0 оценок0% нашли этот документ полезным (0 голосов)

81 просмотров131 страницаMar 16, 2011

© Attribution Non-Commercial (BY-NC)

PDF, TXT или читайте онлайн в Scribd

Attribution Non-Commercial (BY-NC)

0 оценок0% нашли этот документ полезным (0 голосов)

81 просмотров131 страницаAttribution Non-Commercial (BY-NC)

Вы находитесь на странице: 1из 131

Collaborators:

Vivek Kapoor

vivek2.kapoor@citi.com

Citi

Jun Zhang Credit Suisse

Olivia Siu Natixis

May 2008

The views expressed here are those of the author, and do not necessarily represent those of Citi, and no representation as to the

accuracy or completeness of the information provided.

CDX.NA.IG QUOTES

Tranche Price Correlation Delta

0% -3% 500 bps +33.5% 19% 17x

3% -7% 199 bps 5% 7x

7% -10% 64 bps 16% 2.8x

10% -15% 25 bps 21% 1.1x

15% -30% 10 bps 32% 0.3x

Tranche Price Correlation Delta

0% -3% 500 bps +63.5% 8% 14x

3% -7% 255 bps ₋ 7x

7% -10% 57 bps 5% 2x

10% -15% 32 bps 13% 1x

15% -30% 19 bps 30% 0.3x

1

SYNTHETIC CDO TRADING - WHY?

Chase opportunity sloshing across the capital structure

Replication of a CDO tranche by CDS is difficult*:

• Jumps in spread and jumps to default

• Uncertain recovery

• Random realized spread-spread correlation

Tradeoff systematic and idiosyncratic spread convexity

Hunger for yield

#convexity tradeoffs not well captured in risk systems

2

LONG CREDIT INDEX POSITION

selling protection on a pool of CDS

denoted by the solid line, and the

trading book CDS index

default contingent cash flows are

denoted by the dotted line

with this elementary long credit

position

3

SPREAD DELTA-NEUTRAL SELL EQUITY PROTECTION

reference selling protection on an equity

CDS

index tranche referencing a credit index,

and purchasing protection on the

credit index to hedge away to spot

spread delta

synthetic

CDO

C tranches

D with CDS

S index as The pool expected loss refers to

reference

I pool the sum of the contingent legs of

N mezz/senior the CDS in the reference index

D tranche

E (i.e., it is the cost of buying default

X protection on the whole pool if the

trading book equity

pool

tranche

expected CDS contracts were all upfront pay)

loss

4

CASHFLOW & TIME DECAY VIEW AT INCEPTION

80% 40%

CDO P&L (% of tranche notional)

30% cash

60%

cash

20%

40% total P&L positive carry of delta-hedged

positive carry by selling

10% sell equity protection trade

equity tranche protection

20% total P&L

0%

0% -10%

-20% mtm

-20%

mtm -30%

-40%

0 1 2 3 4 5 6 -40%

time elapsed (years) 0 1 2 3 4 5 6

time elapsed (years)

(a) Sell 0%-3% protection

(c) “Spot Delta-hedged” Sell 0-3% protection trade

0%

mtm

-5%

CDS P&L (% of tranche notional)

-10%

negative carry by buying

Un-hedged and “delta hedged”

-15%

-20% cash

delta index protection

sell equity protection trades are

-25%

total P&L

both positive carry

-30%

-35%

-40%

Time decay and total carry view

-45% for 0% - 3% tranche of

0 1 2 3 4 5 6

time elapsed (years) CDX.NA.IG.4 (3/31/2005)

(b) Buy spot “delta” index protection

5

SPREAD DELTA-NEUTRAL STRANGLE

reference

CDS

index

Another popular CDO trade

consists of selling protection on an

equity tranche referencing a credit

synthetic

CDO index, and purchasing protection

tranches

with CDS

on a senior or mezzanine tranche

index as to hedge away to spot spread delta

reference

pool

mezz/senior

tranche

pool

trading book equity

expected

tranche loss

6

SPREAD DELTA-NEUTRAL MEZZ/SR TRADE

reference

CDS

index

In this CDO trade protection

is purchased on a senior or

mezzanine tranche

synthetic

C

CDO referencing a credit index,

tranches

D with CDS and the spot spread delta is

S index as hedged away by selling

reference

I pool protection on the credit index

N mezz/senior

D trading book tranche

E

X

pool

equity

expected

tranche loss

7

MARGINAL VALUE ON DEFAULT (VOD)

There are 125 issuers in the

-0.1%

credit index and associated

CDO analyzed here

Long CDS Index, Carry = 49 bps/yr (right)

single name Value on Default

-1% Delta Hedged Senior Mez Tranche

The horizontal axis is the

(% tranche notional)

(% index notional)

distinct issuers, and the

vertical axis is the marginal

Delta Hedged Equity Tranche (0-3%), Carry = 457 bps/yr (left)

-10%

P&L impact of default (VOD)

of distinct issuers

Straddle (0-3% and 7-10%), Carry = 1149 bps/yr (left)

The greater the carry

-100% of the “delta hedged”

0 50 100 150 200 250 300 350 400 450 500 trade is, the larger

5 year credit spread (bps)

the loss due to

default of any name

Value On Default (VOD) sensitivity for three CDO in the reference pool

strategies and the long credit index trade (CDX.

NA.IG.4, March 31, 2005)

8

CUMULATIVE VALUE ON DEFAULT

100 100

Sell Equity Protection (0-3%) Buy Senior Mez Protection (7-10%)

75

Buy CDS Protection 75 Sell CDS Protection

P&L (% of tranche notional)

Net Net

50

50

25

0 25

(25) 0

(50)

(25)

(75)

(100) (50)

1 5 9 13 17 21 25

1 2 3 4 5 6 7 8 9 10

number of defaults

number of defaults

500

P&L (% of equity tranche notional)

400 Buy Senior Mezzanine Protection (7-10%)

Net

The cumulative VOD shown here is

300 based on sorting the issuers in the

200

order of decreasing spreads and

100

defaulting the top N names

simultaneously

0

1 3 5 7 9 11 13 15 17 19

(CDX. NA.IG.4, March 31, 2005) 9

ONE YEAR DEFAULT RISK PER UNIT CARRY

Shower trade with defaults 0

Long Index

from MC simulation Delta Hedged Equity Tranche (0-3%)

Revalue trade under impact 0.9

Straddle (0-3%, 7-10%)

confidence level

of defaults

0.99

Assess P&L impact at

different confidence levels

0.999

Compare default risk

per unit carry to vanilla

0.9999

long trade: more -1200% -900% -600% -300% 0%

informative than staring VOD (% of carry)

carry (largest for the

straddle) Default risk per unit carry

10

ONE YEAR DEFAULT RISK PER UNIT CARRY

0

0.9 5/16/05 (CDX.NA.IG4)

confidence level

7/29/05

10/12/05

0.99

0.999

0.9999

-2000% -1500% -1000% -500% 0%

VOD (% of carry)

0.9

risk sloshes across the

confidence level

0.99

capital structure

7/29/05

5/16/05

3/31/05

10/12/05

0.999 (b) SPREAD DELTA NEUTRAL

EQUITY TRANCHE TRADE

0.9999 (CDX.NA.IG4)

-500% -400% -300% -200% -100% 0%

VOD (% of carry)

11

CREDIT SPREAD SENSITIVITY

60 50

1 name 2 names 1 name 2 names

P&L (% of tranche notional)

40 5 names 10 names 40

5 names 10 names

15 names 50 names

15 names 50 names

125 names 30

20 125 names

20

0 Spot Spread Delta

10 Neutral Mezz/Sr

(20)

Spot Spread Delta Trade

Neutral Equity Trade

0

(40)

Default

0

5

-5

10

20

50

-20

-10

100

300

1,000

10,000

20,000

(10)

Default

0

5

-5

10

20

50

-20

-10

100

300

1,000

10,000

20,000

spread shift (bps)

250

1 name 2 names

These positive carry trades have

P&L (% of tranche notional)

150

15 names

125 names

50 names

positive systematic convexity &

100 negative idiosyncratic convexity to

50

Spot Spread Delta

Neutral Strangle

spreads

0

The buy mezzanine protection

(50)

trade also exhibits negative

Default

0

5

-5

10

20

50

-20

-10

100

300

1,000

10,000

20,000

spread shift (bps)

higher than spot

12

(CDX. NA.IG.4, March 31, 2005)

TRANCHE PRICING CORRELATION

30

25

P&L (% of tranche notional)

20

15 The implied

10 correlation sensitivity

5

is also an increasing

0

function of the initial

(5)

(10)

Delta Hedged Equity Tranche (0-3%)

trade carry – like the

Delta Hedged Senior Mez Tranche (7-10%)

(15) Straddle (0-3%, 7-10%) credit sensitivity

(20)

-10% -5% 0% 5% 10% 15% 20%

correlation shift

Correlation Sensitivity

For the straddle, the P&L impact is plotted as

a percentage of equity tranche notional

(CDX. NA. IG.4, March 31, 2005)

13

SELL EQUITY TRANCHE PROTECTION P&L

36% -20%

-25%

cash (% of tranche notional)

35%

-30%

34% -35%

-40%

33%

(a) cash -45%

-50%

32%

-55%

-65%

30%

-70%

3/22/05 4/21/05 5/21/05 6/20/05 7/20/05 8/19/05 9/18/05 10/18/05 11/17/05 12/17/05

3/22/05 4/21/05 5/21/05 6/20/05 7/20/05 8/19/05 9/18/05 10/18/05 11/17/05 12/17/05

5%

0%

total P&L (% of tranche notional)

-5%

-10%

-15%

-25%

sell equity tranche

-30% (c) total P&L

-35%

protection position

-40% (CDX.NA.IG.4)

3/22/05 4/21/05 5/21/05 6/20/05 7/20/05 8/19/05 9/18/05 10/18/05 11/17/05 12/17/05

Delta Hedging Daily Delta Hedging Every 2 Weeks

Delta Hedging Every 2 Months Delta Hedging Every 4 Months

14

SELL EQUITY TRANCHE PROTECTION TOTAL

P&L

5% 180 5% 240%

0% 160

0% 220%

% of tranche notional

average spread (bps)

spread dispersion

-5% 200%

% of tranche notional

-5% 140

-10% 180%

-10% 120

-15% 160%

-15% 100

-20% 140%

-20% 80 -25% 120%

-25% 60 -30% 100%

cross-sectional average spread (right)

-40% 60%

-35% 20

3/22/05 4/21/05 5/21/05 6/20/05 7/20/05 8/19/05 9/18/05 10/18/05 11/17/05 12/17/05

3/22/05 4/21/05 5/21/05 6/20/05 7/20/05 8/19/05 9/18/05 10/18/05 11/17/05 12/17/05

5% 40%

0%

% of tranche notional

-5% 30%

-10%

-15% 20%

Total P&L of sell equity

-20%

tranche protection position

-25% 10% (CDX.NA.IG.4)

-30%

implied equity correlation

-35% (right) 0%

3/22/05 4/21/05 5/21/05 6/20/05 7/20/05 8/19/05 9/18/05 10/18/05 11/17/05 12/17/05

15

REALIZED SPREAD MOVE CORRELATION

40%

35%

realized correlation

(a) Realized correlation versus

30%

time interval (CDX.NA.IG.4)

25%

The standard model 15%

Data: 3/22/05-3/2/07

1 day 2 weeks 2 months 4 months

the cost of hedging time interval

5%

spread dispersion 0%

-5%

-10%

-15%

-30%

tranche protection position

-35%

(CDX.NA.IG.4) -40%

3/22/05 4/21/05 5/21/05 6/20/05 7/20/05 8/19/05 9/18/05 10/18/05 11/17/05 12/17/05

Delta Hedging Daily Delta Hedging Every 2 Weeks

Delta Hedging Every 2 Months Delta Hedging Every 4 Months

16

SPREAD DELTA HEDGER: FANTASY VS REALITY

4% coherent spread moves and

fixed implied correlation

2%

0%

P&L (% of tranche notional)

-2%

actual spreads and fixed

-4% implied correlation

-6%

-8%

real case:

-10% actual spreads &

implied correlation

-12%

-14%

-16%

3/22/05

4/21/05

5/21/05

6/20/05

7/20/05

8/19/05

9/18/05

10/18/05

11/17/05

Impact of spread dispersion and implied correlation fluctuations on P&L of a daily

hedged sell equity protection position on CDX.NA.IG.4 (3/22/2005-11/15/2005)

17

dispersion (%) mean 5yr sprd (bps)

index price (bps)

20

40

60

80

100

120

140

160

180

200

0

1/8/05

2/17/05

3/29/05

5/8/05

6/17/05

7/27/05

Index Price

9/5/05

Dispersion (%)

10/15/05

11/24/05

1/3/06

2/12/06

Mean

3/24/06

5/3/06

6/12/06

7/22/06

Equity Implied Corr

8/31/06

CDX IG4 June 2010

10/10/06

11/19/06

12/29/06

2/7/07

3/19/07

4/28/07

0

2

4

6

8

10

12

14

16

18

20

PRICING CORRELATION-DISPERSION REGIMES

(%)

18

SYNTHETIC CDO PRICING EVOLUTION

CDX IG4 June 2010

%/bps

____Equity tranche upfront/index spread

1 ____Equity tranche implied correlation 24

equity upfront / index spread

0.9

goldilocks no more? 20

0.8

16

0.7

0.6 12

0.5

8

0.4

mining of equity value 4

0.3

0.2 0

3/22/05 7/5/05 10/18/05 1/31/06 5/19/06 9/7/06 12/28/06

idiosyncratic negative credit convexity

event event

19

SYNTHETIC CDO PRICING EVOLUTION

1000

CDX.IG.9 Maturity: 2012

100

10

3_7(bps/yr) 7_10(bps/yr)

10_15(bps/yr) 15_30(bps/yr)

30_100(bps/yr)

1

2007-09-05 2007-09-25 2007-10-15 2007-11-04 2007-11-24 2007-12-14 2008-01-03 2008-01-23 2008-02-12 2008-03-03 2008-03-23

20

SYNTHETIC CDO PRICING EVOLUTION

CDX.IG.9 2012

DATE 3_7(bps/yr) 7_10(bps/yr) 10_15(bps/yr) 15_30(bps/yr) 30_100(bps/yr)

(bps/yr) 500 bps/yr)

3/17/2008 192 68.625 894 480 287.5 134.5 69

March08/

Sept 07 3.7 1.9 6.4 9.9 10.8 8.0 8.6

ratio

protection seller - becomes apparent.

