2012 Tokyo Ebig Yau

Embed Size (px)

Citation preview

  • 7/27/2019 2012 Tokyo Ebig Yau

    1/43

    Equity-Based Insurance Guarantees Conference

    June 18, 2012

    Tokyo, Japan

    Market Risk Modeling

    Eric Yau

  • 7/27/2019 2012 Tokyo Ebig Yau

    2/43

    6/6/2012

    1

    Eric YauConsultant, Barrie & Hibbert Asia

    [email protected]

    Eric YauConsultant, Barrie & Hibbert Asia

    [email protected]

    18 June 2012 (1150 1230 hours)18 June 2012 (1150 1230 hours)

  • 7/27/2019 2012 Tokyo Ebig Yau

    3/43

    6/6/2012

    2

    Agenda

    + VA market risk modeling: motivation and building blocks

    + Calibrating to the Japan market

    + Risk factor modeling

    Interest rate

    Equity

    Credit

    + Hedging and hedge projection

    2

  • 7/27/2019 2012 Tokyo Ebig Yau

    4/43

    6/6/2012

    3

    mot i vat i on and bui l d i ng blocks

  • 7/27/2019 2012 Tokyo Ebig Yau

    5/43

    6/6/2012

    4

    Motivation of VA modeling

    VA modeling aims to answer a few fundamental questions:

    ProductDesign /Pricing

    What is the cost of guarantees

    embedded in VA?

    How do product features impact this

    cost?

    HedgingValuation

    management strategies

    affect cashflow profile?

    How should we

    implement a hedging

    What is the right level

    of reserve?

    How hedging

    strategies affect

    4

    strategy and test its

    effectiveness?

    reserves?

  • 7/27/2019 2012 Tokyo Ebig Yau

    6/43

    6/6/2012

    5

    Building blocks of VA modeling

    Economic

    Assumption

    Model

    Assumption

    Model

    Choice

    Market

    Data

    calibration

    Risk Monitoring Reports:

    * Asset and liability valuation*

    Economic scenarios:

    Economic Scenario Generator

    * Risk limits and utilization

    Base and Sensitivities

    Liability

    Portfolio ALM SystemAsset

    Portfolio

    Trading System

    Asset Portfolio Optimizer

    Generate, for both asset and

    liability portfolios:

    * Valuation

    * Hedge Strategy

    * RebalancingRules

    * Risk Limits

    5

    Trading

    Engine

    * Greeks

    * Mismatch position

    va a e

    InstrumentsCalculation

    Engine

  • 7/27/2019 2012 Tokyo Ebig Yau

    7/43

    6/6/2012

    6

    VA risk management and modeling

    Product design and pricing

    - Cost of guarantees

    Valuation

    - Reserve/capital calculation- ens t v ty ana ys s - eserve pro ect on

    Ensure fair and adequate pricing / valuation / capital

    Internal hedging

    -

    Hedge projection

    -- Automated calibration

    - Variance reduction

    Capture key risk exposure of hedging strategy

    6

  • 7/27/2019 2012 Tokyo Ebig Yau

    8/43

    6/6/2012

    7

    Desirable ESG features for VA

    + Integrated modeling

    Interest rates, multiple equity indices, credit risks and alternative assets ons stent mar et-cons stent an rea -wor mo e ng

    + Multiple equity modeling choices E.g., Local volatility, Heston with jumps, to support analysis of model risk

    Exact fit to starting yield curve (for interest rate models)

    Accurate fit to option-implied volatility

    And for hedging:

    + Ability to be run on efficient hardware configurations -

    + Fast calibration tools to facilitate re-calibration / sensitivities

    + Automation of scenario production7

  • 7/27/2019 2012 Tokyo Ebig Yau

    9/43

    6/6/2012

    8

  • 7/27/2019 2012 Tokyo Ebig Yau

    10/43

    6/6/2012

    9

    The market

    Bond / Interest rate market+ Longest available swap / JGB up to 40 years

    + Swap liquid tenors up to ~10 years

    Derivative implied volatilities+ E.g. Nikkei 225 options up to ~10 years

    What happens if we need to discount long term liabilities?

