Title: Estimating Credit Exposure and Economic Capital Using Monte Carlo Simulation Ronald Lagnado Vice Pre
1Estimating Credit Exposure and Economic Capital
Using Monte Carlo SimulationRonald LagnadoVice
President, MKIRiskIPAM Conference on Financial
MathematicsJanuary 11, 2001
2Monte Carlo Simulation for Integrated
Market/Credit Risk
- Random sampling generates potential future paths
of market/credit risk sources - Provides time profile of credit exposure and
distribution of losses - Facilitates effective management of credit limits
and optimal allocation of capital
3Benefits of Monte Carlo Simulation for Credit
Risk Analysis
- Efficient Capital Allocation
- Avoid overstating credit exposure by correctly
aggregating across master agreements, time, and
market scenarios - Account for netting, collateral,
less-than-perfect correlation, mean reversion,
etc. - Prudent Capital Allocation
- Account for default correlation, risky
collateral, margin call lags, correlation
instability, etc.
4MKI Integrated Risk Management Solution
Source Systems
Optional Middleware
Limit Management RV Limits
Source Systems
Consolidation Database - RV Data
Source systems
A P I 's
Source systems
Manual Entry
Price Feed Sources
5Monte Carlo Simulation
Value
Begin With Current Mark-to-Market
Base Mark- to- Market
Time Nodes
1 2 3 4
5 6 7 8 9
Time (Nodes)
6Monte Carlo Simulation
7Monte Carlo Simulation
8Monte Carlo Simulation
9Monte Carlo Simulation
10Monte Carlo Simulation
Value
NEW MARKET DATA
VALUE EVERY DEAL
Base Mark- to- Market
ASSIGN TO PORTFOLIOS
APPLY NETTING, COLLATERAL, ETC.
Time Nodes
1 2 3 4
5 6 7 8 9
Time (Nodes)
11Monte Carlo Simulation
Value
Base Mark- to- Market
Repeat for Successive Time Nodes
Time Nodes
1 2 3 4
5 6 7 8 9
Time (Nodes)
12Monte Carlo Simulation
Distribution of Portfolio Values, Exposures,
etc.
Value
Base Mark- to- Market
Runs
Time Nodes
1 2 3 4
5 6 7 8 9
Time (Nodes)
13Credit Exposure Profiles
Portfolio Exposure Dynamics
Exposure
Max Exposure
Future Potential Exposure
1 Std Dev
Y Std Dev
Mean
Current Exposure
0 1
T
Future Simulation Dates
14Credit Relationships
Counterparty C - Guaranteed or not
Counterparty B - Guaranteed or not
Counterparty A - Guaranteed or not
Master Agreement A2
Master Agreement A1
CSA A12
CSA A11
Trade 10003
Trade 10002
Collateral
Trade 10001
15Counterparty Exposure (Netting)
- Net credit exposure to Counterparty i
16Market Risk Drivers
- Interest Rates
- Base Term Structures
- Spread Term Structures
- Exchange Rates
- Equities
- Indexes
- Individual Stocks
- Commodities
- Spot Prices
- Forward Prices
- Implied Volatility Surfaces
17Example Interest Rate Process
- r vector of interest rates drivers
- ? vector of mean reversion levels
- A matrix of mean reversion speeds
- ? instantaneous covariance matrix
- Z vector of independent Brownian motions
18Example Interest Rate Process
- Integrate over time step discrete VAR(1)
process
19(No Transcript)
20Parameter Estimates USD Libor
- rates 1m 3m 6m 1y 2y
3y 5y 7y 10y - speed 0.51 0.37 0.42 0.51
0.50 0.64 0.78 0.80 0.78 - volatility 0.23 0.19 0.20 0.20
0.16 0.16 0.15 0.14 0.13 - correlation 1.
- 0.39 1.
- 0.34 0.48 1.
- 0.24 0.35 0.53
1. - 0.23 0.35 0.40 0.51 1.