21

PRICING CORRELATION-DISPERSION REGIMES

The market developed a realization of the vulnerability of the sell equity protection

trade to pool idiosyncrasies in May 2005

reaction/over-reaction?

September 2005 saw new heights of richness of equity tranches – together with the

proliferation of leveraged super senior trades

completing capital structure?

At late 2005 and early 2006 levels, an increasing number of market participants

(hedge funds, prop trading) were willing to sell equity protection while banks piled up

senior risk (leveraged super senior)

widespread mining of equity value?

By the start of 2007 the equity correlation had risen to pre May 2005 levels

“fair value” or credit amnesia?

Systematic fears resulted in value slosh up to the senior portions of the capital

structure in late 2007 – early 2008

paranoia? new recognition of systematic negative convexity? Liquidity crisis?

22

ELEPHANT IN THE ROOM

23

OMISSION OF “STANDARD MODEL”

The popular risk-neutral framework of CDO pricing (i.e., static

spread & correlated defaults via some Copula) has not

addressed any theoretical or practical aspect of replication!

– Deltas are found by perturbing the static spread, while valuation is made by

employing static spreads

– Carry of delta hedged trade is fair(?) compensation/cost for what?

– What about replication? How is the embedded spread exchange option

component of an equity tranche priced?

– Impact of spread dispersion and buildup of realized correlation over time

suggests different hedging strategy than a formal 1 bps spread delta

hedging even for short term P&L volatility management for junior tranche

– Systematic convexity of senior tranches not addressed in static spread

framework – spread volatility not even an input

24

OMISSION OF “STANDARD MODEL”

Need to directly address hedging costs and residual risks to discern

relative value and express risk-preference

replication! (pseudo risk neutral eyes wide shut model)

Direct hedge analysis within model of spread with jumps & defaults

with uncertain recovery

– Static hedge analysis/optimization under defaults alone is readily feasible and

insightful

– Market model (and empirical calibration) that couples stochastic spreads and default

needed for dynamic hedge/optimization analysis

of unhedgeable risks!

25

EYES OPEN ASSESSMENT OF VALUE & RISKS

route to valuation while accounting for spread

diffusion, spread jumps, defaults with dependent

uncertain recovery?

Formulate wealth balance of CDO derivative + hedge portfolio

Assess value and hedge ratios that minimize P&L fluctuations under

ensemble of all possible credit outcomes

In the face of residual P&L fluctuations assess pricing based on your

risk preference!

26

OPTIMAL HEDGING

STATIC HEDGING

Plays out a buy and hold assessment of value and risks in a long-

short CDO-hedge portfolio

Sensitive to credit spreads like the standard model as they control

hedging costs

Hedge ratios derived from multidimensional minimization

Carry of optimally hedged portfolio is related to residual risks

DYNAMIC HEDGING

Hedge ratios and value derived from multi-dimensional variational

calculus problem

Simultaneous hedging of spread fluctuations and default is

addressed in this approach

Carry of optimally hedged portfolio is related to residual risks

27

OPTIMAL HEDGING

Change in wealth of CDO trader: W Wtranche Whedge

Hedge error statistical measure:

Establish the hedge and pricing that result in minimizing the

hedge error and a zero expected change in wealth:

minimum & W 0

2 1/ 2

Std. Deviation E[ W W ]

W

Expected

ESF E[ W | W VaR ]

Shortfall

Prob W VaR 1

28

OPTIMAL HEDGING

Static Hedging

Given the spot spreads of the N reference entities determine the tranche

value and the hedge ratios C and i so that the mean change in wealth

over the life of the transaction is zero and the hedge error measure is as

mall as possible.

Dynamic Hedging

Given the spot spreads of the N reference entities determine the tranche

value and the hedge ratio functions C(s1,s2,….,sN), and i(s1,s2,….,sN) so

that the mean change in wealth over every hedge interval t is zero and

the hedge error measure is as small as possible.

29

OPTIMAL STATIC HEDGING RESULTS

80

Sample Problem: Setup 70

ESF0.95

60

30

std dev

2 default models (with identical expected pool loss 20

ESF0.80

0 10 20 30 40 50

and standard deviation): total hedge notional (x tranche)

100

90

80

Structural Variance Gamma (SVG: dashed line) 70

60

40

30

Hedge error minimizing hedge ratio 20

10

0 10 20 30 40 50

Upfront fee that results in a fair bet – on the total hedge notional (x tranche)

average

12

8

Description of un-hedged risks 4

-4

-8

-12

0 10 20 30 40 50

total hedge notional (x tranche)

30

OPTIMAL STATIC HEDGING RESULTS

80

Sample Problem: Price, Hedge Ratio & Error

70

ESF0.95

60

Increasing the hedge notional increases the break-

50

even upfront fee and decreases the no-default carry 40

30

Without any hedge, the breakeven upfront is low, std dev ESF0.80

20

no-default carry is high & P&L uncertainty is high 0 10 20 30 40 50

total hedge notional (x tranche)

100

90

As the hedge notional is increased from zero the 80

70

P&L uncertainty is reduced – it achieves its minima 60

40

30

20

10

0 10 20 30 40 50

The no-default carry is significantly positive at the total hedge notional (x tranche)

12

8

Minimizing the loss tail risk measure requires 4

hedge notionals that eliminate no-default carry and 0

-8

-12

0 10 20 30 40 50

total hedge notional (x tranche)

31

OPTIMAL STATIC HEDGING RESULTS

Sample Problem: Tail Wags the Carry

The tallest spikes in the probability distribution of change in wealth correspond to the no

default carry

Without hedging we have positive no-default carry and tail-losses due to defaults

At 12 x hedge notional the no-default carry is reduced & so are the tail losses due to

default

At 30 x hedge notional the no-default carry is negative and there are tail gains

associated with default

1

0.4

(0 x hedge)

0.35 (0 x hedge)

(12 x hedge)

(12 x hedge) 0.1

0.3 (30 x hedge)

(30 x hedge)

0.25

frequency

frequency

0.2 0.01

0.15

0.1 0.001

0.05

0 0.0001

-100 -80 -60 -40 -20 0 20 40 60 80 100 -100 -80 -60 -40 -20 0 20 40 60 80 100

change in wealth (% tranche) change in wealth (% tranche)

model, bin size = 1% tranche notional, zero mean constraint) 32

OPTIMAL STATIC HEDGING RESULTS

Sample Problem: Optimal Hedging Given Tranche Pricing

• Given the equity tranche upfront what is the optimal hedge ratio?:

maximize market prices hedge error measure:

3.5 1

mean wealth change/hedge error

(0 x hedge)

3 (5 x hedge)

0.1 (10 x hedge)

2.5

mean/ESF0.8

frequency

2 0.01

1.5

0.001

1

0

0 5 10 15 20 25 0.00001

-50 -25 0 25 50 75 100

total hedge notional (x tranche)

wealth change (% tranche)

Sell equity (0%-3%) tranche protection for 60% upfront & 500 bps running

(solid line: PNC; dashed line: SVG)

33

OPTIMAL STATIC HEDGING RESULTS

Sample problem: Optimal Hedging Implied Correlation Skew

50%

45% Variance optimal PNC

ESF0.8 optimal PNC

40%

Variance optimal SVG

implied correlation

30%

25%

20%

15%

10%

5%

0%

equity mezzanine senior

sample problem (0%-3% equity; 3%-7% mezzanine; 7%-10% senior)

34

OPTIMAL STATIC HEDGING RESULTS

Summary

Addresses replication insofar as the hedge ratios accomplish hedge error

minimization and “value” is explicitly related to cost of hedging

Standard model does not address the mechanics of replication

Carry of optimally hedged portfolio is related to residual risks

Standard model is silent on the carry of the “delta-hedged” position

Produces skew even when the objective default measure is described by a

Normal Copula with 1 asset correlation – i.e., skew is endemic to the

mechanics-imperfections-costs of hedging across the capital structure

Standard model attributes correlation skew to parametric complexity rather than directly

to replications costs and/or errors

Employing expected shortfall minimization with random recoveries, the

confidence level can be adjusted to provide “market calibration”

Standard model calibration invokes a rich correlation parameterization with no

reference to hedging mechanics

35

OPTIMAL DYNAMIC HEDGING RESULTS

Need integrated spread-default model to perform analysis

Simple example: CIR spread with spread dependent objective hazard rate:

sk 1 sk ( sk ) t sk t k

36

OPTIMAL DYNAMIC HEDGING RESULTS

Sample Problem: ATM No-Knockout Payer Swaption

te = 0.5 yr; T = te + 5 yr; N = $100,R = 0.4

spot spread = strike spread = 300 bps/yr;

Spread Dynamics: CIR (fitted to historical)

Default Dynamics: obj haz rate = x risk–neutral haz rate

1600

5 yr par CDS spread for GM (bps/yr)

1400

1200

1000

800

600

400

200

0

01/01/02 07/20/02 02/05/03 08/24/03 03/11/04 09/27/04 04/15/05 11/01/05 05/20/06 12/06/06

37

OPTIMAL DYNAMIC HEDGING RESULTS

A Monte Carlo Simulation Approach:

k k+1=T

s 1k 1 , C ( s 1k 1 ) payoff1

default

w1k

s 1k s k2 1 , C ( s k2 1 ) payoff 2

wk2

s k2

s k3 C ( sk3 1 ) 0

ski 20

This path defaults C ( ski 1 ) 0

before t(k)

s kn wkn s kn 1 , C ( s kn 1 ) payoff n

2

such that W tk , tk 1 0 and E[ W W ] is minimized

38

OPTIMAL DYNAMIC HEDGING RESULTS

Sample Problem: ATM No-Knockout Payer Swaption Solution Steps

(1) Generate MC realizations of the state of the credit underlying the swaption – i.e.,

realizations of spread time series and time of default

(2) For each MC path find option value at exercise – in the case of defaults with and

without knock-out that terminal value will be zero

(3) Given the option value at tk+1 (initially tk+1 = te) find functions C(s) and (s) at

time step tk that minimize hedge error for zero expected change in wealth.

This variational problem (the unknowns are functions) is rendered finite dimensional

by using basis functions – following the methodology of Bouchaud et al [2000] applied

to equity derivatives.

(4) Repeat (3) until first time step to find hedge-optimal breakeven option value

sufficiently small time-intervals agree with the risk-neutral framework

framework will delineate costs and errors of the hedging strategy and help express a

risk preference

39

OPTIMAL DYNAMIC HEDGING RESULTS

Sample Problem: ATM No-Knockout Payer Swaption

1

2

8 day

0.8

4 day 8day

1.6

2 day 4day

1 day 0.6

2day

1.2

1day

0.4

0.8

0.2

0.4

0 0

-3 -2 -1 0 1 2 3 -4 -3 -2 -1 0 1 2 3

Probability density of residual wealth change of variance optimal hedged ATM no-

knockout payer swaption under varying hedging intervals

=0 = 0.3 =1

Hedge W W

C 0 0 C 0 0 W

C 0 0

Interval C0 C0 C0

($) ($) ($) ($) ($) ($)

(days) (%) (%) (%)

8 3.77 60.91 16.9 3.80 72.30 47.3 3.81 81.89 60.85

4 3.78 61.42 12.1 3.79 72.47 45.8 3.81 81.73 60.12

2 3.78 61.62 8.7 3.80 73.03 44.1 3.81 81.47 59.79

1 3.78 61.74 6.2 3.80 70.66 42.9 3.82 82.23 59.02

40

OPTIMAL DYNAMIC HEDGING RESULTS

Sample Problem: ATM No-Knockout Payer Swaption

2.2 1

pdf

2

pdf(conditioned on no-default)

probability density function

1.8

0.1

1.6

1.4 = 0.3

1.2 0.01

1

0.8

0.6 0.001

0.4

0.2

0.0001

0

-30 -25 -20 -15 -10 -5 0 5

-5 -4 -3 -2 -1 0 1 2 3

total change in wealth ($) total change in wealth ($)

knockout payer swaption under 1-day hedging interval

41

OPTIMAL DYNAMIC HEDGING RESULTS

Sample Problem: ATM No-Knockout Payer Swaption

400 90

spread 800 90

380

85

hedge notional

spread (bps/yr)

360 700 85

spread (bps/yr)

80

340 600 80

75

320 500 75

70 400 spread 70

300

default event default event

hedge notional

280 65 300 65

260 60 200 60

0 0.02 0.04 0.06 0.08 0.1 0 0.05 0.1 0.15 0.2 0.25 0.3

2 2.75

40

option & hedge P&L ($)

50 2.5

0

option 2.25

20 2

-2 30 hedge

1.75

0 total

-4 10 1.5

1.25

-20 option 1

-6 -10

hedge 0.75

-40 total -8 0.5

-30 0.25

-60 -10 0

-50 -0.25

0 0.02 0.04 0.06 0.08 0.1

0 0.05 0.1 0.15 0.2 0.25 0.3

time (yrs) time (yrs)

MC realization (a) MC realization (b)

knockout payer swaption under 1 day hedging interval with = 0.3 42

OPTIMAL DYNAMIC HEDGING RESULTS

Sample Problem: ATM No-Knockout & Knockout Payer Swaption

No-Knockout Knockout

1 1

pdf pdf

probability density function

0.1 0.1

0.01 0.01

0.001 0.001

0.0001 0.0001

-30 -25 -20 -15 -10 -5 0 5 -20 -10 0 10 20 30 40 50

total change in wealth ($) total change in wealth ($)

E W no default 0 E W no default 0

knockout and knockout payer swaption under 1-day hedging interval with =

0.3

43

OPTIMAL DYNAMIC HEDGING RESULTS

CDS Swaption Dynamic Hedging Key Features

Optimal hedge ratios depend on the real-world propensity of jumps to

default and its interaction with spread dynamics (not just spread volatility)

The irreducible hedging error is dependent on interactions betweens

diffusive movements in spread and jumps and the option payoff structure

Optimal dynamic hedging based on observable spread can reduce the tail

risks associated with defaults from the static hedging values albeit only to an

irreducible level

Ongoing Work

Full term structure of spreads with spread jumps and defaults and random

recovery

Index swaptions hedging, pricing, and risk

Nth to default baskets & CDOs hedging, pricing, and risk

Integrated market-credit risk assessment

44

SUMMARY

redundant derivative!