    9

  • 7/27/2019 2012 Tokyo Ebig Yau

    11/43

    6/6/2012

    10

    Linkage to liability valuation

    + This generally apply to a number of areas

    Interest rate volatility

    Equity volatility10

  • 7/27/2019 2012 Tokyo Ebig Yau

    12/43

    6/6/2012

    11

    Simple extrapolation: interest rate

    12%USD government forward

    8%

    9%

    10%

    11%

    strate

    rates assuming constant

    rate beyond 30 years

    for1985-2007

    5%

    6%

    7%

    Forwardin

    ter

    Very conservative and willgenerate very high

    volatility in the MTM value

    of ultra long-term cash

    2%

    3%

    0 10 20 30 40 50 60 70 80 90 100

    Maturity(years)

    .

    11

  • 7/27/2019 2012 Tokyo Ebig Yau

    13/43

    6/6/2012

    12

    Macroeconomic extrapolation

    Three questions:

    latility

    Limiting, unconditionalforwardrate/IVassumption1) What is the longest

    market interest rate that

    we can observe?

    ate/Option

    V

    Marketforwards

    2) What is an appropriateassumption for the very

    long-term

    unconditional or

    InterestRorwar ra e

    3) What path should beset between the longest

    12

    0 10 20 30 40 50 60 70 80 90 100 110 120

    Term(years)mar e ra e an e

    unconditional forward

    rate?

  • 7/27/2019 2012 Tokyo Ebig Yau

    14/43

    6/6/2012

    13

    Extrapolation of interest rate

    A common approach

    + Fitting to available market data+ Setting a target for the ultra long-term forward rate

    + Developing an economically sensible functional form

    11%

    12% 11%

    6%

    7%

    8%

    9%

    10%

    ardinterestrate

    5%

    6%

    7%

    8%

    9%

    ardinterestrate

    2%

    3%

    4%

    5%

    0 10 20 30 40 50 60 70 80 90 100

    For

    Maturity(years)1%

    2%

    3%

    4%

    0 10 20 30 40 50 60 70 80 90 100

    For

    Maturity (years)

    + Unconditional anchor: stability in mark-to-model valuations

    13

  • 7/27/2019 2012 Tokyo Ebig Yau

    15/43

    6/6/2012

    14

    Extrapolation of derivative implied vol

    + A perfect fit to market data? + Economically robust, stable

    + How should we extrapolate?

    extrapolation lead to stable,

    sensible valuation

    14

  • 7/27/2019 2012 Tokyo Ebig Yau

    16/43

    6/6/2012

    15

    Example: equity implied vol

    + One possible extrapolation approach:

    Real-world volatility estimate as the limit of extrapolation... ... Adjusted for empirical option implied volatility / real-world volatility bias

    How fast do we revert to this long-term position?

    N225Im liedVolatilitiesandExtra olation

    30%

    35%

    40%

    45%

    latilities

    10%

    15%

    20%

    25%

    EquityImpliedV

    Q42007

    Q42008

    Q42009

    Q42010

    15

    0%

    5%

    0 5 10 15 20 25 30 35 40 45 50

    Term(years)

    Q42011

  • 7/27/2019 2012 Tokyo Ebig Yau

    17/43

    6/6/2012

    16

  • 7/27/2019 2012 Tokyo Ebig Yau

    18/43

    6/6/2012

    17

    What makes VA modeling difficult?

