0%
5%
10%
15%
20%
0 3 6 9 12 15 18 21 24
Time (months)
Pote
ntia
l Ex
posu
re (
% N
otio
nal)
Page 5E 5Evan Picoult, Citigroup January, 2005
Potential Exposure Of A Single TransactionEXPOSURE PROFILES AND MARKET RATE SCENARIOS Example 2: I.R. Swap, We pay fixed and receive floating every 6 months for three years.
Two Random scenariosof the path of six month LIBOR over the 36 month life of swap.
Profile of market valueof swap over its life for each scenario.
Exposure Profile of swap over its life for each scenario .
Scenario 2: Random change in 6 month LIBOR
5%
6%
7%
8%
9%
10%
11%
0 6 12 18 24 30 36Time (months)
Inte
rest
Rat
e
Scenario 2 : Profile of IR Swap Value
-2%-1%0%1%2%3%4%5%6%7%
0 6 12 18 24 30 36
Time (Months)
Pote
ntia
l Ex
posu
re (
% N
otio
nal)
Scenario 2: IR Swap Exposure Profile
0%
1%
2%
3%
4%
5%
6%
7%
0 6 12 18 24 30 36
Time (Months)
Pote
ntia
l Ex
posu
re (
% N
otio
nal)
Scenario 3 : Profile of IR Swap Value
-2%-1%0%1%2%3%4%5%6%7%
0 6 12 18 24 30 36
Time (Months)
Pote
ntia
l Ex
posu
re (
% N
otio
nal)
Scenario 3: IR Swap Exposure Profile
0%
1%
2%
3%
4%
5%
6%
7%
0 6 12 18 24 30 36
Time (Months)
Pote
ntia
l Ex
posu
re (
% N
otio
nal)
Scenario 3: Random change in 6 month LIBOR
5%
6%
7%
8%
9%
10%
11%
0 6 12 18 24 30 36Time (months)
Inte
rest
Rat
e
Page 6E 6Evan Picoult, Citigroup January, 2005
Potential Exposure Of A Single Transaction
Three Exposure Profiles for a two year US$/GBP forward FX transaction. At threeconfidence levels:
- 99% CL Exposure Profile
- 97.7% CL Exposure Profile
- Expected Positive Exposure Profile
99.0% CL Profile.
97.7% CL Profile.
Expected Profile
Forward FX Exposure Profiles at Three Confidence Levels
0%
10%
20%
30%
40%
50%
60%
0 3 6 9 12 15 18 21 24
Time (Months)
Expo
sure
Pro
file
(% N
otio
nal)
Three Exposure Profiles for a three year fixed/floating US$ interest rate swap. At three confidence levels:
- 99% CL Exposure Profile
- 97.7% CL Exposure Profile
- Expected Positive Exposure Profile99.0% CL Profile.
97.7% CL Profile.
Expected Profile
Int. Rate Swap Exposure Profiles at Three Confidence Levels
0%
1%
2%
3%
4%
5%
6%
7%
0 6 12 18 24 30 36
Time (Months)
Expo
sure
Pro
file
(% N
otio
nal)
On the basis of thousands of such simulations we can represent the potential exposure over time statistically, at different confidence levels:
Statistical Picture Of Potential Exposure
Page 7E 7Evan Picoult, Citigroup January, 2005
Potential exposure for a counterparty with multiple transactionsTWO METHODS FOR MEASURING COUNTERPARTY EXPOSURE (CE):
SIMPLE “ADD-ON” METHOD
CE TRANSACTION = CURRENT MTM + “WORST CASE” POTENTIAL INCREASE IN VALUE
= CURRENT MTM + NOTIONAL PRIN. * CREDIT EXPOSURE FACTOR
CE CP PORTFOLIO = Σ CE TRANSACTION
PORTFOLIO SIMULATION METHOD
CE CP PORTFOLIO = THE EXPOSURE PROFILE OF COUNTERPARTY
COUNTERPARTY EXPOSURE PROFILE
0
25
50
75
100
125
150
0 6 12 18 24 30 36 42 48 54 60
TIME (months)
POTE
NTI
AL
REP
LAC
EMEN
T C
OST
($
mm
)
Potential increase in value per unit of notional principal.
Potential exposure to a counterparty, at a high C.L., over lifetime of transactions with counterparty.
Assumes:
- No additional transactions
- Contractual cash flows set and settle over time.
- All legally enforceable riskmitigant agreements are taken into account.