CDO tranche under spread dispersion, spread jumps, and defaults with

uncertain recovery, it has not resulted in a stable view of how value should

be partitioned across the capital structure.

across the capital structure – market agents experience in real time the

challenges of hedging and tend to adjust valuations in response.

45

SUMMARY

structure) is needed to understand risk-return tradeoffs.

precursor to finding value by addressing the hedge performance over the

ensemble of credit outcomes and minimizing a statistical hedge error

measure.

analysis as well as optimal hedging analysis.

to be an input into a CDO hedging-valuation model.

46

REFERENCES

Bouchaud, J-P., M. Potters, Theory of Financial Risks, From Statistical

Physics to Risk Management, Cambridge University Press, 2000

Douglas, R., Credit Derivative Strategies: New Thinking on Managing Risk &

Return. Bloomberg Press 2007

in CDOs, DefaultRisk.com, April 2006

Management, The Handbook of Structured Finance, by Arnaud

de Servigny and Norbert Jobst, McGraw-Hill 2007

Swaptions, working document, 2007

47

Optimal Static Hedging of Defaults in CDOs

Andrea Petrelli, Olivia Siu, Jun Zhang, Vivek Kapoor

April 2006

Abstract

The optimal static hedging of a CDO tranche position with a portfolio of bonds that constitute the

CDO reference pool is addressed here. The hedge ratio and tranche pricing that result in a fair bet on

the average and minimum hedge error measures are found for synthetic CDO tranches, employing

two default models (1) Reduced form Normal Copula; (2) Structural Variance-Gamma. The

sensitivities of the break-even spread, optimal hedge ratios, and un-hedged risks to the underlying

credits and the CDO structure are illustrated. The relationship between the no-default carry and

residual default risks of hedged CDO tranches are illustrated. In the same framework hedging a bond

with a CDS is also examined. The residual hedge error dependence on recovery uncertainty and

deviation of bond price from par are shown.

I. Introduction

The state of practice of assessing the financial impact of jumps in market variables on derivative

positions is far from ideal: (1) the mechanics of theoretical perfect replication that are the foundation

of pricing models for derivatives are challenged in the face of jumps of random magnitude and

uncertain timing, let alone practical difficulties with replication; (2) many pricing models in practice

are continuous-diffusion-process based and do not entertain jumps (see Cont & Tankov [2004] for an

overview). Controlling the risk profile of derivative trading, however, requires understanding P&L

impacts due to realistic changes in pricing input variables, which can involve sudden moves not

captured by diffusive processes. Furthermore, managing a derivative trading book requires

understanding and anticipating the impact of jumps in basic market variables on more exotic pricing

model inputs. In the face of jumps in basic market parameters, significant segments of market

participants can become risk-aware and risk-averse, and that can manifest as a correction in implied

parameters of pricing models. For example: (1) the 1987 equity market crash and its impact on

volatility skew resulting from a greater recognition of fat tails and heteroskedasticity of return

distributions, and (2) the May 2005 investment grade CDO equity tranche correlation correction

resulting from a recognition of un-priced cost of hedging idiosyncratic spread jumps within the

standard model, as analyzed by Petrelli et al [2006]. All these challenges get compounded when the

derivative references multiple issuers, and its payoff is triggered by jumps alone – in the case of

CDOs, triggered by an issuer state variable switching from no-default to default. This work examines

the basic synthetic CDO contract and how the impact of defaults on a tranche investment might be

offset by taking a position in the reference pool assets.

Not all jumps of issuer state from no-default to default come as surprises. The credit spread revealed

in the CDS market will often advertise distress. For a CDS position, marked to market daily with

prescient knowledge of recovery, the impact of default on the day of default does not have to result in

a significant P&L event if default occurs after the credit spread of that name has already widened

The views expressed here are those of the authors, and do not necessarily represent those of their employers.

significantly. Of course recovery is not perfectly known beforehand and the credit spread of an issuer

itself can have sharp moves en-route to default – which can cause jumps in P&L en-route to default.

Whether default arrives as a shock to market participants or as a gradual deterioration of a credit, the

standard synthetic CDO contract payout occurs only after a default occurs – hence it would be

inconceivable to attempt to devise a vanilla synthetic CDO pricing model without jumps to default.

The standard synthetic CDO model (e.g., Li [1999]) considers jumps to default, given the reference

asset spread term structures, and does not consider diffusion or jumps in the credit spreads. In that

model, a risk-neutral description of time-to-default is effected by fitting marginal default

probabilities to observable credit spreads. The joint distribution of issuer default-time is described by

a Copula approach with the implied asset correlation found by fitting a modeled CDO tranche price

to market. While that model assesses risk-neutral expectations of tranche losses due to jumps to

default to value the tranche - it is silent on replication-hedging in the face of defaults or diffusion or

jumps in the spread. This paper assesses the costs and irreducible errors of static hedging the

financial impact of defaults in Synthetic CDO contracts.

The CDO hedging-pricing problem is cast in a framework of continuous premium payments and

constant interest rates for the sake of simplicity of notation. The significant methodological

simplification made here is that of single-period static hedging, as opposed to dynamic hedging. The

reasons for this simplification are: (1) little work has been done on direct hedge performance of CDO

tranches, and prior to grappling with the dynamic hedging problem, an understanding of the static

hedging problem is needed; (2) the ideal endpoint of dynamic hedging in pursuit of assessing the

unique value of a derivative contract, a risk-free replicating portfolio, is not even theoretically

achievable in the face of jumps of uncertain timing and magnitude, so residual hedge errors are

important to understand, even in a static hedging framework, which will remain an important subset

of the dynamic hedging problem; (3) a dynamic analysis requires a coupled model of spreads and

defaults which is beyond the scope of this paper.

Reference Pool

Consider a CDO reference pool of n bonds of notional ni (1 ≤ i ≤ n ). The total reference pool

notional N = ∑i =1 ni . These reference bonds can default, and in the event of default recover a fraction

n

Ri of notional. The pool default loss process, pL(t), is a superposition of delta-functions centered

at τ i , the time to default of reference bond i, and with each default contributing a loss of

(1 – Ri)ni:

n

pL(t ) = ∑ ni (1 − Ri )δ (τ i ) (1)

i =1

The cumulative pool loss (cpL) and recovered amount (cpR) are superposition of Heaviside functions

corresponding to a time integral of the loss and recovery process:

⎧0 if τ i > t ⎫

n n

⎪ ⎪

cpL(t ) = ∑ ni (1 − Ri )I i (t ) ; cpR(t ) = ∑ ni Ri I i (t ) ; I i (t ) = ⎨ ⎬ (2)

i =1 i =1 ⎪ 1 if τ ≤ t ⎪

⎩ i ⎭

2

Tranche

We adopt a synthetic CDO tranche specification, where the lower strike (k1), upper strike (k2),

upfront payment fraction (u) and running spread (s) define the tranche. The tranche notional as a

function of time, tn(t) = k 2* (t ) − k1* (t ) , where:

In this vanilla rendition of synthetic CDOs, the tranche amortizes from the bottom due to defaults,

with associated contingent payments made by protection seller. A tranche can also amortize from the

top, due to recoveries associated with defaults, to ensure that the outstanding tranche notional

matches the un-defaulted reference pool notional. The standardized synthetic CDOs based on credit

indexes follow aforementioned amortization rules.

The present value of the cash-flows for the tranche investor (i.e., the tranche default protection seller)

consists of received upfront payments and tranche spread on outstanding tranche notional and

outgoing default-contingent payments:

T

ΔWtranche = u (k 2 − k1 ) + s ∫ tn(τ ) exp[− rτ ]dτ + ∫

T

( )

d k1* (τ ) −rτ

e dτ (4)

0 0

dτ

In (4) the risk-free discount rate r is taken as a constant and the tranche premium is assumed to be

paid continuously. These simplifications are made for compactness of notation – it is feasible to

relax these assumptions to handle a term structure of interest rates and discrete premium payments

without significant extra computational effort.

We will attempt to hedge the default risk of a CDO tranche with a position in the reference bonds of

notional hi, market price equal to f i × hi , and coupon ci. The change in wealth of this bond portfolio is

given by:

n

ΔWbond = ∑ ΔWi (5)

i =1

⎧ ⎧ T

⎫⎫

⎪no default over t ∈[0,T ] : hi ⎨− f i + exp[− rT ] + ci ∫ exp[− rτ ]dτ ⎬⎪

⎪

⎪ ⎩ 0 ⎭⎪⎪

ΔWi = ⎨ ⎬

(6)

⎪ ⎧⎪ τi

⎫⎪ ⎪

⎪ default at τ i ∈ (0, T ] : hi ⎨− f i + Ri exp[− rτ i ] + ci ∫ exp[− rτ ]dτ ⎬ ⎪

⎩⎪ ⎪⎩ 0 ⎪⎭ ⎭⎪

The hedging could be done by a portfolio of CDS on the CDO reference issuers, as is customary with

much synthetic CDO trading activity. Hedging default risk with a CDS is theoretically identical to

hedging with a par bond under certain conditions (see Appendix-A).

3

Optimal Static Hedging

We attempt to find the bond hedge notionals (hi) and tranche pricing (u and s) such that the change in

wealth of the hedged portfolio consisting of bonds and a CDO tranche is zero on the average, and a

certain hedge error measure Θ is as small as possible:

ΔW = 0 (8)

Minimize [ Θ ] (9)

As ΔWbond accounts for the coupon and market-value of the CDO reference pool bonds, the credit

spreads of the reference pool is an explicit input for our analysis. Both the average change in wealth

and the hedge error measure are quantified based on an objective probability measure which in

principal needs to be inferred from empirical observations of the market risk drivers that are defaults

in this study. This approach is parallel to the risk neutral derivative pricing approach insofar as if a

perfect hedging strategy exists then the hedge error measure will become identically zero and the

price that is fair on the average will become the unique arbitrage-free price that is independent of the

objective probabilistic measures. An illustration of that is provided in Appendix-A in the context of

hedging a par bond with a CDS, where under certain conditions static hedge optimization analysis

results in (1) pricing that is identical to the risk neutral approach, and (2) a perfect theoretical hedge,

independent of the real-world probabilistic description of the risk driver (Appendix-A).

The reader is referred to Leungberger [1998] for a basic exposition of optimal static hedging. The

work of Bouchaud and co-workers on variance optimal hedging (Bouchaud and Potters [2003]) and

expected shortfall optimal hedging (Pochart and Bouchaud [2003]) illustrates hedge optimization in a

dynamic multi-step framework. Bouchaud and co-workers directly address hedging-replication and

its limits as a pre-cursor to pricing a derivative without making special assumptions on market

stochastic dynamics that enable hedge error elimination (e.g., Brownian motion without jumps).

Bouchaud et al do not formally invoke the existence of a risk-neutral measure without a

demonstration of theoretical mechanics of replication. The optimal hedging approach has the

attraction that a delineation of a hedging strategy and risk-assessment takes place en-route to

valuation. These insights come at the cost of extra effort in (1) solving the optimization problem to

assess hedge ratios, and (2) specifying real-world probabilistic measures associated with the

underlying asset.

We employ two hedge error measures:

Standard Deviation ( (

Θ = σ ΔW ≡ E[ ΔW − ΔW ] ) )

2 1/ 2

(10)

4

At high confidence levels ESFα & VaRα captures aspects of “tail risks.” The standard deviation

measure quantifies “body risk” insofar as it is driven from the central regions of the wealth change

distribution (whose expectation we are constraining to zero in search of a fair price).

If hedging error can be eliminated altogether and perfect replication is theoretically possible then the

model pricing results are not sensitive to the choice of hedge error minimization objective function,

which is the central attraction of the risk-neutral pricing theory’s postulate of derivative contract

value being equal to the cost of perfect replication. However, even when replication is not

theoretically possible, fitting prices to purported risk-neutral pricing models takes place in the day-to-

day practice of marking to market/model complex derivative contracts. In contrast to that practice,

our goal is to explicitly illustrate a specific static hedging scheme and the irreducible residual risks

associated with it, albeit under idealized conditions.

No-Default P&L

In the event no default occurs the change in wealth of a CDO tranche trader is given by:

n ⎧ T

⎫

ΔWbond ,no−default = ∑ hi ⎨− f i + exp[− rT ] + ci ∫ exp[− rτ ]dτ ⎬

i =1 ⎩ 0 ⎭ (14)

T

ΔWtranche,no−default = u (k 2 − k1 ) + s (k 2 − k1 )∫ exp[− rτ ]dτ (15)

0

This no-default wealth change can be divided by the transaction tenor T to express a no-default carry-

rate:

ΔWno−default

no − default carry ≡ (16)

T

Computational Framework

For the CDO hedging problem there is a vector of bond hedge notionals hi and tranche prices that are

unknowns. The zero mean change in wealth constraint (Equation (8)) results in expressing the

tranche price in terms of the bond hedge notionals. The bond hedge notionals are then found by

numerically minimizing the hedge error measure. The hedge ratio and average break-even price are

directly coupled as we are simultaneously enforcing the two constraints in equation (8) and (9).

A Monte-Carlo (MC) simulation of the time-to-default for the CDO reference pool is performed and

the time-to-defaults are used to define unit face bond wealth change measures as well as unit spread

tranche measures for each MC path. These intermediate measures are stored in computer memory,

and the optimization problem is solved by performing 1 MC simulation (100,000 realizations). The

doubling and halving of the number of MC realizations does not materially change the results of the

optimization problem presented here.