    Embedded derivative nature

    + Path-dependent payoff+ Multiple assets, multiple time periods

    Uncertainty

    + Insurance risks

    + Management actions

    + Policyholder options

    ut are po cy o ers pr ce-sens t ve or u y rat ona

    Complex exposure to various financial risk factors

    17

    - , - , -

    for a realistic estimation of risk exposure and valuation

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    19/43

    6/6/2012

    18

    Sample ESG model structure

    Property returnsAlternative asset returns

    (e.g. commodities)

    Credit risk model

    Corporate bond returnsEquity returns

    Initial swap and

    government nominal

    bonds

    Nominal short rate

    Real-economy; GDP

    and real wages

    Exchange rate

    (PPP or IRP)

    Inflation

    expectations

    Real short rateIndex linked

    government bonds

    Realised Inflation and

    alternative inflation

    Foreign nominal

    short rate and

    inflation

    o n s r u on

    Correlation relationships between the shocks to different models

    Economically rational structure 18

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    20/43

    6/6/2012

    19

    Real-world vs Market-consistent

    A clarification of terminology

    - -

    Question to answer: What is the probability

    distribution of future asset

    prices?

    What is the current

    market-consistent value of

    future cashflows?

    Usage: Financial projections for

    ALM, cashflow testing,probability of ruin analysis

    Fair valuation of liabilities

    (and Greeks)

    Calibration: Calibrated to best-estimate Calibrated to market

    arge s op on- mp e vo a es

    Risk premium: Y N

    19

    The section below focuses on market-consistent modeling

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    21/43

    6/6/2012

    20

    n eres ra e mo e s

    20

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    22/43

    6/6/2012

    21

    Key consideration

    + Provide good fit to market option-implied volatility surface

    + Take yield curve as input+ Flexibility in volatility factor specification and modeling

    From a modeling perspective it generally means

    + A number of popular yield curve choices for MC modelling

    Hull-White

    ox- ngerso - oss

    Heath-Jarrow-Morton

    +

    In particular LIBOR Market Model is a market standard for rate derivatives trading+ Fast robust calibration tools should be available for fre uent re-

    calibrations

    21

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    23/43

    6/6/2012

    22

    Low interest environment

    + Negative interest rate issues with

    Gaussian models like HW Lognormal models with displacement

    + Theoretically acceptable, but in practice

    22

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    24/43

    6/6/2012

    23

    Model choice matters

    Hull-White /

    Vasicek

    Black-

    Karasinski

    LMM DDLMM + SV

    Fit to initial

    yield curve

    Depends on

    implementation

    Depends on

    implementation

    Fit to swaption

    pr ces

    Calibrationefficiency

    Negative

    interest rate

    Yes No No Yes

    23

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    25/43

    24

    qu y mo e s

    24

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    26/43

    25

    Black-Scholes model as a starting point

    + Model assumption affects fair valuation / pricing of liability

    + Among other assumption: Returns are normally distributed

    Volatility is constant

    Constant volatilit assum tion

    0.7%

    0.8%

    0.9%

    1.0%

    y

    Historic (20th Century)

    Stochastic Vol Model

    Normal Distribution

    35.00%

    40.00%

    0.2%

    0.3%

    0.4%

    0.5%

    0.6%

    Frequenc

    10.00%

    15.00%

    20.00%

    25.00%

    30.00%

    25

    0.0%

    0.1%

    -30% -25% -20% -15% -10%

    Equity returns in excess of ri sk-free rates

    0.

    25 0

    .5

    0.

    75

    1

    2

    34

    57 1

    0

    0.00%

    5.00%

    Maturity

    Strike

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    27/43

    26

    Key consideration

    + Provide good fit to market option-implied volatility surface

    + Support fast and frequent re-calibration+ Provide simultaneous fit to multiple equity indices vol surfaces

    From a modeling perspective it generally means

    + Going beyond Black-Scholes (constant volatility), e.g.

    Stochastic volatility

    orre a on e ween re urn an vo a y

    Mean reverting volatility and volatility clustering

    + But still provide (semi-) analytical calculation

    Technology infrastructure for daily recalibration

    26

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    28/43

    27

    Market-consistent valuation

    + Model should first fit to market prices of derivativesMarket Volality Sur face Model Volality Surface

    35.00%

    40.00%

    35.00%-40.00%

    30.00%-35.00%

    25.00%-30.00%

    20.00%-25.00% 30.00%

    35.00%

    40.00%

    35.00%-40.00%

    30.00%-35.00%

    25.00%-30.00%

    20.00%-25.00%

    5.00%

    10.00%

    15.00%

    20.00%

    25.00%

    .