Page 8E 8Evan Picoult, Citigroup January, 2005
MEASURINGCOUNTERPARTY CREDIT EXPOSURE
PORTFOLIO SIMULATION METHOD
Page 9E 9Evan Picoult, Citigroup January, 2005
Counterparty Exposure Portfolio Simulation
O DETAILED CONTRACT TERMS AND CONDITIONS.
CreditAdmin O TABLES OF LEGAL AGREEMENTS
- NETTING- MARGIN
O COLLATERAL
CreditAdmin O TABLES OF DEFAULT TRANSACTION PROFILES
FX FX DEBT I.R. EQ. COMM. COLLATERAL PRODUCT PROCESSOROPT SEC. DER. DER. DER. SYSTEM SYSTEMS
Detailed T&Cof Transaction
COUNTERPARTYCREDIT DATA BASE
COUNTERPARTY’S:- TRANSACTION DETAILS.- RISK MITIGANT DATA.
COUNTERPARTY’S:EXPOSURE PROFILE.
CE SERVER(analytical engine)
ANALYTICALENGINE
TABLES OF HISTORICAL VOLATILITIES AND CORRELATIONS
DAILY FEEDS OF CURRENT MARKET DATA MARKET DATA
Page 10E 10Evan Picoult, Citigroup January, 2005
General Method To Measure Counterparty’s Exposure Profile:
1) SIMULATE A PATH, P, OF MARKET RATES OVER TIME, M(t)P
- Start with current market rates.- Simulate a scenario (or path) of market rates at many future dates, over many years,
using tables of volatilities and correlations.
2) FOR SIMULATED PATH, P, MEASURE THE POTENTIAL MARKET VALUE OVER TIME OF EACH TRANSACTION WITH COUNTERPARTY K.
- Start with feed of transaction details and legal information.- For each simulated scenario, calculate the potential market value of each contract
at many future dates, using the contract’s terms and conditions, revaluation formula and the simulated state of the market.
- For each simulated scenario, at each future point in time, transform thepotential market value of each contract into the potential exposure of the portfolio through aggregation rules that take risk mitigants and legal context into account.
- i.e. For the counterparty K, for path M(t)P derive Exposure(t)K,P
3) THEN FOR SIMLUATED PATH, P, DERIVE COUNTERPARTY K’S POTENTIAL EXPOSURE OVERTIME
Loop
ove
r tho
usan
ds o
f pat
hs P
.
4) AFTER SIMULATING THOUSANDS OF POTENTIAL PATHS OF MARKET RATES, M(t)PCALCULATE EXPOSURE PROFILE OF COUNTERPARTY: THE POTENTIAL EXPOSURE AT SOME HIGH CONFIDENCE LEVEL, AT A SET OF FORWARD DATES
Page 11E 11Evan Picoult, Citigroup January, 2005
The Counterparty’s Exposure Profile:
THE EXPOSURE PROFILE EXPOSURE PROFILEPotential replacement cost of portfolio of contracts, over time, calculated at some confidence level, assuming:- no additional transactions- netting and margin taken
into account
COUNTERPARTY EXPOSURE PROFILE
0
25
50
75
100
125
150
0 6 12 18 24 30 36 42 48 54 60
TIME (months)
POTE
NTI
AL
REP
LAC
EMEN
T C
OST
($
mm
)
Exposure Profile LimitExposure Profile for credit line limits is typically calculated at a very high CL (e.g. 97.7%).
Exposure profile can be calculated at other CL, including expected positive profile or even a negative profile – how much one’s firm may owe counterparty in the future.
Page 12E 12Evan Picoult, Citigroup January, 2005
ECONOMIC CAPITAL
MEASUREMENT ISSUES FOR LOAN PORTFOLIO
Evan Picoult, Citigroup January, 2005 Page 13 Page 13
EC Definition• Economic Capital (also called “Economic Risk Capital” or “Risk Capital”) is a
measure of risk.
• Risk in this context means the potential unexpected loss of economic value over one year, calculated at a very high confidence level (99.97% CL).
• Thus EC measures risk from an insolvency or debt holders perspective (potential loss of value) rather than from an equity investment perspective (undiversified volatility of returns).
• Here is an example of EC for a loan portfolio:
Probability Distribution of Potential Credit Loss for a Portfolio of Many Obligors
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
0-20-40-60-80-100-120-140-160Potential Credit Loss ($mm)
Prob
abili
ty o
f Cre
dit L
oss
Economic capital
= Unexpected Loss
= Loss at very high CL– Expected loss.