5

III. Default Models

Many of the challenges of understanding CDO risk-return are not tied to any particular model

specification of portfolio credit. Even if one postulated the simplest probabilistic default model, e.g.,

a model with independent and statistically homogeneous Poisson default arrivals, the nonlinearity

introduced by tranching coupled with the multidimensional nature of the problem and the default

event driven payoffs makes the hedge analysis problem non-trivial. Indeed, CDO tranche payoff

replication by bond/CDS positions and its cost and limitations have not been established for any

simplified case of a CDO contract or an Nth-to-default basket contract, even though the standard

CDO model takes risk-neutral expectations of cash-flows under defaults alone without any spread-

time-dynamics. A notable exception is the work of Sircar and Zariphopoulou [2006] that addresses

CDO economics without formally invoking risk-neutrality, similar to this paper. Sircar &

Zariphopoulou [2006] address CDOs from a utility valuation perspective for a long only CDO

investor, whereas here we analyze the cost of hedging and hedge performance of long-short CDO

tranche hedging strategy.

We adopt two approaches for the objective measure model of the default process over which we seek

to hedge P&L uncertainty arising from default events: (1) Reduced form model of Poisson arrival of

defaults with Normal Copula based dependence; (2) Structural model of defaults with Variance

Gamma firm value drivers. The first approach ushered in a mark-to-model dynamic and is taking

hold in the accounting of synthetic CDO trading P&L. The latter approach provides an alternative

and the potential of integrated spread evolution and default modeling as well as integrating credit and

equity modeling.

The structural Variance Gamma (VG) applications in CDOs heretofore have been in developing a

risk neutral description of marginal defaults and fitting VG dependence parameters to observed

tranche prices – much like the reduced form approaches, albeit invoking the firm value process en-

route to fitting parameters to observable CDS and tranche pricing (e.g., Cariboni & Schoutens [2004],

Luciano and Schoutens [2005], & Moosbrucker [2006]). Joshi and Stacey [2006] have recently

adopted the gamma process business time-information arrival concept of the VG approach in a

reduced form modeling framework and shown the ability of that model to more naturally fit the

correlation skew. With these works, there are now a handful of attractive practical approaches that

provide workable portfolio loss descriptions that can be fit to standard index tranche prices in the risk

neutral framework. These approaches provide attractive alternatives to the base correlation approach

of fitting prices and marking to market non-standardized tranches.

Here we employ a flat default hazard rate and a uniform asset correlation. Time to default is

simulated using a latent variable single factor approach with both the market and idiosyncratic drivers

of randomness taking place via standard Normal independent random variates. The specification of

the default hazard rate can be made based on rating or default probability estimates provided by

KMV or Kamakura. The hedging strategy and related risk management mandate requires making

assumptions on the objective default probabilities and correlations.

Here we employ a structural modeling approach to simulating time-to-default. The latent variable

“firm-value” is simulated as a VG process, i.e., a Brownian motion sampled over a random

“economic-time increments” (i.e., Gamma process). We calibrate the VG parameters to fit the

6

assumed objective default probability of reference pool issuers and control the dependence of defaults

by controlling the correlation between the Gamma processes (economic-time elapsed) and the

Brownian motion for the different issuers. To facilitate an “apples to apples” comparison between

the two models we chose calibration parameters that match the T period first and second statistical

moments of the reference pool loss distribution (see Appendix-B).

In both these approaches we are simulating defaults based on a presumed real-world default rate as

well as dependence structure. The challenge of creating a forward looking objective stochastic

default model conditioned on the present state of the world requires understanding historical default

experience, spreads, ratings and downgrades en-route to default. Here we simply employ two generic

descriptions of real-world measure defaults and examine the question of hedging P&L in the face of

defaults over the term of the transaction.

Model Parameters

Tenor T = 5 yrs

Interest rate r = 5%/yr

Pool Information

Number of issuers n = 125

Reference notional ni = $0.8 m ∀ i

Total pool notional N = $100 m

Bond coupon ci = 5.78 %/yr ∀ i

Bond unit price fi= 1 (par bond) ∀ i

Recovery rate Ri = 0.3 ∀ i

Hazard rate λi = 0.65 %/yr ∀ i

Asset correlation 25%

GBM drift μi = 0.0 (1/yr) ∀ i

GBM volatility σ i = 0.20 (1/yr1/2 ) ∀ i

Gamma volatility ν = 2 yr

Default threshold ϖ = 0.39152623

GBM dependence parameter β = 0.24105

Gamma dependence parameter κ =1

Tranche Information

Name k1 k2 upfront fixed running (bps/yr)

Equity 0 3 yes 500

Mezzanine 3 7 no -

Senior 7 10 no -

7

The parameters for the two default models result in identical first two statistical moments for T period

portfolio loss (see Appendix-B). For the static hedging analysis we employ these models to simulate

time-to-default for the issuers in the reference pool in a Monte-Carlo setting. Given an ensemble of

reference pool default times we determine the hedge error measure as a function of the bond hedge

notional. For an initially homogeneous pool the results can be displayed easily in plots. The optimal

hedge solution can be found for an inhomogeneous pool also using the downhill simplex approach.

80 100

90

70

hedge error (% tranche)

ESF 0.95 80

60 70

60

50

50

40 40

30

30

std dev 20

ESF 0.80

20 10

0 10 20 30 40 50 0 10 20 30 40 50

total bond hedge notional (x tranche) total bond hedge notional (x tranche)

(a) (b)

12

no-default carry (%tranche/yr)

-4

-8

-12

0 10 20 30 40 50

total bond hedge notional (x tranche)

(c)

Figure 1. Sell equity (0%-3%) tranche protection (solid-lines: PNC; dashed-lines: SVG)

In the sell-equity tranche protection trade, increasing the hedge notional increases the break-even

upfront (Figure 1b), to be able to pay for the hedge and still be make a fair bet (based on an

expectation of wealth change). The no-default carry also decreases as the bond hedge notional

increases (Figure 1c). On the extreme end of there being no hedge, the breakeven upfront is low, the

no-default carry is high, and the P&L uncertainty measures are high (Figure 1a). As the hedge

notional is increased the P&L uncertainty is reduced (Figure 1a). The body risk measured by the

wealth change variance achieves its minima at a much smaller hedge ratio than expected shortfall

measures corresponding to tail losses at confidence levels higher than 80%.

As the total bond hedge notional increases (relative to the un-hedged position) the losses are reduced

relative to the no-hedge case: the variance measure is reduced on account of diminished losses in the

8

events of default. The variance measure penalizes gains as well as losses and therefore achieves its

minima and then sharply increases with hedge notional at smaller hedge notionals (~ 10 × tranche

notional) compared to the tail risk measures. Increasing the hedge notional after a point starts to

create more scenarios where there are losses on account of cost of hedging in scenarios that do not

experience defaults and there are gains in the events of significant defaults. The no-default carry is

significantly positive at the hedge notional that minimizes the wealth change variance. In contrast,

minimizing the loss tail risk measure requires hedge notionals that eliminate no-default carry and can

result in negative no-default carry, which is the cost of limiting tail default loss, and therefore require

a higher upfront price on the tranche to be on the average a fair bet.

0.4 1

(12 x hedge) (12 x hedge)

0.3 0.1

(30 x hedge) (30 x hedge)

0.25

frequency

frequency

0.2 0.01

0.15

0.1 0.001

0.05

0 0.0001

-100 -80 -60 -40 -20 0 20 40 60 80 100 -100 -80 -60 -40 -20 0 20 40 60 80 100

change in wealth (% tranche) change in wealth (% tranche)

(a) (b)

Figure 2. Sell equity (0%-3%) tranche protection wealth change distribution (SVG model,

bin size = 1% tranche notional)

The tall spikes in the probability distribution of the change in wealth shown in Figure 2 correspond to

the no default carry. The unhedged wealth change distribution is quite disperse because of

significant positive no-default carry and tail-losses due to tail default risk. At a 12 × hedge notional

the tail losses are significantly less than the unhedged case, and the spike associated with the no

default carry occurs at a smaller fraction of the tranche notional. A further reduction of tail losses

occurs at a 30 × hedge notional, which is associated with a negative no-default carry. The wealth

change variance for the 30 × hedge notional case is higher than the 12 × case due to the infrequent

wealth change gains associated with a large number of defaults – i.e., the wealth gain tail (right tail)

as opposed to the wealth loss tail that controls the variance in the unhedged case. Figure 2 provides a

direct display of the no-default carry versus tail default risk tradeoff embedded in the sell equity

tranche protection position.

9

Sell Mezzanine Tranche Protection

60 1000

900

50

hedge error (% tranche)

40

700

std dev

30 600

500

20

400

10

300

0 200

0 5 10 15 20 0 5 10 15 20

total bond hedge notional (x tranche) total bond hedge notional (x tranche)

(a) (b)

3

no-default carry (% tranche/yr)

2

1

0

-1

-2

-3

-4

-5

-6

0 5 10 15 20

total bond hedge notional (x tranche)

(c)

Figure 3. Sell mezzanine (3%-7%) tranche protection (solid-lines: PNC; dashed-lines: SVG)

0.8 1

(0 x hedge)

0.7 (0 x hedge) (8 x hedge)

(8 x hedge) 0.1 (20 X hedge)

0.6

(20 X hedge)

0.5

frequency

frequency

0.01

0.4

0.3 0.001

0.2

0.0001

0.1

0 0.00001

-100 -80 -60 -40 -20 0 20 40 60 80 100 -100 -80 -60 -40 -20 0 20 40 60 80 100

wealth change (% tranche) wealth change (% tranche)

(a) (b)

Figure 4. Sell mezzanine (3%-7%) tranche protection wealth change distribution (SVG

model, bin size = 1% tranche notional)

10

In contrast to the sell equity protection trade, for the sell mezzanine protection trade, the hedge

notional that minimizes the wealth change variance (Figure 3a) is associated with a negative no-

default carry (Figure 3c). The expected shortfall minimizing hedge ratios associated with confidence

levels equal to or higher than 80% vary much more with confidence levels compared to the equity

tranche sell protection position.

The hedge notional that minimizes the expected shortfall at low confidence levels can be small

because the subordination affords this tranche no defaults in many more scenarios than the equity

tranche. In other words, if at a certain low confidence level, the defaults on the pool do not hit the

mezzanine tranche, then minimizing expected shortfall at that confidence level may involve not

hedging at all. However as the confidence level is raised, the expected shortfall minimizing hedge

ratios start to become large and exceed the hedge ratio corresponding to the minimum wealth change

variance, similar to the equity tranche example of Figure 1. The minimum hedge error for the

mezzanine tranche sell protection position is however smaller than the corresponding measures for

the sell equity protection trade.

40 600

35

break-even spread ((bps/yr)

500

hedge error (% tranche)

30

400

25

ESF 0.95

20 300

std dev

15

200

10

100

5

ESF 0.80

0 0

0 2 4 6 8 10 12 0 2 4 6 8 10 12

total bond hedge notional (x tranche) total bond hedge notional (x tranche)

(a) (b)

1.5

1

no-default carry (% tranche/yr)

0.5

0

-0.5

-1

-1.5

-2

-2.5

-3

-3.5

-4

0 2 4 6 8 10 12

total bond hedge notional (x tranche)

(c)

Figure 5. Sell senior tranche (7% - 10%) protection (solid-lines: PNC; dashed-lines: SVG)

11

The sell senior protection trade is qualitatively similar to the sell mezzanine tranche protection. Note

that the 80 percentile expected shortfall minimizing hedge ratio is in fact not hedging at all in this

example (Figure 5a). This is because there is less than a 20% chance of the pool losses exceeding 7%

of the pool notional over 5 years. Like the sell mezzanine tranche protection position and unlike the

equity tranche trade, the variance minimizing hedge ratio is associated with a negative no-default

carry for the sell senior tranche protection position.

1

1

(0 x hedge)

(0 x hedge) (6 x hedge)

0.8 (6 x hedge) 0.1 (12 x hedge)

frequency

(12 x hedge)

frequency

0.6

0.01

0.4

0.2 0.001

0

0.0001

-100 -75 -50 -25 0 25 50 75

-100 -75 -50 -25 0 25 50 75 100

change in wealth (% tranche) change in wealth (% tranche)

(a) (b)

Figure 6. Sell senior (7%-10%) tranche protection wealth change distribution (SVG model,

bin size = 1% tranche notional)

In the examples above the asset correlation was set to 25% to find the optimal static hedge and the

breakeven price corresponding to that optimal hedge for the Normal Copula model. For the VG

structural model we have chosen model parameters to produce the same first two statistical moments

of portfolio loss over the term of the CDO transaction as produced by the Normal Copula model. We

now “calibrate” the standard Normal Copula risk neutral model to find the implied correlation that

reproduces the optimal static hedge breakeven price.

To assess the tranche specific implied correlation we use the tranche specification of previous sample

calculations. As the variance optimal static-hedge breakeven pricing is quite different from ESF

optimal breakeven pricing we get large differences between the implied correlations found from these

two measures. However even for this hypothetical example with a homogeneous pool one gets an

implied correlation skew across the tranches. The ESF0.8 optimal implied correlation skew lies below

the variance optimal pricing correlation and bears greater qualitative resemblance to investment grade

index tranche market skews.

The ESF0.80 optimal skew for the PNC model is not qualitatively distinct from the SVG model –

remembering that we had chosen the parameters of these models to produce the same first two

statistical moments of portfolio loss distributions. While it is common to invoke complexity of the

portfolio default loss process to “explain the skew,” our results show that the simple invocation of a

12

default risk-averse tranche protection seller-hedger (say, for example, seeking to minimize her

ESF0.80) and imperfections in replicating the CDO tranche contract under defaults (even for PNC

model of defaults) support the existence of a skew.

50%

Variance optimal PNC

45%

ESF0.8 optimal PNC

40% Variance optimal SVG

35%

implied correlation

30%

25%

20%

15%

10%

5%

0%

equity mezzanine senior

Figure 7. Optimal static default hedge break-even price implied correlation for sample

problem shown in Figures 1-3.

While refining the modeling of portfolio loss distributions is potentially important and certainly

interesting, examining limitations of replication and explicitly assessing hedging strategies is

essential in discerning relative value. While this work accounts for the initial credit spread of the pool

references, as they are a part of the trader’s wealth change, it does not consider the temporal

evolution of the credit spreads. Further progress in this area requires a coupled spread-default model.