    15.00%-20.00%

    10.00%-15.00%

    5.00%-10.00%

    0.00%-5.00% 5.00%

    10.00%

    15.00%

    20.00%

    25.00% 15.00%-20.00%

    10.00%-15.00%

    5.00%-10.00%

    0.00%-5.00%

    0.2

    5 0.

    50.

    75

    1

    2

    3

    45

    7 10

    0.00%

    Maturity

    Strike

    0.

    25 0

    .5

    0.

    75

    1

    2

    3

    4

    5

    7 10

    .

    Maturity

    Strike

    s mp e ac - c o es mo e canno a vo a y sur ace - u

    market implies one

    27

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    29/43

    28

    What is SVJD?

    + The SVJD model is a combination of two well known models of

    quantitative finance

    The Heston Stochastic Volatility Model

    er on s ump us on o e

    + Benefits:

    Fairly realistic but parsimonious model

    Generally provides a good fit to volatility surface at both long and short maturities

    Semi-analytic (i.e. fast) valuation formulae for vanilla option prices

    28

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    30/43

    29

    Technical specification

    + Two parts:

    Stochastic volatility part, Heston model (red) Jump diffusion part, Merton model (blue)

    29

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    31/43

    30

    Example simulationSimulation of the SVJD Model

    Example Simulation

    45%200.00

    10 Year Total Return Index

    Index No Jumps Index Jumps Only Index Stochastic Volatility

    30%

    35%

    40%

    120.00

    140.00

    160.00

    .

    ility

    ex

    15%

    20%

    25%

    60.00

    80.00

    100.00

    Stochastic

    Volati

    TotalReturn

    In

    0%

    5%

    10%

    0.00

    20.00

    40.00

    30

    Month

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    32/43

    31

    Calibrating to implied vol surface

    Market Feature Model Component Key Parameters

    Implied volatility term

    structure

    Stochastic variance Initial variance, mean

    reversion level, andspeed of mean reversion

    Lon term skew/smile Stochastic Variance Return-variance

    correlation and volatility

    of variance

    Short term skew/smile Jump Diffusion Jump parameters

    + Realistic description of underlying dynamics is key

    volatility surface

    31

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    33/43

    32

    re mo e s

    32

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    34/43

    33

    Impact of credit models

    + The impact of credit risky bonds can be highly significant

    On both guarantee costs / pricing

    And required hedge portfolio

    2.60%

    2.80%

    3.00%

    )

    2.00%

    2.20%

    2.40%

    'tees(%

    InitialFund

    premium GMAB & GMDB for a 45year-old male for 20 years 50%

    invested in equities and 50% in a

    1.40%

    1.60%

    .

    Costof-yr on un ...

    33

    1.00%

    .

    Govt AA BBB

    AssumedBondStrategy

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    35/43

    34

    Credit Modelling: Key Features

    A good credit risk model should be

    + Arbitrage-free

    + Stochastic credit rating migrations and defaults+ Stochastic variations in credit spreads

    + Integrated with other market risks (e.g. equities and interest rates)

    o e ynam c ers or eren app ca ons:

    + Real world modelling Pricing matrix more severe than underlying transition matrix

    =>

    + Risk neutral modelling

    Pricing matrix as severe as underlying transition matrix

    => risk neutral

    34

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    36/43

    35

    hedge projection

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    37/43

    36

    Projection of hedging strategies

    + Real-world model capturing hedging strategys key risk exposures

    Vega: Increases in option-implied volatility levels

    - -

    + Integrated modelling of equity total returns and changes in the level

    of equity option-implied volatilities

    25%

    30%

    35%

    40%

    45%

    50%

    gVolatility

    0.5

    0.6

    0.7

    0.8

    0.9

    1.0

    Correlation

    UK Volatility (LHS)

    US Volatility (LHS)

    UK vs US Correlation (RHS)