Expected loss should be covered by reserves and/or pricing.
Expected LossLoss at high CL
Economic Capital
Page 14E 14Evan Picoult, Citigroup January, 2005
Probability Distribution of Potential Credit Loss for a Portfolio of Many Obligors
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
0-20-40-60-80-100-120-140-160Potential Credit Loss ($mm)
Prob
abili
ty o
f Cre
dit L
oss
The probability distribution of potential credit loss, and the ratio UL/EL, depends on the composition of the portfolio and the definition of credit loss.
Expected Loss (EL)Loss at a very high CL (e.g. 99.9%)
Economic Capital for Credit Risk to cover Unexpected Loss (UL)
EXAMPLE: EC FOR CREDIT RISKDefinition Of Economic Capital
Economic Capital For Credit Risk = A measure of risk: The unexpected loss, at a high confidence level, in excess of the expected loss.
Evan Picoult, Citigroup January, 2005 Page 15 Page 15
Economic Loss - Loan Portfolio - Default Only Analysis
ASSUME SOURCE OF CREDIT RISK IS DEFAULT AND RECOVERY ONLY.
• FACTORS NEEDED TO SIMULATE LOSS DISTRIBUTION:
- Credit exposure per obligor
- Probability distribution of exposure at default, for contingent credit.
- Probability of default and correlations of probability of default
- Probability distribution of loss given default (LGD) (i.e. 1 – recovery%).
• There are several very different ways of modeling the potential loss distribution due to default and recovery.
• A robust method will model and capture the relative degree of risk diversification or risk concentration in the portfolio.
Evan Picoult, Citigroup January, 2005 Page 16 Page 16
Evan Picoult, Citigroup January, 2005 Page 17 Page 17
Drivers of EC
Idiosyncratic risk Systemic Risk
Uncertainty in the relative number of obligors that will default.
UL/EL tends to decrease as number of obligors increases.
UL/EL tends to decrease if diversify exposures across many industries / countries.
Uncertainty in relative size of exposure at default.
UL/EL tends to decrease as size of exposures becomes more uniform.
Uncertainty in LGD Relevant for portfolios with:• Few obligors• Inhomogeneous exposure
Economic Loss - Loan Portfolio - Default Only Analysis
- Probability distribution of migration of PD (internal risk rating).
- Volatilities and correlations of change in spread, given rating.
ASSUME SOURCE OF CREDIT RISK IS ECONOMIC LOSS
• FACTORS NEEDED TO SIMULATE LOSS DISTRIBUTION:
- Credit exposure per obligor
- Probability distribution of exposure at default, for contingent credit.
- Probability of default and correlations of probability of default.
- Probability distribution of loss given default (LGD).
Components of default only perspective
Component of long term simulation of obligor’s spread
• THERE ARE SEVERAL VERY DIFFERENT WAYS OF MODELING THE POTENTIAL LOSS DISTRIBUTION DUE ECONOMIC LOSS.
Page 18E 18Evan Picoult, Citigroup January, 2005
ECONOMIC CAPITAL FOR COUNTERPARTY RISK
DEFAULT ONLY PERSPECTIVE
Page 19E 19Evan Picoult, Citigroup January, 2005
Economic Loss - Counterparty Risk - Default Only Analysis
• EC For Counterparty Risk vs. EC For Loan Risk From A Default Only Perspective.
LOAN PORTFOLIO COUNTERPARTY CREDIT RISK
Default and recovery √ √DRIVERS OF WIDTH OF LOSS DISTRIBUTION
Inter-counterparty portfolio w.r.t. default. √ √
Inter-counterparty portfolio w.r.t. exposure √
TYPES OF DIVERSIFICATION BENEFITS
Variable exposure √
Intra-counterparty portfolio √
• Full simulation of EC from a default only perspective:
– For a given path of the market over time, the risk free value of each contract and the corresponding conditional exposure of each counterparty can be fully specified.
– For each path we can therefore simulate thousands of scenarios of default and recoveries, exactly as we would for a loan portfolio from a default only perspective.
– We can then loop over thousands of potential scenarios of changes in market rates.
Page 20E 20Evan Picoult, Citigroup January, 2005
Economic Loss - Counterparty Risk - Default Only AnalysisEC BY FULL SIMULATION: GENERAL METHOD, FIVE STEPS
1) SIMULATE A PATH, P, OF MARKET RATES OVER TIME M(t)P
Same as for Exposure Profile.