The consideration of spread changes and dynamical changes in actuarial default probabilities will

alter the hedging picture from the one presented here. To the extent default occurrence is tied to

spread movements, dynamically hedging spread movements will result in decreasing the wealth

change uncertainty due to defaults. However, to the extent defaults can happen suddenly, spreads can

jump, and the realized coherence of spread moves can vary randomly, there is no reason to believe

that the wealth change distribution width will shrink to zero under dynamic hedging, as would be the

case with perfect replication. The complex reality of multi-name coupled spread movements and

defaults introduces further risks, like uncertainty of the realized correlation of spread moves, and

does not change the irreducible nature of default risk associated hedging errors.

In CDO vernacular the Normal Copula risk neutral approach is sometimes described as being

analogous to the Black-Scholes pricing approach. Where the underlying follows geometric Brownian

motion, the classical Black-Scholes approach creates a replicating portfolio en-route to assessing

derivative contract value. In contrast there is no non-trivial CDO tranche valuation problem where

the contract payoff can be replicated by holding positions in the reference assets. The Normal Copula

model of jump to defaults in the context of CDO tranche contracts is sufficiently complex to thwart a

perfect replication strategy. Insofar as CDO valuation models do not address hedging and

replication, they are simplistic compared to the classical equity option pricing models.

13

A general criticism or challenge of the direct hedging cost and hedging error approach to pricing a

derivative contract is that (1) it requires a specification of the “objective” probabilistic description

(including parameter values) of the underlying processes, and (2) in the absence of a perfect hedge it

also requires a specific hedge error function. From the point of view of understanding a trading

book’s risk-reward profile, point (1) is without much merit as a criticism because there is no wishing

away the objective measure of market variables in derivative contracts where replication is not even

theoretically feasible and dealing with un-hedgeable risks requires having available historical facts

and an opinion on market events that can create losses. Point (2) raises the issue of what to make of a

market agent’s capricious choice of hedge error function. From a proprietary trading point of view,

having a risk-reward criterion is natural and it is unnecessary for that to conform to the “market.” An

acceptance of the reality of irreducible replication errors should be accompanied by an acceptance of

the possibility that rational modeling approaches can justify a range of model prices. Under liquid

trading conditions and standardized contracts, the prevalent price will be observable and may reflect

the specific risk appetites and risk preference of particularly active market segments. Where the

derivative contract is one of a kind (bespoke), the accounting imperatives to enforce uniformity in

marking are being addressed by calibrating implied parameters of convenient risk-neutral models

using, say, a “base-correlation” inferred from standardized tranches to mark to model bespoke

transactions. Notwithstanding the challenges of implementing accounting rules governing illiquid

derivative positions, the elucidation of unhedgeable risks is a prerequisite of effective risk

management. The market dynamics are likely to be strongly dependent on: (1) at what execution

price a derivative trading strategy risk-reward makes sense for a market agent seeking to participate

in the product; (2) given the derivative contract price, which hedging strategy is optimal for a market

agent.

What is a market agent to do if replication of a structured credit derivative contract is fundamentally

infeasible? The break-even price then only captures the cost of hedging on the average – found by an

expectation taken over an assumed and/or empirically motivated real world probability distribution.

As such one should then be prepared for the actual price to deviate from this average as market

agents build in their different risk aversion preferences and have different views on the objective

probabilistic measures for different pricing variables. Certainly for CDO tranche trading, hedge error

elimination in a dynamic framework has not been shown. Here we have shown the irreducible hedge

errors that arise due to defaults in the framework of static hedging. Much of portfolio management,

trade strategy development and risk management must contend with the realities of imperfect

replication due to both practical reasons (e.g., transaction costs) as well as fundamental theoretical

reasons (e.g., jumps, fat-tails, multi-asset options).

Let us say that a market agent is seeking to minimize a P&L uncertainty measure and the tranche

pricing that is associated with minimizing the hedge error measure and which results in a fair bet on

the average is “x” but the market is trading at “y”. That market agent has an option to (1) not

participate in the market in the face of irreducible hedge errors and pricing that does not correspond

to minimizing hedging error or (2) devise a hedging strategy around the market that trades at “y.” As

the derivative pricing model’s goal of eliminating P&L hedging error while making fair bets cannot

be met, we explore the alternative of maximizing expected change in wealth per unit hedge error

measure:

14

⎡ ΔW ⎤

Maximize ⎢ market prices⎥ (17)

⎢⎣ Θ ⎥⎦

In the face of unhedgeable risks, we assume that the market agent seeks to maximize expected P&L

per unit hedge error measure given the observable market price. Practically speaking, the hedge error

measure constraining the market agent can be her risk-appetite, risk capital-regulatory constraints

and/or aspirations for solvency. This optimization problem can be posed many different ways

depending on the precise circumstances of the market agent.

For the model problem analyzed here we consider optimally hedging a sell equity tranche protection

position. Figure 1 depicts the results for pricing and hedging that result in a fair bet and the minimum

hedge error measure. Now let us assume that the market is pricing the tranche at 60% upfront and

500 bps running – then ask the question: What is a market agent to do? For agents seeking to

maximize expected P&L per unit P&L uncertainty measures, we depict the ratios of expected wealth

change and its uncertainty in Figure 8.

3.5

mean wealth change/hedge error

2.5

mean/ESF0.8

2

1.5

0

0 5 10 15 20 25

total bond hedge notional (x tranche)

Figure 8. Sell equity (0%-3%) tranche protection for 60% upfront and 500 bps

running (solid line: PNC; dashed line: SVG)

If the market agent can find alternative investments with greater expected rewards per unit risk, then

he is likely to not participate in this particular market at the current pricing levels. If the market agent

finds the expected rewards per unit risk to be very attractive, then there will be an active bid from her

to sell equity protection. If many market agents start sharing this view on the equity tranche, that will

tend to drive the upfront payment on the equity tranche down, with all else remaining equal.

The wealth change distribution in Figure 9 shows quite clearly the reduction of the no-default carry

with an increase in the hedge notional and a reduction in tail losses. However beyond a certain point,

an increase in hedging while reducing no-default carry, results in a lower marginal impact on tail

losses – e.g., in going from a hedge notional of 5 × tranche notional to 10 × tranche notional.

Therefore the ratio of the expected wealth change divided by the tail loss measure ESF0.8 achieves its

maximum value around a hedge notional of 5 × tranche notional, in Figure 8. The body risk measure

15

of wealth change standard deviation continues to decrease as the hedge notional increases from 5 to

10 × tranche notional. Therefore the ratio of the expected wealth change to wealth change standard

deviation achieves its maximum at a higher hedge notional than the ratio of the expected wealth

change to ESF0.8 as shown in Figure 8.

0.4 1

(0 x hedge)

(0 x hedge)

0.35 (5 x hedge)

(5 x hedge) 0.1 (10 x hedge)

0.3 (10 x hedge)

0.25

frequency

0.01

frequency

no-default carry

0.2

0.15 0.001

0.1

0.0001

0.05 tail-losses

0 0.00001

-50 -25 0 25 50 75 100 -50 -25 0 25 50 75 100

wealth change (% tranche) wealth change (% tranche)

(a) (b)

Figure 9. Wealth change distribution for sell equity (0%-3%) tranche protection for

60% upfront and 500 bps running (SVG model, bin size = 1% tranche notional)

Even more prevalent than the sell equity tranche and hedge trade is the buy mezzanine protection

(from real money investor) and hedge trade. This is because many mezzanine investors are

historically not hedgers. Hedging of a mezzanine or a senior tranche position is often done by the

broker-dealer or hedge fund that purchased protection on either a customized pool or in the

standardized tranche market. In this trade the act of hedging by the protection purchaser by going

long the underlying credits helps pay for the premium needed to purchase protection. The hedging of

a purchase mezzanine tranche protection position balances expected cash-flows and also results in the

popular positive carry without any spot delta in the standard CDO model. The hedging of the

purchase default protection position also creates tail loss risk that increases with no-default carry, as

depicted in Figure 10.

Hedging a sell equity tranche protection position with a purchase senior tranche protection position

(i.e., a straddle), depicted in Figure 11, is quite different from hedging it with a short position in the

underlying assets (Figure 9). As the senior tranche protection does not result in any payoff to the

protection purchaser until the pool losses exceed the senior tranche lower strike, there is only a very

modest reduction in the left tail (losses) of the wealth change distribution in going from no hedging to

hedging by purchasing protection on 1 to 2 times the equity tranche notional. In fact as the hedge

notional is increased, the right tail of the distribution becomes more pronounced, associated with the

rare events in which there are enough losses in the pool for the senior tranche to result in a payoff to

the protection purchaser.

16

0.8 1

0.7

(0 x hedge) (0 x hedge)

0.1

0.6 (5 x hedge) (5 x hedge)

(10 x hedge) (10 x hedge)

0.5

frequency

frequency

0.01

0.4

0.3 0.001

0.2

0.0001

0.1

0 0.00001

-100 -75 -50 -25 0 25 50 75 100 -100 -75 -50 -25 0 25 50 75 100

wealth change (% tranche) wealth change (% tranche)

(a) (b)

Figure 10. Wealth change distribution for purchase mezzanine (3%-7%) tranche protection for 260

bps running (SVG model, bin size = 1% tranche notional)

0.4 1

(1 x hedge) 0.1 (1 x hedge)

0.3 (2 x hedge) (2 x hedge)

0.25 (5 x hedge) (5 x hedge)

frequency

0.01

frequency

0.2

0.15 0.001

0.1

0.0001

0.05

0 0.00001

-50 -25 0 25 50 75 100 -100 -50 0 50 100 150 200 250 300

wealth change (% tranche) wealth change (% tranche)

(a) (b)

Figure 11. Wealth change distribution for sell equity (0%-3%) tranche protection for 60% upfront

and 500 bps running and purchase senior (7%-10%) tranche protection for 60 bps running to hedge.

(SVG model, bin size = 1% equity tranche notional)

17

While carry versus credit risk tradeoff is central to structured credit derivative structuring and product

evolution, pricing is potentially far from any theoretical perfect hedge or even optimal hedge situation

because (1) it is unclear what the perfect replicating strategy is in the face of multi-name contracts

whose payoffs are driven by jump events, as elaborated here; (2) historically many structured credit

market agents are not hedgers. Consider a market agent who can only be long investment grade or

higher rated bonds (including CDO tranches) and who is not in a position to include a short in the

underlying reference assets to optimize a portfolio. For such a narrowly framed long-only agent, the

pertinent choices are comparing a long position in the underlying assets of the CDO versus being

long risk by being a tranche investor. Cost of funding and the differences in leverage afforded to this

long-only agent in pursuing his long-only strategy are likely to be decisive factors rather than any

hedge optimization-replication arguments pursued in this paper. In this circumstance there can be a

hedger on one side of the trade and a long only-investor on the other side.

The CDO trader-hedger may also be narrowly framed in a different way from the long only investor.

She may be constrained by a regulatory risk-management and a historical limit structure that pre-

dates CDO trading. Despite the revolution in structured credit products, traditional top-down risk

management frameworks are poorly suited to interact with the fabric of CDO tranche trading risk-

return. These frameworks generally ignore the CDO capital structure and associated credit non-

linearity and are aften driven solely by spot spread delta exposure and/or some ad-hoc notional limit,

both of which are grossly insufficient to describe any CDO strategy risk return. The CDO trader-

hedger could be hedge funds without such narrow and inadequate but binding constraints, but with a

variety of time-horizon dependent risk-appetites, depending on promised lock-up periods and

liquidity, and funding costs and margining requirements. The fitted implied parameters of any

pricing model, in the face of such a variety of market agents, should be expected to reflect demand-

supply and liquidity flows in addition to any specific hedging strategy and estimated cost of

replication and irreducible risks in replication.

V. Summary

Static hedging of default risks of a sell CDO tranche protection position was analyzed in this paper.

This framework seeks to explicitly assess the tranche value based on the cost of attempted replication

– along with assessing irreducible hedging errors due to the complexity arising from the multi-name,

jump dependent payoff contracts of CDOs. The hedge error measures of wealth change variance and

expected shortfall were considered. The bond/CDS hedge notional that minimizes the hedge error

measures for a sell tranche protection position was demonstrated for two different multi-name default

models: (1) Reduced form Normal Copula; (2) Structural Variance Gamma.

For long tranche risk positions the hedge ratios that minimize the wealth change variance were

smaller than those that minimize the expected shortfall at high degrees of confidence. As the

expected shortfall only penalizes losses exceeding a certain threshold, for senior tranches with

sufficient subordination, the minimization of expected shortfall at low to modest confidence levels

can imply much smaller hedge notionals than that required by the minimization of wealth change

variance measure which penalizes gains as well as losses. The break-even average cost of hedging is

an increasing function of the hedge ratio and the no-default carry is a decreasing function of the

optimal hedge ratio. Variance-optimal hedged sell equity tranche protection positions were found to

have positive no-default carry, whereas expected shortfall optimal hedging of sell equity protection

positions at high confidence levels was associated with lesser or negative no-default carry. Variance-

18

optimal hedged sell mezzanine and senior tranche protection positions have negative no-default carry

(and positive no-default carry for mezzanine-senior protection buyer).

In the face of unhedgeable risks, one has to contend with the reality that the hedge ratios that

minimize different hedge error functions are different. Given a traded price, a market agent can

hedge such that his expected change in wealth is maximized relative to his choice of hedge error

function. For both the default models, we provided examples of assessing hedge notionals so as to

minimize hedge error measures or maximize (positive) expected wealth change per unit hedge error

measure. The tradeoffs between carry and default risks are illustrated by the wealth change

probability distributions evaluated at different hedge ratios. Positive no-default carry was shown to

be associated with tail losses – as visible by the left tail of wealth change distributions. Negative no-

default carry was associated with tail gains – as visible by the right tail of the wealth change

distributions.

While the complexity of the portfolio default loss process has been widely used to accommodate the

correlation skew in the risk-neutral framework (by invoking multiple Copulas, base correlation, etc),

we have shown that a default risk-averse tranche protection seller-hedger, say, for example, seeking

to minimize her ESF0.80, and the imperfections in replicating the CDO tranche contract (even for PNC

model of defaults) also support the existence of the correlation skew. Sircar and Zariphopoulou

[2006] arrive at a similar conclusion based on utility valuation and a long-only tranche risk investor.