    0%

    5%

    10%

    15%

    20%

    -63

    -68

    -73

    -78

    -83

    -88

    -93

    -98

    -03

    5Y

    Rolli

    0.0

    0.1

    0.2

    0.3

    0.4

    5Y

    Rollin

    36

    De

    De

    De

    De

    De

    De

    De

    De

    De

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    38/43

    37

    Long-term Hedging P/L Analysis

    + Long-term profitability of delta hedging strategy driven by how

    realized volatility behaves relative to implied volatility levels

    Expected real-world volatility vs option-implied volatility

    Variability of real-world volatility

    15.0%1.50%

    -2.5%

    0.0%

    2.5%

    5.0%

    7.5%

    10.0%

    12.5%

    -0.25%

    0.00%

    0.25%

    0.50%

    0.75%

    1.00%

    1.25%

    PriceChane(%

    )

    edgingLosses

    rlyinhgfund)

    CumulativeDeltaHedgingP/L

    -17.5%-15.0%

    -12.5%

    -10.0%

    -7.5%

    -5.0%

    -1.75%-1.50%

    -1.25%

    -1.00%

    -0.75%

    -0.50%

    DailyS&P500

    Cumulative

    (%ofunde

    37

    - .- .

    1 3 5 7 9 1113151719212325272931333537

    TradingDay(October1st November20th'08)

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    39/43

    38

    Variance reduction for risk assessment

    + Hedge projection requires Greeks projection

    Estimation of Greeks at each point in each real-world simulation requires

    stochastic-on-stochastic (theoretically)

    + Variance reduction for Greeks calculation

    E.g. Least Squares Monte Carlo to efficiently capture future liability valuation

    38

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    40/43

    39

    Example LSMC for projected Greeks

    + Simple Black-Scholes example

    projection of 10-year vanilla put option

    + 10,000 outer scenarios and 2 inners per outer

    Higher numbers of inner sims can be used to increase accuracy

    + Fit cubic function and differentiate to estimate delta

    0.4

    0.5

    True value (Black-Scholes)

    I-0.2

    -0.1

    0

    0.5 1 1.5 2 2.5

    Value (after 1 year) Delta (after 1 year)

    y = -0.1367x3 + 0.74x2 - 1.4881x + 1.1539

    0.1

    0.2

    0.3

    OptionValue@Year1 In a -scenaro es mae

    Regression estimate

    Differentiate

    fitted polynomial

    -0.8

    -0.7

    -0.6

    -0.5

    -0.4

    -0.3

    ption

    Delta@Year1

    True delta (Black-Scholes)

    Regression estimate

    39

    0

    0.5 1 1.5 2 2.5

    Put

    Equity Price @ Year 1

    -1

    -0.9

    Put

    Equity Price @ Year 1

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    41/43

    40

    Concluding remarks

    + Market risk modeling impacts

    Fair and adequate pricing / valuation / capital

    Design / risk exposure of hedging strategy

    + Model choice and calibration, among others, have significant impact

    40

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    42/43

    41

    6/6/2012

  • 7/27/2019 2012 Tokyo Ebig Yau

    43/43

    42

    Copyright 2012 Barrie & Hibbert Limited. All rights reserved. Reproduction in whole or in part is

    prohibited except by prior written permission of Barrie & Hibbert Limited (SC157210) registered in

    Scotland at 7 Exchange Crescent, Conference Square, Edinburgh EH3 8RD.

    .

    estimates included in this document constitute our judgment as of the date indicated and are subject

    to change without notice. Any opinions expressed do not constitute any form of advice (includinglegal, tax and/or investment advice). This document is intended for information purposes only and is

    not intended as an offer or recommendation to buy or sell securities. The Barrie & Hibbert group

    excludes all liabilit howsoever arisin other than liabilit which ma not be limited or excluded at

    law) to any party for any loss resulting from any action taken as a result of the information provided in

    this document. The Barrie & Hibbert group, its clients and officers may have a position or engage in

    transactions in any of the securities mentioned.

    Barrie & Hibbert Inc. and Barrie & Hibbert Asia Limited (company number 1240846) are both wholly

    42

    owne su s ar es o arr e ert m te .

    www.barrhibb.com