5) AFTER SIMULATING THOUSANDS OF POTENTIAL PATHS OF MARKET RATES, M(t)PCALCULATE FULL LOSS DISTRIBUTION AND DERIVE THE FULL SIMULATION ECONOMIC CAPITAL FOR COUNTERPARTY RISK.
2) FOR SIMULATED PATH P, MEASURE THE POTENTIAL MARKET VALUE OVER TIME OF EACH TRANSACTION WITH FOR COUNTERPARTY K. Same as for Exposure Profile.
3) FOR SIMLUATED PATH P, DERIVE COUNTERPARTY K’S POTENTIAL EXPOSURE OVER TIME.i.e. For each counterparty K, for path M(t)P derive Exposure(t)K,P Same as for Exposure Profile.
Loop
ove
r tho
usan
ds o
f pat
hs P
.
4) USING THE SET OF EXPOSURE PROFILES {Exposure(t)K,P } FOR ALL COUNTERPARTIES K, GENERATED BY MARKET PATH P:
CALCULATE THE POTENTIAL LOSS DISTRIBUTION BY SIMULATING THOUSANDS OF SCENARIOS OF DEFAULT AND RECOVERY FOR THE SET OF COUNTERPARTIES K.
Loop
ove
r tho
usan
ds o
f sce
nario
s of
def
ault
and
reco
very
.
Page 21E 21Evan Picoult, Citigroup January, 2005
Economic Loss - Counterparty Risk - Default Only Analysis
IS IT POSSIBLE TO DEFINE A “LOAN EQUIVALENT” FOR COUNTERPARTY EXPOSURE?
• A “Loan Equivalent” is the fixed exposure profile, per counterparty, that would generate the same economic capital as the actual varying potential exposure.
Note:
• The use of a fixed exposure profile per counterparty to calculate total EC for the credit risk of all counterparties entails the incoherent summation of each counterparty’s potential exposure and has the same potential error as the simple, incoherent summation of the potential exposure of each transaction with a counterparty to get the counterparty’s total potential exposure.
HOW GOOD AN APPROXIMATION IS THE EXPECTED EXPOSURE PROFILE AS A LOAN EQUIVALENT? IN WHAT CONTEXT, IF ANY, IS IT A GOOD APPROXIMATION?
WHY DOES QUESTION ARISE?• Internally, transactors need an easy method for evaluating return on risk – i.e. return on EC.
Therefore they need to have a loan equivalent of notional (and equivalent tenor) to plug into EC tables as a function of tenor and obligor risk rating.
• Basel II species risk weight functions as a function of risk parameters: PD, LGD, EAD and M. Therefore need to identify the “loan equivalent” that multiplies the Basel II risk weight.
Evan Picoult, Citigroup January, 2005 Page 22 Page 22
Economic Loss - Counterparty Risk - Default Only Analysis
EC USING EXPECTED POSITIVE EXPOSURE PROFILE AS A LOAN EQUIVALENT:
1) CALCULATE THE EXPECTED POSITIVE EXPOSURE PROFILE, EPE(t)K , OF EACH COUNTERPARTY K
a) Simulate a path, P, of market rates over time M(t)P
Loop
ove
r tho
usan
ds
of p
aths
P.
b) For simulated path, P, measure the simulated market value over time of each transaction with counterparty K.
c) For simulated path, P, derive counterparty exposure over time.i.e. For the counterparty K, for path M(t)P , derive Exposure(t)K,P
d) Derive each counterparty’s expected positive exposure profile
Evan Picoult, Citigroup January, 2005 Page 23 Page 23
3) CALCULATE THE ECONOMIC CAPITAL FROM THE LOSS DISTRIBUTION DERIVED FROM EACH COUNTERPARTRY’S EXPECTED EXPOSURE PROFILE.
2) USING THE SET OF EXPECTED POSITIVE EXPOSURE PROFILES {EPExposure(t)K } FOR ALL COUNTERPARTIES
Calculate the potential loss distribution by simulating thousands of scenarios of default and recovery for the set of counterparties K.
Loop
ove
r tho
usan
ds o
f de
faul
t sce
nario
s.
EPExposure(t)K
Economic Loss - Counterparty Risk - Default Only AnalysisDEFINING THE “LOAN EQUIVALENT” FOR ECONOMIC CAPITAL
From a default only perspective.