The framework demonstrated here is not tied to any special stochastic description of the market. The

approach encourages developing realistic descriptions of markets because their features are not

fundamental inconveniences in the optimization framework that does not fall apart simply because

risks are not perfectly hedgeable. This framework binds together some of the trading risk

management and derivative price modeling objectives by focusing on the costs of replication and

recognizing the reality of imperfect replication. In contrast the approach of formally invoking risk-

neutrality without any explicit analysis of hedging strategy is focused on providing versatile loss

distributions to fit prices, assuming a complete market and perfect replication.

An important challenge of the optimal hedging approach is computationally solving the constrained

optimization problems that represent hedging strategies. In a static one-period framework, CDO

hedge optimization involved solving a multi-dimensional minimization problem under constraints. In

a dynamic multi-period setting, optimal CDO hedging involves solving multi-variate variational

problems which are computationally much more challenging than the static problem. The

specification of an empirically sound coupled model of spreads and defaults is also challenging, and a

prerequisite for any dynamic hedge optimization analysis. The challenges of implementing an

optimal hedging analysis are amply balanced by the potential of better integrating pricing, hedging, &

risk management and developing views on relative value en-route to pricing.

19

Appendix-A

In this Appendix, the variance of the change in wealth of a portfolio of a defaultable bond and a

credit default swap is minimized in a static hedging framework, accounting for recovery uncertainty

and differences between market and face values of the bond. Analytical expressions are derived for

(1) the portfolio composition that minimizes the variance of wealth change; (2) the mean wealth

change based break-even premium of the default swap; (3) the un-hedgeable wealth change variance.

In hedging against the default of a bond by purchasing default swap protection, when the bond has a

market value that is different from its face value, it is not possible to eliminate credit risk – the

sensitivities of the residual wealth change variance are presented here. It is shown that the residual

wealth change variance increases as the square of the difference between the market and par value of

the bond, and uncertainty in recovery also increases the residual risk.

To focus on the random default time and the associated random recovery, we simplify other

parameters of the problem. We do not consider interest rate uncertainty and for simplicity of

presentation we do not consider the term structure of interest rates and default probability – however

the hedging framework for all the examples described here lends itself readily to computations with

interest rate and default probability term structures.

Our analysis requires prescribing objective default probabilities. We adopt the popular and

convenient exponential parameterization for the default time pdf

In this parameterization the mean time to default and the standard deviation of the time to default are

the inverse of the default hazard rate λ: t d = σ td = 1 / λ . The approach outlined here is applicable

regardless of the parameterization for default time uncertainty. Recovery is assumed to be uncertain

and results are presented in terms of the mean and variance of recovery. Correlated recovery and

default time can be handled in the static hedging framework computationally. However, the simple

analytical results presented here assume statistical independence between default time and recovery.

Default probabilities (and the associated hazard rates) are provided by a variety of rating agencies in

different forms (e.g., Moody’s, Standard and Poor’s). Approaches that analyze details of firm’s

assets and liabilities can also be brought to bear on the problem of estimating its probability of default

(e.g., KMV, Kamakura, etc). We do not discuss the relative merits of these different sources here.

We show later that in certain circumstances the optimal-hedge portfolio is independent of the

objective default probabilities – this corresponds to the complete market risk-neutral approach.

However, on relaxing the two main assumptions implicit in the risk-neutral approach, namely fixed

recovery and bond market value equal to par, we show how the objective default probabilities can

explicitly determine the optimal portfolio and more importantly the residual risks which require the

trader to take a view on the objective measures of the variables that determine his P&L uncertainty.

20

Change in Wealth of a Portfolio of a Defaultable Bond and CDS

A risky bond with a par value of nu pays a continuous coupon of c dollars per notional dollar per unit

time over the time interval [0,T]. The market value of the defaultable bond is pu. The protected

notional of the CDS is nr and the premium paid to purchase default protection is sr. The CDS

counterparty pays (1-R)nr in the event of default (Figure A1).

sr

-pu, c , nu

CDS Trading Bond

Book -pu, c, Rxnu

(1-R)nr

If default does not occur over the life of the bond, then the change in wealth of the portfolio over the

time interval [0,T] is

T

ΔW = ΔWs = − p u + nu e − rT + (cnu − s r n r )∫ e − rτ dτ (A2)

0

If default occurs prior to maturity, the change in wealth over the time interval [0,T] follows

td

ΔW = ΔWd = − p u + [ Rnu + (1 − R )n r ]e − rt d

+ (cnu − s r n r )∫ e − rτ dτ (A3)

0

In (A2) and (A3) the interest rate is given by r and all cash-flows accrete at that rate. The random

parameters involved in the change in wealth are the time to default and recovery. Assuming these

parameters to be independent and the time to default following (A1),

e − (λ + r )T

ΔW = .

(λ + r ) (A4)

[n {c(e

u

( λ + r )T

− 1) + (λ + r ) + λR (e ( λ + r )T

}

− 1) − p u e ( λ + r )T

(λ + r ) + n r e ( ( λ + r )T

) ]

− 1 (λ (1 − R ) − s r )

For the mean change in wealth to be zero the default swap spread sr must follow

ΔW = 0 ⇒

sr =

[ ]

nu c(e ( λ + r )T − 1) + (λ + r ) + λR (e ( λ + r )T − 1) − p u e ( λ + r )T (λ + r )

+ λ (1 − R )

(A5)

(

n r e ( λ + r )T − 1 )

21

Variance Optimal Hedge

( )

The variance in the change in wealth, σ Δ2W = E[ΔW 2 ] − ΔW , requires finding expectations of the

2

squares of cash-flows given in (A2). The variance of the wealth change, subject to the constraint

(A5), is evaluated analytically. Setting its derivative with respect to the default swap notional to zero

yields the variance minimizing portfolio:

= 0 ⇒ nr = nr* = u

dnr ϖ4

(A6)

( λ + r )T ( λ + r )T ( λ + 2 r )T

ϖ 1 = λ (1 − R )e (1 − e ) , ϖ 2 = r (1 − R )e

rT

(2e − e rT

− 1)

ϖ 3 = re ( λ + r )T

[{(1 − R ) + (1 − R ) + σ R }+ e {(1 − R ) − 2(1 − R ) + σ R }− e

2 2 rT 2 2 (λ + 2 r )T

{(1 − R ) 2 − (1 − R ) + σ R2 }] −

+ r[(1 − R ) 2 + σ R2 ]

ϖ 4 = r ((1 − R ) 2 + σ R2 )(e ( λ + r )T (1 + e rT − e ( λ + 2 r )T ) − 1)

Substituting nr = nr* into (A5) yields the zero-mean wealth change premium sr* corresponding to the

minimum variance portfolio of a defaultable bond and a credit default swap. The analytical

expression for the wealth change variance and its residual value is complicated and not reproduced

here. We present analytical results for some limiting cases and sample calculations for the general

case.

For this case the randomness in recovery has no bearing on hedging and pricing, and the change in

wealth of the hedged portfolio over time interval [0,T] is identically zero. For the idealized problem

formulated here, the portfolio that results in minimum wealth change variance and on the average no

change in wealth is

nr* = nu

(A7)

sr* = c-r

σ Δ2W * = 0 (A8)

The results (A7) and (A8) apply regardless of recovery uncertainty and are identical to that obtained

in the risk-neutral approach (e.g., Jarrow and Turnbull [1995]; Duffie and Singleton [1999] ). These

results are also independent of the objective default probability – much like the binomial branch

option pricing problem that is the basis of the popular tree-based option pricing models (Cox and

Rubinstein, 1979). Next we look at situations where it is not possible to eliminate risks even in the

idealized theoretical framework explored here.

22

Perfectly Known Recovery Case ( σ R2 = 0 )

For this case the general results for nr* and sr* can be used after setting σ R2 = 0 , in (A6). The general

expression for the residual variance simplifies considerably:

( nu − p u ) 2 δ 1

σ 2

ΔW *= (A9)

δ2

δ 1 = r 2 (1 + e 2( λ + r )T ) − (λ + r ) 2 e λT (1 + e 2 rT ) + 2λ (λ + 2r )e ( λ + r )T , δ 2 = r 2 (e ( λ + 2 r )T − 1)(e ( λ + r )T − 1) 2

The residual wealth change variance is proportional to the square of the difference between the face

value and the market value of the defaultable bond, whose possible credit event we seek to hedge by

purchasing a regular default swap.

We present sample calculations for a specific case in Figure A2 and in Tables A1 and A2 (case 1).

Points to note are:

(1) The default swap notional nr* and the break-even spread sr* are influenced by differences in

par and face value of the defaultable bond and the default hazard rate.

(2) The dependence of the break-even CDS spread and optimal hedge notional on the hazard rate

is relatively mild, which indicates the ability of the optimal hedge to minimize dependence

on objective measures, albeit not eliminating it and hedging errors.

(3) The residual wealth change variance is an increasing function of the differences between the

price and notional of the underlying bond and an increasing function of the default hazard

rate.

The derivations for nr* and sr* for the general case are in (A5) and (A6). Here we present numerical

results similar to the previous case but with the additional effect of recovery variance in Figures A3-

A4 and Tables A1-A4. The optimal notional and break-even spread are functions of recovery

variance (equation (A6)) – although the dependence is mild as the recovery variance appears as an

additive term in both the denominator and numerator in (A6). The residual wealth change variance

increases with recovery uncertainty.

Figure A5 shows results of Monte-Carlo simulation for the example setup in Table A1. The

agreement between the Monte-Carlo simulation technique described in the main text and the

analytical results of the Appendix provides a check on both approaches. The MC technique also

provides a way to show the value at risk and expected shortfall error measures that capture tail risks.

The hedge ratios for the optimization of wealth change variance can be quite different from the

optimization of value at risk and expected shortfall.

23

1.2

0.1

n / nu

*

r

1

0.08

0.8

0.06

0.8

8

0.9 0.04

λ (yr -1 )

1 0.02

1.1

p u / nu 1.2

0.1

0.1

s (1/yr)

*

r 0.05 0.08

0 0.06

0.8

8

0.9 0.04 λ (yr -1 )

1 0.02

1.1

pu / nu 1.2

0.03

σ ΔW *

0.02 0.1

0.01 0.08

nu 0 0.06

0.8

8

0.9 0.04

λ (yr -1 )

1 0.02

1.1

p u / nu 1.2

T = 5 yr, c = 10%/yr, r = 5%/yr, R = 0.5 , σ R / R = 0

24

attribute symbol value

case 1: 0%

σR

Recovery std dev case 2: 50%

(% of R )

case 3: 75%

1.17% 1.63% 1.91%

(% bond notional)

Table A2. Variance optimal hedge sample results for problem setup in Table A1

25

1.1

0.1

n / nu

*

r

1

0.9

0.08

0.06

0.8

8

0.9 0.04

λ (yr -1 )

1 0.02

1.1

p u / nu 1.2

0.1

0.075 0.1

s (1/yr)

*

r 0.05 0.08

0.025

0 0.06

0.8

8

0.9 0.04 λ (yr -1 )

1 0.02

1.1

pu / nu 1.2

0.04

σ ΔW *

0.1

0.02 0.08

nu 0 0.06

0.8

8

0.9 0.04

λ (yr -1 )

1 0.02

1.1

p u / nu 1.2

T = 5 yr, c = 10%/yr, r = 5%/yr, R = 0.5 , σ R / R = 0.75

26

0.94

0.92 0.1

n / nu

*

r

0.9 0.08

0.88

0

0.06

λ (yr -1 )

0.25 0.04

0.5 0.02

σR /R 0.75

1

0.1

0.08

0.08

0.078

0

0.06

λ (yr -1 )

0.25 0.04

0.5

σR /R

0.02

0.75

1

σ ΔW *

0.03

0.02 0.1

nu 0.01 0.08

0

0

0.06

λ (yr -1 )

0.25 0.04

0.5

σR /R

0.02

0.75

1

T = 5 yr, pu/nu = 0.9, c = 10%/yr, r = 5%/yr, R = 0.5

27

attribute symbol value

case1 0%

σR

Recovery std dev case 2 50%

(% of R )

case 3 75%

0.079% 0.106% 0.124%

(% bond notional)

Table A4. Variance optimal hedge sample results for problem setup in Table A3

28

2

hedge error measure (% bond notional)

1.8

1.6

1.4

1.2

1

0.8

0.6

0.4

80 85 90 95 100

CDS hedge notional (% bond notional)

890

870

CDS breakeven spread (bps/yr)

850

830

810

790

770

750

80 85 90 95 100

Figure A5. Hedge ratios and hedge error measures: known recovery case

T = 5 yr, pu/nu = 0.9, c = 10%/yr, r = 5%/yr, R = 0.5 , σ R / R = 0

29

Bond-CDS Hedge Error Minimization Summary

Two assumptions are made in the wide-spread applications of risk-neutral pricing of default swaps:

(1) the risky bond whose credit risk is being hedged has a price equal to its par value; (2) the recovery

amount is deterministic. Both of these assumptions can be far from realistic. Changing credit ratings

and interest rates can result in the risky bonds trading at a significant discount or premium from par.

Recoveries in cohorts of similar subordination vary significantly, making it impossible to have

perfect foresight about recovery while entering into a swap agreement. This work relaxes these two

assumptions by explicitly analyzing hedge performance with the impact of credit risk (default time

and extent) being minimized by a regular default swap. The key analytical results are the amount of

notional in a CDS contract and the premium associated with the protection that result in both

minimization of the variance in change of wealth over the life of the swaps and zero mean change in

wealth. The expressions show that recovery variance and differences of market values from par

value have important consequences for the composition, break-even price, and residual risks of the

hedged portfolio.

The residual wealth change variance in a portfolio of a defaultable bond and a regular default swap is

proportional to the squared difference between its par value and its market value. When the market

value and par value are identical the residual variance is identically zero, irrespective of recovery

uncertainty. When the market value and par value are distinct, the residual variance is an increasing

function of recovery uncertainty.

The smaller the market value is compared to the par value, the smaller the notional protected in the

default swap and the larger the break-even default swap spread (for a fixed coupon rate) will be in the

minimum variance hedged portfolio. The notional of the default swap and the break-even spread

depend on the objective default probability (i.e., objective hazard rates) when the residual risk is not

zero although that dependence is weaker than the dependence of the residual risk on the objective

hazard rates. When the residual risk is zero (i.e., the par and market value are the same), the

optimally hedged portfolio is independent of the objective default probability and is consistent with

the risk neutral approach to marking to market of default swaps.

