Beruflich Dokumente
Kultur Dokumente
R is k Management
Presented by Joshua Corrigan, Charles Qin
Co-authored by Gene Leynes, Chris Onken
© Milliman
C ontents
1. Needs and Applications
2. What is a Replicating Portfolio
3. Methodologies
4. Case Studies
5. Implementation Challenges
C ontents
1. Needs and Applications
2. What is a Replicating Portfolio
3. Methodologies
4. Case Studies
5. Implementation Challenges
R is k Management
Best Practice
Dynamic Hedging Real World Replication
Stochastic
Most variable annuities
(light or full)
C ontents
1. Needs and Applications
2. What is a Replicating Portfolio
3. Methodologies
4. Case Studies
5. Implementation Challenges
Definition
• A replicating portfolio is an asset portfolio consisting of:
–Real assets with liquid market prices
–Real illiquid assets
–Theoretical assets
that replicates the scenario-dependent cash flows of an asset or
liability
Theoretical static Theoretical Real world
assets dynamic assets dynamic hedging
T heoretic al F ramework
Implementation of
Principle of No Arbitrage
Replicating Portfolios
• If two instruments or • If the cash flows of a portfolio
portfolios have identical cash of assets and the liabilities
flows under all economic are identical under the base
conditions... and stressed economic
scenarios...
• then the prices of the two • then the market value of the
instruments must be equal liabilities must equal that of
the asset portfolio
Illus trative E xample 1
1.00 unit
1.5-years fixed coupon bond
6% semi-annual coupon
Capital
MC Liability Values
T he G AM Model
• Fit GAM curves to (base & stressed) liability cash flows at each time step
GAM
• Apply optimisation, which balances between both goodness of fit and smoothness
Fitting (robustness)
• Goodness of fit measured with standard statistics, e.g. generalised cross validation (GCV)
and R2
Diagnostics • Out-of-sample tests performed with additional economic scenarios
Liability Claim
Amount
Index Level
Liability Claim Amount
T he G AM Model
Index Level
T he E quity R eplic ation Model
• Models net assets (assets less liabilities) at next time step as a multivariate polynomial of
pre-selected risk factors
Parametric • Determine coefficients via optimisation, by utilising both normal and extreme economic
Fitting scenarios
120,000
GAM R2 = 99.95%
RP Portfolio
100,000
100,000
RP Liability Values
Liability Values
80,000
60,000
50,000
40,000
20,000
0 0
0% 50% 100% 150% 200% 250% 300% 0 50,000 100,000 150,000
300,000
GAM R2 = 99.65%
RP Portfolio
300,000
250,000
RP Liability Values
Liability Values
200,000
200,000
150,000
100,000
100,000
50,000
0 0
0% 50% 100% 150% 200% 250% 300% 0 100,000 200,000 300,000 400,000
300% 0.06%
200% 0.04%
100% 0.02%
0% 0.00%
-100% -0.02%
-200% -0.04%
-300% -0.06%
Time Step
C as e S tudy – C omplex G MAB
Liabilities MG-Hedge Values RP Values
Bond
European Options
116,948
368,834
• Asian and look-back equity
Asian Options
Look back Options
22,067
-346,121 options required for replicating
Total Base Values 164,800 161,729
-20% Equity Stress
Capital Strain
192,287
27,487
188,423
26,694
dynamic lapses and ratchets
Delta
Delta MG-Hedge Values RP Values
1,026 1,007 • Excellent replication of base
0-1 Year
Rho MG-Hedge Values RP Values
3 0
liability values
2-5 Year 20 17
6-10 Year 80 89
11-20 Year 203 199 • Minimal errors in the sensitivities
21-30 Year 0 0
Parallel 306 306
to key risk variables
Vega MG-Hedge Values RP Values
1-2 Year 535 475
3-4 Year
5-6 Year
370
733
721
807
• Very accurate capital strain from
7-8 Year
9-10 Year
889
3403
820
3187
equity stress
Parallel 5,931 6,010
C as e S tudy – R ange F orward
1,000,000 1,500,000
MG-Hedge Liabilities
800,000
GAM R2 = 99.85%
RP Portfolio 1,000,000
600,000
400,000
RP Liability Values
500,000
Liability Values
200,000
0
0
-200,000
-400,000
-500,000
-600,000
-800,000 -1,000,000
40% 90% 140% 190% -1,000,000 -500,000 0 500,000 1,000,000 1,500,000
Candidate
Liability Review
and Clean Asset
Data
Identification
Construct
Scenario Optimisation Review RP Asset
Volatility and Revise
Data Process Positions
Surfaces
Goodness
Of Fit Results
C hallenges
• Business requirements • Timelines of data dependencies
– Scope: single / multiple blocks, full • Robustness versus fit trade-off
balance sheet
• Goodness of fit thresholds of
– Characteristics: accuracy, acceptability
frequency, level of detail
• Choice of most appropriate
method
Expertise and experience can make all the difference
C onc lus ion
• Replicating portfolios have numerous applications:
–Risk management: assessment, hedging,
–Capital management: APRA Pillar 2 requirements
–Internal MI and education