Define α as the ratio of Economic Capital calculated in two different ways:
i) EC calculated with full simulation of both defaults and variable exposure.
ii) EC calculated with simulated defaults using a fixed EPE profile for each counterparty
T) CL, (P; Cap Econ
T) CL, (P; Cap Econ )(
ONLY DEFAULT SIM;FIXED_EPE_
ONLY DEFAULT FULL_SIM; =P; CL,Tα
Where: P = Particular portfolio of counterparties with transactions and assumptions about portfolio, e.g. PDs, Correlations, etc.
CL = Confidence Level that EC is measured at.T = Time Horizon over which EC measured.
Therefore: LOAN EQUIVALENTK = α * EPEK for counterparty K
Questions:- What factors does α depend on
Evan Picoult, Citigroup January, 2005 Page 24 Page 24
? How much does α vary, firm by firm? - Is α stable over time i.e. how stable are characteristics of portfolios?
Economic Loss - Counterparty Risk - Default Only AnalysisISDA TESTS
• Create test portfolios and calculate α as a function of the characteristics of the portfolio:- Efective number of counterparties- Effective number of market factors- Probability of default of counterparties- Correlation of default.- Initial MTM of counterparties’ portfolios.- Other factors.
• Initial Proposal: Evan Picoult, Citigroup
• Simulations of Stylized Portfolios: Eduardo Canabarro, GS (now Lehman)
• Analytical Calculations: Tom Wilde, CSFB
• Measurement of α for Real Portfolios: Several Firms
• SUMMARY CONCLUSION: For Large Market Makers α ≈ 1.10
See: - ISDA web site. Papers on Counterparty Risk to Basel Committee, June 2003- Risk Magazine, September, 2003
Evan Picoult, Citigroup January, 2005 Page 25 Page 25
Results from slide of Tom Wilde, CSFB
Page 26E 26Evan Picoult, Citigroup January, 2005
Asset corr'n
Spot value+/-
No factor
s
No. of cpties PD Conf
levelSystematic
riskActual
Portfolio AReference Portfolio B α = Α/Β
λ u K N p q Analytic M-Carlo Analytic M-Carlo Analytic M-CarloAnalyticBase case Percentile Percentile Percentile
22% 1.36 3 200 0.3% 99.9% 10.19 13.14 12.96 12.06 12.02 1.09 1.08Sensitivity to asset correlation
0% 1.36 3 200 0.3% 99.9% 0.51 6.09 NA 4.26 NA 1.43 1.4612% 1.36 3 200 0.3% 99.9% 5.31 8.99 8.91 7.43 7.73 1.21 1.1524% 1.36 3 200 0.3% 99.9% 11.30 14.08 13.96 13.04 13.05 1.08 1.0750% 1.36 3 200 0.3% 99.9% 30.69 32.70 32.50 32.06 31.82 1.02 1.02
Sensitivity to current market values22% 0 3 200 0.3% 99.9% 5.65 8.42 8.23 6.24 6.18 1.35 1.3322% 1 3 200 0.3% 99.9% 8.26 10.96 10.81 9.61 9.61 1.14 1.1222% 2 3 200 0.3% 99.9% 14.28 17.80 17.64 16.95 16.96 1.05 1.0422% 3 3 200 0.3% 99.9% 21.24 25.95 25.73 25.19 25.26 1.03 1.02
Sensitivity to the number of market risk factors22% 1.36 1 200 0.3% 99.9% 10.19 13.22 13.11 12.02 12.02 1.10 1.0922% 1.36 5 200 0.3% 99.9% 10.19 13.07 12.93 12.10 12.02 1.08 1.0822% 1.36 10 200 0.3% 99.9% 10.19 12.97 12.91 12.01 12.02 1.08 1.0722% 1.36 50 200 0.3% 99.9% 10.19 12.96 12.89 12.00 12.02 1.08 1.07
Sensitivity to number of counterparties22% 1.36 3 20 0.3% 99.9% 1.02 3.54 3.72 2.81 2.85 1.26 1.3122% 1.36 3 50 0.3% 99.9% 2.55 5.21 5.26 4.27 4.37 1.22 1.2022% 1.36 3 100 0.3% 99.9% 5.10 7.79 7.83 7.08 6.92 1.10 1.1322% 1.36 3 500 0.3% 99.9% 25.48 28.92 28.36 27.81 27.31 1.04 1.04
Sensitivity to probability of default22% 1.36 3 200 0.1% 99.9% 4.55 7.03 6.93 6.01 6.16 1.17 1.1222% 1.36 3 200 0.5% 99.9% 14.56 17.59 17.56 16.44 16.50 1.07 1.0622% 1.36 3 200 1.0% 99.9% 23.10 26.60 26.50 25.09 25.20 1.06 1.0522% 1.36 3 200 5.0% 99.9% 59.40 65.00 64.55 61.90 61.84 1.05 1.04
Sensitivity to confidence level22% 1.36 3 200 0.3% 99.0% 4.37 6.08 6.11 5.68 5.56 1.07 1.1022% 1.36 3 200 0.3% 99.5% 5.85 7.90 7.90 7.18 7.23 1.10 1.09
Actual portfolio A• This is the portfolio with full
stochastic exposures and correlations as per the settings described above.