30

Appendix-B

In the Variance-Gamma structural approach defaults are based on a model of evolution of a firm’s

value return, which follows a geometric Brownian motion (with drift and volatility parameters μi and

σ i ) evaluated at stochastic time clock governed by increments of gamma processes (Madan et al

[1998]):

Δf i

= μi g i (Δt ;1,ν ) + σ iWi ( g i (Δt ;1,ν )) (B1)

fi

We adopt single factor approaches to correlate the gamma stochastic clock processes and the

Brownian motion underlying the firm-value evolution. For the Brownian motion we have

Wi = β Wm + 1 − β 2 Z i (B2)

where Wm and Zi are independent standard Wiener processes. The increments of the gamma

processes for the different issuers are assumed to follow

The processes gmarket and ui are increments of independent gamma processes. The marginal density of

the increments of the gamma process g (Δt ; m, n ) follows

m 2 Δt

−1 ⎡ m ⎤

m 2 Δt g n

exp ⎢− g ⎥

⎛m⎞

f (g Δt ) = ⎜ ⎟ ⎣ n ⎦

n

(B4)

⎝n⎠ ⎛ m Δt ⎞

2

Γ ⎜⎜ ⎟⎟

⎝ n ⎠

where Γ (.) denotes the gamma function and the characteristic function and the first 2 moments of

the increment process are given by

m2

− Δt

⎛ u ⎞

E[exp{iug }] = ⎜1 − i ; E [g ] = mΔt ; E[( g − g ) ] = nΔt

n 2

⎟ (B5)

⎝ m/n⎠

We employ a default-barrier model where the first passage of the firm value beneath a static and

uniform barrier (ϖ : fraction of initial firm-value) triggers default. We do not explore here all the

features of a structural model, such as random recoveries or the possibility of expressing a dynamic

coupling between credit spreads, default and recovery. Such a coupled spread-default probability-

recovery stochastic description will be necessary if the costs and efficacy of dynamic hedging are to

be assessed en-route to assessing fair-value.

31

We simulate the firm value over a monthly time grid ( Δt = 1 / 12 (yr) ). To facilitate a meaningful

comparison of the features of the VG structural model and the Poisson-Normal Copula reduced form

model, we calibrate the VG model to reproduce two statistics that are central to describing the

portfolio losses: (1) T period default probability for issuers; (2) T period portfolio loss standard

deviation. For the Normal Copula example of the main text the prescribed hazard rate of 0.65%/yr

results in a 5 year default probability of 0.0319. A set of VG parameters that effects such a marginal

description of 5 year default probability is:

ν = 2 yr;ϖ = 0.39152623

For a 125-issuer initially homogeneous asset pool with recovery set to 30 %, the 5 year pool loss

standard deviation using 25% asset correlation in a Normal Copula model is 3.32% of the initial pool

notional. That portfolio-loss standard deviation can be matched by adjusting β and κ . The sample

results in the main section are presented using a common stochastic time increment process.

κ = 1; β = 0.254105

1.4

1.2

1

firm value

0.8

0.6

0.4

0.2

0

0 1 2 3 4 5

time (yr)

Figure B1. VG firm value sample paths for 2 issuers. Three realizations depicted in

32

References

Bouchaud, J-P., M. Potters, Theory of Financial Risks, From Statistical Physics to Risk Management,

Cambridge University Press, 2000.

Cariboni, J, W, Schoutens, Pricing Credit Default Swaps Under Levy Models, preprint, 2004.

Cont, R., P. Tankov, Financial Modelling with Jump Processes, Chapman &Hall/CRC, 2004.

Cox, J. C., S. A. Ross, M. Rubinstein, Option Pricing: A Simplified Approach, Journal of Financial

Economics, 7, 229, 1979.

Duffie, D., K. J. Singleton, Modeling Term Structures of Defaultable Bonds, The Review of

Financial Studies, vol. 12, no. 4, pp 687-721, 1999.

Jarrow, R., S. Turnbull, Pricing Options on Financial Securities Subject to Default Risk, Journal of

Finance, 50, 53-86, 1995.

Li, D. X., On Default Correlation: A Copula Function Approach, Working Paper Number 99007,

The RiskMetric group, 1999.

Luciano, E., W. Schoutens, A multivariate Jump-Driven Asset Model, preprint, December 2005.

Madan, D. B., P. P. Carr, E. C. Chang, The Variance Gamma Process and Option Pricing, European

Finance Reviews, 2, 7-105, 1998

Moosbrucker, T., Pricing CDOs with Correlated Variance Gamma Distributions, preprint, January

2006.

Petrelli, A., J. Zhang, J., N. Jobst, V. Kapoor, A Practical Guide to CDO Trading Risk Management,

The Handbook of Structured Finance, edited by Arnaud de Servigny and Norbert Jobst, McGraw-

Hill, 2006.

Pochart, B., J.-P. Bouchaud, Option Pricing and Hedging With Minimum Local Expected Shortfall,

2003

Sircar, R., T. Zariphopoulou, Utility Valuation of Credit Derivatives and Applications to CDOs,

preprint, July 2006.

33

Contact information:

Andrea.Petrelli@credit-suisse.com

Olivia.Siu@morganstanley.com

Jun.Zhang@credit-suisse.com

Vivek.Kapoor@ubs.com

34

May 2008

Andrea Petrelli1, Jun Zhang1, Olivia Siu2, Rupak Chatterjee3, & Vivek Kapoor3,4

Abstract. Analyzed here is a Cliquet put option (ratchet put option) defined as a resettable strike put

with a payout triggered by the reference asset falling below a specified fraction of its value at a prior

look-back date. The hedging strategy that minimizes P&L volatility over discrete hedging intervals is

assessed. Examples are provided for an asset exhibiting jumpy returns (kurtosis > 3) and temporal

correlation between the squared residual returns. The limited liquidity of the asset limits the discrete

hedging frequency. Each of the realities of discrete hedging intervals and fat-tailed asset return

distributions render the attempted replication imperfect. A residual risk dependent premium is added to

the average cost of attempted replication (i.e., average hedging cost) based on a target expected return on

risk capital. By comparing the P&L distribution of a derivative seller-hedger with that of a delta-one

trader holding a long position in the underlying asset, relative-value based bounds on pricing of vanilla

options and Cliquets are presented.

Keywords: Gap-risk, Cliquet, Crash-Cliquet, Kurtosis, Hedging, Residual risk, Option traders P&L

______________________________________________________________________________

1. Introduction

The P&L of a seller and hedger of a Cliquet contract on an asset with limited liquidity and with

jumpy returns is analyzed here. The hedge ratio that minimizes P&L volatility, the average

hedging cost, and the hedge slippage probability distribution are assessed by applying the

Optimal Hedge Monte-Carlo (OHMC) methodology developed by Bouchaud & Potters [2003].

The computed probability distribution of the option-seller-hedger’s P&L reflects the stochastic

characteristics of the asset, the hedging strategy, and the Cliquet contract. We determine the

risk-premium that needs to be added to the average hedging cost to render the risk-return of the

derivatives trader (that sells and hedges the option) to be no worse than a delta-one trader who

is long the underlying asset.

Cliquets in the equity markets are often in the form of out-of-the money put Cliquets that

are used to protect the holder from a market crash scenario (i.e., crash Cliquet, gap risk Cliquet).

Most gap-risk Cliquets are defined as forward starting put spreads (e.g., 85-75 strike Cliquet put

spread). This structure is similar to a tranche of a market value CDO (with attachment points of

15% and 25%). Hedging the mezzanine tranche of such a CDO involves trading the underlying

assets to protect against gap risk. These structures share a common feature: sudden large moves

of the underlying asset can cause economic loss. The OHMC methodology proposed below will

explicitly include such moves through a process that exhibits excess kurtosis and results in

credit-type loss mechanisms.

1 2 3 4

Credit-Suisse; Natixis; Citi; corresponding author; email: vivek.kapoor@mac.com

1,2,3

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 2

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Petrelli et al [2006] analyzed optimal static hedging of multi-name credit derivatives (i.e.,

synthetic CDO tranches). Kapoor et al [2003] employed a GARCH(1,1) model in the OHMC

framework to examine the risk return characteristics of two-tranche structures supported by a

volatile asset. In that work the underlying asset values were reported monthly and the hedging

interval was also monthly. The errors incurred in replicating the senior tranche were compared

to that incurred in replicating the junior tranche, in addition to computing the average cost of

attempted replication. The impact of knockout and running premium for a Cliquet contract

results in a wealth change formulation that is quite similar to that for a CDS swaption problem

analyzed in Zhang et al [2006]. This work builds up on the formulations in Kapoor et al [2003]

and Zhang et al [2006] and applies it to a Cliquet contract, and further analyzes residual risks and

return on risk capital. The ultimate goals of our series of works on optimal hedging are to be

able to handle hedging, attempted replication, and assess residual risks of multi-asset options - be

they credit type assets or equity type assets. This is also a prerequisite to developing a

satisfactory valuation model for multi-name credit derivatives. The reader is referred to Petrelli

et al [2007] for documentation of dynamic-hedge performance of CDO tranches, and to Laurent

et al [2008] for a direct study of replication of CDO tranches.

In the OHMC approach the asset underlying the derivative is simulated based on a model

that seeks to capture its real-world characteristics. The hedge ratio and pricing functions are

sought at every time step to keep the hedged derivative position as flat as possible between

successive hedging intervals. The numerical solution for hedging starts at the time-step prior to

the option expiry. The hedge ratio and pricing functions are a solution to the variational-calculus

problem of minimizing a statistical hedging error measure between two time-steps while keeping

the trading book flat on the average. Thus the OHMC methodology puts itself in the shoes of a

derivatives trader attempting to replicate the option payoff. Like the derivatives trader, the

OHMC methodology is concerned with residual risk accompanying any hedging strategy in the

real-world. OHMC seeks to deliver to the derivatives trader information on the average hedging

costs and the residual risks. By comparing the average cost of hedging and the residual risks

with the amount of money someone is willing to pay the trader for that option, an opinion on the

attractiveness of the trade can be developed.

The optimal hedging methodology adopted here follows the approach taken by Bouchaud and

co-workers: see Bouchaud and Potters [2003] for an introduction to OHMC. Other foundational

studies of optimal hedging include Schweizer [1995], Laurent & Pham [1999], and Potters et al

[2001]. All of these works are focused on the cost of option replication by analyzing dynamic

hedging explicitly and as a pre-requisite to valuation.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 3

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

The works on optimal hedging mentioned above and the approach pursued here are easily

distinguished from the formal risk neutral valuation approaches insofar as follows:

• OHMC addresses the mechanics of hedging, the average hedging costs and the hedge

slippage distribution, and establishes the theoretical reasons for lack of perfect

replication. Taking expectations under a de-trended underlying process does not directly

address hedging and replication errors (in risk-neutral modeling hedging errors are

assumed to be zero under ideal conditions).

valuation - with demonstrable theoretical perfect replication - is limited to very specific

descriptions of underlyings that are not empirically observed in even the most vanilla and

liquid financial instruments, let alone exotic underlyings. For example, the daily return

kurtosis of US large cap stocks is on the average ~ 20: see Bouchaud & Potters [2003].

The kurtosis of the daily and monthly returns of the S&P500 total return index is many

multiples of 3 (the kurtosis of Geometric Brownian Motion).

The residual risk associated with any attempted replication strategy is of paramount

importance to a derivatives trader trying to carve the risk-return profile of a trading book, and

making the binary decision of selling an option at a given price in the first place. OHMC is

focused on hedging strategies, their expected costs, and residual risks.

The central idea of derivative replication is establishing costs and trading strategies for

eliminating risk. However, valuation modeling has come to limit itself to taking expectations of

option payoffs under a de-trended underlying process – without establishing the mechanics to

achieve replication or estimates of the residual risks. Such formal risk neutral models (that do

not establish replication but presume it to be theoretically possible) are generally fit to market

prices – without offering any analysis of hedging and its limitations. The risk-neutral label

seems to be earned merely by taking LIBOR discounted averages of option payoff evaluated

using a de-trended description of the underlying relative to the cost of carry (which in recent

environments has itself jumped around!). Such formal valuation models do not differentiate

options based on the relative sizes of replication errors endemic to the option contract and the

underlying process. While such formalism based valuation modeling is taking hold in

accounting practices, it is largely an exercise of fitting model parameters to observed derivative

prices. Such formal risk neutral models do not directly help understand risk-return

characteristics or hedge performance of vanilla options or exotics. The main role of such risk

neutral valuation models seems to be facilitating upfront P&L for exotics by employing

parameters fitted to vanilla options, with the presumption that a delta to the underlying and

vanilla options can perfectly replicate an exotic.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 4

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

100%

90%

80% kurtosis = 15

70% kurtosis = 3

60%

50%

40%

30%

20%

10%

0%

0 20 40 60 80 100

hedging frequency (#/month)

(a)

Figure 1a. Dependence of hedging error on hedging frequency and return kurtosis for a 1 month at the money put

option (see Example 1 in Appendix-II for further details). The hedge error measure displayed here is the standard

deviation of hedging error divided by the average hedging cost. The black line shows results for a stylized asset

(stylized-asset 3 defined in section 4) with a return kurtosis of 15. The rose-pink-line shows results for the same

asset, but without any excess kurtosis, i.e., with a kurtosis of 3 which corresponds to a Geometric Brownian Motion

(GBM) rendition of the asset.