- 0.22 0.33 0.38 0.49 0.97
1. - 0.20 0.31 0.36 0.46 0.93
0.95 1. - 0.19 0.29 0.34 0.44 0.88
0.91 0.96 1. - 0.17 0.27 0.31 0.42 0.83
0.87 0.93 0.96 1.
21(No Transcript)
22(No Transcript)
23(No Transcript)
24Option Exposure Comparison of Exact Results
with Monte Carlo
- Equity Index Call Option
- expiration 2 years
- implied volatility 20
- initially at-the-money
- Underlying Stochastic Parameters
- drift 15
- volatility 20
- Monte Carlo Simulation Weekly Time-Steps
- Exact Results Obtained with Gauss-Hermite
Quadrature
25(No Transcript)
26(No Transcript)
27(No Transcript)
28Simulation of Dynamic Collateral and Margin
Call Lags
- Example
- Single Counterparty
- Single Transaction 2-year equity call option
- Margin Call Parameters
- Threshold 30 Million
- Margin Call Lag 4 weeks
- Delivery Lag 1 week
- Excess Collateral Returned Immediately
- Monte Carlo Simulation 10000 paths
29(No Transcript)
30(No Transcript)
31(No Transcript)
32Losses and Capital Calculation
- Model Requirements
- Exposure Profiles
- Credit Quality Migration and Default (Correlated)
- Stochastic Recovery
- Benefits
- Loss Reserves and Economic Capital
- Capital Allocation across Business Units
- Performance Measures (RAROC)
- Incremental Capital and Capital-Based Pricing
33The Losses Distribution
Distribution of Losses (Integrated Market/Credit
Risk Simulation)
Losses PDF
0
PV(Losses))
34The Losses Distribution
Distribution of Losses (Integrated Market/Credit
Risk Simulation)
Losses PDF
Expected Losses
0
PV(Losses))
35The Losses Distribution
Distribution of Losses (Integrated Market/Credit
Risk Simulation)
Losses PDF
Expected Losses
Unexpected Losses
0
PV(Losses))
36The Losses Distribution
Distribution of Losses (Integrated Market/Credit
Risk Simulation)
Losses PDF
Expected Losses (Reserves)
Unexpected Losses (Economic Capital)
0
PV(Losses))
37Credit Migration Model
- Markov chain with transition probability
matrix -
- probability of
migrating from rating to rating during the
time interval -
38Credit Migration Model
- Time Inhomogeneous
-
- Time Homogeneous
-
39Typical Transition Matrix (1-Year)
40Credit Quality Migration and Default
Correlation
- Factor Model for Asset Value Return
- For each counterparty
41Credit Migration Quantiles
BBB
BB
A
B
AA
CCC
AAA
D
0
Change in Firm Value (Normalized)
42Relating Asset Returns to Default Correlation
- Asset-Return Correlation
- Default Correlation
-
43(No Transcript)
44Losses
- discrete time nodes
- market risk driver path
- idiosyncratic credit driver path
- default stopping time
-
45Loss Statistics (Simplified Case)
- Single-period Independent exposure and default
-
46Loss Statistics (Simplified Case)
- Single-period
- Constant and identical exposures
- Identical default probabilities and correlations
-
47Loss distributions 500 counterparties, constant
exposures, p 0.05
48Tolerance Intervals
- Ordered sample of losses from Monte Carlo
simulation - Estimated quantile
- Distribution of order statistics
-
49Tolerance Intervals
- Construct non-parametric
confidence interval for estimated quantile -
50Convergence of Unexpected Losses
- 500 counterparties, 550 deals, 1 year horizon
51Summary
- Monte Carlo simulation is preferred approach for
integrated market/credit risk analysis - Reveals distributions of future credit exposure
and losses to default - Facilitates efficient capital allocation by
correctly aggregating exposure across time and
market scenarios - Leads to prudent capital allocation by accounting
for market complexities