Reference portfolio B• This portfolio is as A but with
each exposure fixed = EPE of the corresponding portfolio A counterparty.
α = Aq/Bq
• alpha measures the extra risk arising from the fact that exposures are variable and correlated.
Agreement • The analytic results agree well
to MC – fortunately!
2005 addendum:
α = 1.2 to take into account “general wrong way risk” –negative correlation of changes in general level of yield curve and credit spreads.
ECONOMIC CAPITAL FOR COUNTERPARTY RISK
ECONOMIC LOSS PERSPECTIVE
Page 27E 27Evan Picoult, Citigroup January, 2005
Economic Loss - Counterparty Risk - Full Economic Loss Analysis
FIRST QUESTION:
WHAT SHOULD BE THE EFFECT OF CREDIT SPREADS / RISK RATING ON DERIVATIVE VALUATION?
If all derivatives are (and should be) marked-to-market by discounting expected future cash flows at LIBOR Bid/offer midpoint, then valuation would be:
- Independent of counterparty risk rating and- Independent of counterparty credit spread.
In that case, changes in risk rating or spreads would not cause a change in economic value. Default only and economic loss analysis would be the same.
Page 28E 28Evan Picoult, Citigroup January, 2005
Economic Loss - Counterparty Risk - Full Economic Loss AnalysisHOW SHOULD RISKINESS OF OBLIGOR / MARKET SPREADS AFFECTTHE CURRENT MARKET VALUE OF FORWARD FX AND DERIVATIVES?
• BOND/LOAN VALUE = PV of cash flows discounted at Risk Free Rate – RISK PREMIUMLoan
= PV of cash flows discounted at (Risk Free Rate + SPREAD)
• RISK PREMIUM Loan ≅ PV Risk Free * Duration * Average SpreadLoan
• DERIVATIVE VALUE = PV of cash flows discounted at Risk Free Rate – RISK PREMIUM
CONTEXT FOR ASCERTAINING RISK PREMIUM OF DERIVATIVES. WHY THE CALCULATION IS NOT IDENTICAL TO THAT FOR LOANS:
1) PORTFOLIO ANALYSIS OF EXPOSURE
METHOD FOR CALCULATING RISK PREMIUM MAY DEPEND ON THE PURPOSE OF THE CALCULATION:
- Pricing
- Cost of credit
2) TIME VARYING, UNCERTAIN FUTURE EXPOSURE
3) FUTURE ASSET OR LIABILITY?– In general, at any future date
- Obligor could owe us (asset) or- We could owe obligor (liability)
– Therefore which party’s risk ratings should be taken into account?
Evan Picoult, Citigroup January, 2005 Page 29 Page 29
LET US CALL THE CREDIT RISK PREMIUM OF THE COUNTERPARTY’S PORTFOLIO ITS CVA.CVA = Credit value adjustment for counterparty’s credit risk
THEREFORE, THE MARKET VALUE OF DERIVATIVE PORTFOLIO WITH A COUNTERPARTY IS:= Σ MARKET VALUE (discounted at risk free rate) - COUNTERPARTY RISK PREMIUM
= Σ MARKET VALUE (discounted at risk free rate) - CVA
SIMULATING LOSS DISTRIBUTION OF DERIVATIVE PORTFOLIO: ECONOMIC LOSS ANALYSIS
MEASURING THE CVA OF A COUNTERPARTY: Modification of a proposal by Bollier & Sorensen.
TWO PERSPECTIVES ON CVA: A UNILATERAL AND A BILATERAL PERSPECTIVE.
UNILATERAL CVA: CVACOUNTERPARTY K, UNILATERAL = CVA+CNTPY K
BILATERAL CVA: CVACOUNTERPARTY K, BILATERAL = CVA+CNTPY K - CVA–
CNTPY K
Credit premium of own firm’s expected asset from derivatives
Credit premium of CP’s expected asset from derivatives.
Calculated on a portfolio basis, taking into account potential future exposure
Economic Loss - Counterparty Risk - Full Economic Loss Analysis
Evan Picoult, Citigroup January, 2005 Page 30 Page 30
Economic Loss - Counterparty Risk - Full Economic Loss Analysis
SIMULATING LOSS DISTRIBUTION OF DERIVATIVE PORTFOLIO: ECONOMIC LOSS ANALYSIS
Market Value CP Portfolio = Σ MVCP Portfolio (risk free) - CVA CP Portfolio
CVA CP Portfolio_Unilateral = CVA+CNTPY
CVA CP Portfolio_Bilateral = CVA+CNTPY - CVA-
CNTPYExpected amount CP will owe to own firm.
A Counterparty's Expected Positive Exposure Profile
0
15
30
45
60
75
0 6 12 18 24 30 36 42 48 54 60
Time (Months)
Pote
ntia
l Exp
osur
e ($
MM
)
CVA+CNTPY K (Credit premium of own’s firm potential asset due to CP K)
)*** JJJ,KJ,K dftSpread CP ForwardExposure Expected ( J
∆∑ +=
Calculate and sum over each forward period J, for CP K.
Expected amount own firm will owe to CP.
Evan Picoult, Citigroup January, 2005 Page 31 Page 31
A Counterparty's Expected Negative Exposure Profile
0
15
30
45
60
75
0 6 12 18 24 30 36 42 48 54 60
Time (Months)
Pote
ntia
l Exp
osur
e ($
MM
) CVA-CNTPY K (Credit premium of CP K’s potential asset due to own firm)
Calculate and sum over each forward period J, for CP K
)*** JJJJ,K dftSpreads Firm' Own ForwardExposure Expected( J
∆∑ −=
Economic Loss - Counterparty Risk - Full Economic Loss Analysis
SIMULATING LOSS DISTRIBUTION OF DERIVATIVE PORTFOLIO: ECONOMIC LOSS ANALYSIS
Market Value CP Portfolio = Σ MVCP Portfolio (risk free) - CVA CP Portfolio
CVA CP Portfolio_Bilateral = CVA+CNTPY - CVA-
CNTPY (bilateral perspective)
CVA CP Portfolio_Unilateral = CVA+CNTPY (unilateral perspective)
EXAMPLES
Example 1:
• ASSUME:– Only One Swap With Counterparty– Counterparty And Own Firm Have Same Risk Rating.– Potential Change in Value Has Symmetric Shape For Pay or Receive Fixed Swaps.
(e.g. flat yield curve).
• CONSEQUENCE FOR CVA unilateral and bilateral
Example 2:
• ASSUME:– Only One Swap With Counterparty– Counterparty Is BBB And Own Firm Is AA.
• CONSEQUENCE FOR CVA unilateral and bilateral
Evan Picoult, Citigroup January, 2005 Page 32 Page 32
Counterparty Risk Issues – Maturity and CVA
Deriving M for Unilateral CVA:
• For simplicity, let us define Spread = average spread for counterpartyEPE1 yr = Average EPE over one year horizon.
Therefore:
∑=
=portfolio oflife Full
1kkkk df∆tEPE * Spread CVA
∑=
=portfolio oflife Full
1kkkk df∆tEPE * Spread CVA ∆∆
• In analogy to the relationship between the change in the credit spread of a bond and its duration, we can define the effective M by the equation:
M * EPE * Spread CVA year1∆∆ ≡
• We therefore have the definition of M, from a unilateral perspective:
∑
∑
=
== Year 1
1 kkkk
Maturity
1 kkkk
dft
dft
EPE
EPE M
∆
∆ M is simply the ratio of:
The area under the full-lifetime EPE curve divided by the area under the 1-year discounted EPE curve.
Evan Picoult, Citigroup January, 2005 Page 33 Page 33
For fuller discussion including M for bilateral CVA, see Evan Picoult and David Lamb (2004) Economic Capital for Counterparty Credit Risk, Chapter in Economic Capital: A Practioner Guide, London, Risk Books
Counterparty Risk Issues – Maturity and CVA
A Counterparty's Exposure Profile
0
25
50
75
100
125
150
0 6 12 18 24 30 36 42 48 54 60
Time (Months) ==>
Pot
entia
l Exp
osur
e ($
MM
)
Lifetime area under discounted curve
140.7 EPEPortfolio ofLife Full
1 kkkk dft =∑
=∆
A Counterparty's Exposure Profile
0
25
50
75
100
125
150
0 6 12 18 24 30 36 42 48 54 60
Time (Months) ==>
Pote
ntia
l Exp
osur
e ($
MM
)
EPE 75.9 EPE Year 1Year 1
1 kkkk dft ==∑
=∆
One year area under discounted curve
Therefore M = (140.7/75.9) yrs= 1.85 yrs= 22.2 months
Evan Picoult, Citigroup January, 2005 Page 34 Page 34
Economic Loss - Counterparty Risk - Full Economic Loss Analysis
CONSEQUENCE FOR PSE ECONOMIC CAPITAL.
TO SIMULATE ECONOMIC LOSS DISTRIBUTION:
• SIMULATION NEEDED TO INCORPORATE POTENTIAL DEFAULT AND RECOVERY
- Simulate potential exposure
- Simulate potential defaults
- Simulate potential recovery, given default
• SIMULATION NEEDED TO INCORPORATE POTENTIAL LOSS DUE TO CHANGES IN RISK RATINGS AND/OR CREDIT SPREADS:
- Need to simulate how CVA could change over time per obligor due to:
- Potential changes in exposure profile at a future date- Due to changes in market rates and volatilities/correlations.
- Potential changes in obligor’s risk rating / credit spread at a future date.
- A double level of simulation
Page 35E 35Evan Picoult, Citigroup January, 2005
Page 36E 36Evan Picoult, Citigroup January, 2005
Hedging Counterparty Risk
Market Value CP Portfolio = Σ MVCP Portfolio (risk free) - CVACP Portfolio
Gives rise to market risk Gives rise to counterparty credit risk
∑=
=portfolio oflife Full
1kkkk df∆tEPE * Spread CVA
• For a specific portfolio of U.S. Dollar LIBOR interest rate swaps, the discounted area under the EPE curve, Σ(EPEk ∆tk dfk), will be a function of the terms and conditions of all the swaps and the structure and volatility of the U.S. Dollar LIBOR yield curve.
• Therefore for such a portfolio:
)()(
}){}{(
x contracts;f * Spread x contracts;f * Spread
σr(t) contracts;f * Spread
df∆tEPE * Spread CVA
k
kkkportfolio oflife Full
1k
==
=
= ∑=
kk ,
∆x*∆Spread*xSpread
f ∆x*xf * Spread df∆tEPE * ∆Spread ∆CVA
2portfolio oflife Full
1kkkk ∂∂
+∂∂
+= ∂∑=
Page 37E 37Evan Picoult, Citigroup January, 2005
Hedging Counterparty Risk
∑=
=portfolio oflife Full
1kkkk df∆tEPE * Spread CVA
∆x*∆Spread*xSpread
f ∆x*xf * Spread df∆tEPE * ∆Spread ∆CVA
2portfolio oflife Full
1kkkk ∂∂
+∂∂
+= ∂∑=
Hedge potential changes in market rates in proportion to CP spread.
Determines notional value of interest rate derivatives need to buy to hedge interest rate risk (or more general hedges needed to hedge more general market risk).
Potential risk of correlation of change in spread and changes in market rates.
Determines degree of correlation risk, which may not be possible to hedge.
Hedge spread risk in proportional to discounted EPE curve.
Determines notional value of credit default swaps to buy.
Effectively have transformed counterparty credit risk into market risk.
SUMMARY
• PORTFOLIO SIMULATION OF COUNTERPARTY EXPOSURE PROFILE.– Described method.
• ECONOMIC CAPITAL FOR LOAN CREDIT RISK– Described default only simulation.– Described loss of economic value simulation.
• ECONOMIC CAPITAL FOR COUNTERPARTY RISK - DEFAULT ONLY:– Described full coherent simulation of potential exposure and default.– Described approximation using incoherent simulation with expected
exposure profile scaled up by factor α .
• ECONOMIC CAPITAL FOR COUNTERPARTY RISK - ECONOMIC LOSS:– Need to first define Credit Value Adjustment (CVA) for credit risk of
counterparty’s portfolio. – Need to simulate default, recovery and changes in CVA over time.
• DYNAMICALLY HEDGING COUNTERPARTY RISK
Evan Picoult, Citigroup January, 2005 Page 38 Page 38