90 5

4

100 80

3

70

hedge-ratio %

2

daily P&L ($)

asset value ($)

95 60 1

50 0

90 40 -1

30 -2

-3

85 asset-value ($) 20 -4

hedge-ratio (%) 10 -5

80 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

hedge interval (business days)

business days

(b) (c)

Figure 1b&c. Sample path behavior of OHMC analysis with daily hedging for a 1 month at the money put option

for a stylized asset with return kurtosis equal to 15 (see Example 1 in Appendix-II for further details). The

evolution of the asset-behavior and the hedge ratio is shown in (b). The daily P&L is shown in (c) with

contributions from the option position and the hedge position (discounted to start of hedging interval). The P&L

plot does not show the risk premium the option seller will add to the average hedging cost to get compensated for

the residual P&L risk. The sample path shown here corresponds to the 1 year 99.9% confidence level equivalent

total wealth change over the options life.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 5

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Minor jumpiness of returns (excess kurtosis) rules out perfect hedging even in the continuous

hedging limit for vanilla options – as depicted in Figure 1 - based on an example of an at-the-

money put further detailed in Appendix-II. This fact is brazenly ignored by the mainstream

valuation modeling but experienced by the derivative trader herself. The unsophisticated model

user can fall into the convenient trap of believing that the only consideration in a derivatives

trade is how correct is the volatility surface (or parameters of stochastic volatility models) in

valuing a derivative and calculating the standard greeks (delta, vega, gamma). Accounting

departments (also called Product Control) further reinforce this risk-free replication belief based

modeling regime – by embracing the unique derivative price found after calibrating parameters

to observed vanilla prices. This ignores the bid-offer of prices in vanilla derivatives and what

they may reflect about the residual risks in attempting to replicate a vanilla option. At greater

peril, for exotics, this ignores the fact that the residual risks in replicating an exotic may be quite

different than the vanilla option and a-priori one should have no expectation that calibration to

some presumed mid price of vanillas results in a replicating strategy for the exotic.

Risk management departments (also called Risk Control) are charged with feeding the

pricing model sensitivity outputs into a VaR model, and generally defer P&L deliberations to

accounting. Accounting departments in turn often defer to a valuation model that purports itself

to be risk-neutral and does not a-priori communicate estimates of hedging errors. VaR models

vary in terms of granularity of market risk factors (index, vs. single name, etc) and their

capabilities in making assessments of P&L with sensitivities and-or a complete revaluation. Due

to its simplifications, often VaR is not reported at a trade level. With the valuation model based

on presuming perfect replication, and the VaR model typically broad-brushed and not focused on

the quirks of exotics, it is quite possible that trades are executed – possibly strongly motivated by

the upfront P&L or significant carry – yet without any careful assessment of risks. Therefore a

proliferation of risk neutral valuation models has not generally been accompanied by an

improvement in risk management. In-fact, valuation models that purport to be risk-neutral and

do not advertise irreducible hedging errors, perpetuate the incorrect belief that a delta-hedged

position is close to risk free and actually aid and abet the taking of un-sized risks, despite the

seeming oversight of valuation modeling, risk-control, and product control.

The diligent trader/risk manager will typically get to understand the exotic derivative

over time and may know how its risk sensitivities are different enough from the vanilla

derivative such that calibration of model with vanillas does not ensure a plausible pricing of risk-

premiums that are endemic to attempting to replicate the exotic. The “risk-neutral” valuation

models complete silence on hedge performance (as a part of valuation) renders them of little

value in developing an exotics trading strategy with clearly documented risk-return tradeoffs.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 6

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

The amount of compensation that a market agent that sells an option (and attempts to

replicate) demands for unhedgable risks is for that agent to opine on and for trading

counterparties to be the ultimate arbiter of. The amount of risk premiums the market will bear

will depend on demand and supply, market sentiments, and the extent to which the residual risks

are diversifiable in a practical trading book. Replication and diversification are at the heart of

making derivative trading decisions, and derivative valuation models that are silent about even

theoretically unhedgable risks are of little value in guiding trading and risk management in

expressing a risk preference. The OHMC approach has the potential to more directly tie together

valuation modeling, hedge performance analysis, and risk management, by relating the average

hedging costs and hedge slippage distribution to the distribution of the underlying as a precursor

to valuation, and therefore prior to the stage where upfront P&L or carry motivations are

entrenched. The purpose of this article is to demonstrate the practical feasibility and the utility

of the optimal hedging approach for Cliquet contracts. It will be demonstrated that the OHMC

framework can be utilized as a consistent hedging, pricing, risk-management engine that can

serve the needs of traders, risk managers, and product control (see Table 1).

Cliquet contracts are traded for liquid public market assets as well as more customized

assets that could represent a trading strategy itself, in the form of either a rule based strategy, or a

hedge fund, or an associated index. The values of such customized assets are often reported at a

much lower frequency than the liquid public market assets, and the liquidity time interval over

which any hedger can adjust the hedging portfolio can be even larger than the interval over

which asset-values are reported. The hedging of Cliquet contracts for such imperfectly liquid

assets is one of the focus of this work, hence the explicit treatment of hedging frequency. The

bespoke baskets motivating this work are more widely held as long positions, therefore hedging

requiring going long the underlying bespoke basket may be easier than going short. However it

is possible to go short by total return swaps with counterparties that want to go long the bespoke

basket. In this work we invoke an ability to go long as well as short the asset underlying the

Cliquet contract, with limited frequency of adjusting the hedge due to contractual limitations on

redemptions of customized baskets.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 7

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

and assesses real-world residual

i. Assumes perfect replication is

risks that are generally found to

theoretically possible and residual

be significant compared to

risks are non-existent

average hedging costs

Hedging & ii. Continuous/instantaneous hedging

ii. Can addresses hedging

replication with no transaction costs

frequency and transaction costs

iii. Naïve deltas assuming (i)and (ii)

iii. Produces hedge ratios that

without any reference to real-

minimize desired hedging error

world risks

measure (e.g., P&L volatility or

expected shortfall)

i. Independent of underlying

Underlying i. De-trended martingale process

process (i.e., applicable to non-

description only

markov, fat tails, jumps etc)

free replication assumption – no

i. Scenario evaluation/back-testing

direct risk metric output from

is performed as a part of

model

assessing cost of hedging and

Risk hedge slippage measures which

ii. Sensitivity of valuation models

drive valuation

management can be used to evaluate risk

needs measures external to valuation

ii. Risk measures (VaR and

model (VaR, expected shortfall,

expected shortfall) are collateral

etc)

model output en-route to

assessing hedging strategy

iii. Loss scenarios can be assessed by

perturbing valuation model inputs

Table 1. Comparison of Risk Neutral Approach with Optimal Hedge Monte-Carlo (OHMC)

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 8

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Organization

The reader completely new to optimal hedging analysis is recommended to start with Appendix-

II which introduces it in the context of vanilla options and presents sample calculations on

hedging error, average hedging costs and hedge ratios. The reader with background on optimal

hedging analysis should jump right into the Cliquet formulation, and Appendix-II is then best

read prior to section 5 of the main text.

The mechanics of the Cliquet contract are presented in section 2, including a formulation

of a Cliquet option seller’s P&L. Section 3 presents the OHMC analysis of the P&L of the

Cliquet option seller – hedger, and the approach to finding the optimal hedge ratios and the

average hedging cost as well as residual risks. Also presented in section 3 are relative value and

risk capital measures that utilize the assessments of average hedging costs and residual risks and

provides guidance on valuation. Appendix-I and III present details of the OHMC

implementation for Cliquet contracts. The OHMC method to analyze Cliquets pursued here does

not hinge on any special or convenient stochastic description of the asset. Rather, it assesses the

average hedging cost and deviations around those averages, given any description of the asset

and Cliquet contract parameters. For the purpose of presenting specific examples we use a

GARCH(1,1) description of the asset and employ three stylized asset descriptions. Section 4

presents the GARCH(1,1) description and a method of moments approach that can be used to

calibrate its parameters to data. Section 5 presents some specific examples. A discussion of this

work and concluding remarks are presented in section 6.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 9

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

2. Cliquet Contract

The basic Cliquet put option contract consists of a series of forward starting European puts. On

pre-specified dates separated by the time-interval between checking payout trigger conditions,

τroll, the value of the reference asset is compared to its value one look-back interval, τlook-back,

earlier. If the asset has fallen beneath a specified fraction, called strike K1, of its value one look-

back interval earlier, then a payout to the option purchaser is made by the option seller. That

payout is the fractional amount by which the asset has fallen below the strike K1 multiplied by a

reference notional value – with possibly a maximum payout fraction established by a lower strike

K2 < K1 (sometimes called a “bear-spread”) . In the knock-out variant of the Cliquet contract, the

contract terminates after one payout event. In return for the possible payout, the option

purchaser pays the option seller a running premium and/or possibly an upfront payment. The

costs and efficacy of hedging such Cliquet contracts are analyzed here in the framework of

minimizing P&L volatility in between hedging intervals. There can be many other variants of

the Cliquet contract - the framework developed here can handle any path-dependent derivative.

T

τ roll

τ hedge τ look−back

τ obs

Figure 2. Schematic of time-scales pertinent to the Cliquet contract. The option tenor is denoted

by T. The most granular time-interval over which asset value observations are available is

denoted by τ obs . The hedging interval is denoted by τ hedge . The look-back interval over which a

decline in asset value triggers payoff and a termination of the contract is denoted by τ look−back .

The time-intervals over which the drop in asset value is checked is denoted by τ roll .

One of the motivations for this study are Cliquets on assets with limited ability to

rebalance hedges, due to contractual limitations on redemptions, and where the value of the

underlying assets can be reported less frequently than a typical publicly traded stock or bond or

CDS contract. The redemption time-interval and the asset value reporting interval are two time

scales that are characteristics of the underlying asset are a constraint on the Cliquet option trader.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 10

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Additionally, the frequency of checking the payout condition and the look-back interval (over-

which a decline of asset value triggers payout) are time-scales pertinent to a Cliquet trading

strategy. These time-scales are depicted in Figure 2. Devising a hedging strategy that is

cognizant of these five time-scales and the asset characteristics (including its volatility clustering

time-scale introduced later) is our purpose.

A-priori we know that the replication cannot be perfect, unless the description of the

underlying asset is contrived. We illustrated that for vanilla options in this paper (Figures 1 and

Appendix-II). One of our main goals is to demonstrate the utility of elucidating the residual

risks while attempting to replicate as much as possible. By combining the average hedging costs

and the residual risks, we seek to develop a framework for relative value metrics that will drive

the prices.

The terminology and symbols needed to specify a Cliquet contract and a protection seller’s P&L

are enumerated here.

Hedging interval:

(tk , tk +1 ]

s(tˆi )

Payout trigger time: tˆi* = min tˆi ∋ < K1

s(tˆi − nlb )

Reference notional: Ψ

s(tˆi* )

Payout amount: P(tˆi* ) =Ψ K1 − max , K 2

s(tˆi*− nlb )

Trigger indicator: I (tk , tk +1 ] =

0 no payout trigger over (tk , tk +1 ]

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 11

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

t*

tˆi* if I (t k , t k +1 ] = 1

χ (t k , t k +1 ) = ∫ df (t k ,τ )dτ ; t* =

tk t k +1 otherwise

Discounted payout:

P&L of sell Cliquet protection position between time tk and tk+1(discounted to tk):

∆Wt kcliquet (tk , tk +1 ) = C (tk ) + ηΨχ (tk , tk +1 ) − ω (tk , tk +1 ] − {1 − I (tk , tk +1 ]}C (tk +1 )df (tk , tk +1 )

(1)

The P&L of the sell Cliquet position written above is for contracts with possibly a combination

of a running premium and an upfront payment. This specification of Cliquet P&L is used later to

propagate the optimal hedging solution from the Cliquet expiry to the initial time step. A similar

formulation can be made for any other exotic Cliquets. We focus on the one-touch knockout put

variant of the Cliquet.

In addition to the variables that impact the P&L on the sell Cliquet protection position, the other

main object of interest for the Cliquet trader is Φ (tk ) , the amount of asset to hold at time step tk

to hedge the Cliquet position. The P&L generated by the hedge is determined by the change in

asset values, the carry costs for owning the economics of the hedge, and the discount rates:

s(t k )

∆Wtkhedge(t k , t k +1 ) = Φ (t k )s(t k +1 ) − df (t k , t k +1 )

DF(t , t )

k k +1

(2)

To account for different carry costs (i.e., different funding rates) and possibly dividends or

subscription fees associated with the asset, we employ a funding discount factor, DF(tk, tk+1),

which is possibly distinct from the risk-free discount factor df(tk, tk+1). Long dated derivative

contracts with relatively large hedge ratios can be quite sensitive to the funding rates of the

option seller-hedger. For such contracts, the prevailing funding rates of the different market

players can be as important as the differences in perceptions about the randomness of the asset

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 12

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

returns in determining the demand and supply in option markets. In reality the funding rates can

also be random, and in some situations can be a dominant determinant of the option cost. While

the formulation made here can handle random funding rates and discount rates, we only focus on

hedging the impact of the asset value, and invoke constant funding and discount rates for the

example calculations made here.

Cliquet Seller-Hedger’s P&L

We change slightly the notation of the Cliquet sellers-hedgers P&L ((1) & (2)) to develop the

notation of the OHMC algorithm. The contract value at time tk, C(tk), is denoted as a function of

the spot at time step k: i.e, as Ck(sk). The hedge amount at time tk, Ф(tk), is denoted as a

function of the spot at time step k, sk: i.e., Фk(sk). This notation emphasizes that the value and

hedge amount are viewed as time dependent functions of the spot asset value.

k k k

(3)

+ Φ k (s k )[s k +1 − s k / DF (t k , t k +1 )]df (t k , t k +1 )

To keep the notation compact and yet general we rewrite the above as

k

(4)

OHMC Problem

In the OHMC approach a MC simulation of asset evolution is used to evaluate Gk and Hk for

every random realization. Based on that, all terms of the wealth balance (4) can be directly

computed, other than the yet unknown deterministic functions of value and hedge ratio Ck(sk) and

Фk(sk). These two functions are found by imposing a constraint of zero average change in wealth

and minimum wealth change variance:

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 13

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

E[∆Wtk (t k , t k +1 )] = 0

(5)

Minimize

σ ∆2W tk (tk ,tk +1 ) [(

= E ∆Wtk (t k , t k +1 ) − ∆Wtk (t k , t k +1 ) 2 )] (6)

Appendix-I & III describes the algorithm to determine the unknown functions Ck(sk) and

Фk(sk) to satisfy (5) and (6) given a simulated MC ensemble of sk that also provide Gk, and Hk.

Each of the realities of fat return tails (kurtosis > 3) and discrete hedging individually rule out

perfect replication (see Figures 1 and Appendix-II). Certainly the combination of fat-tails and

discrete hedging render the residual risk to be of direct interest to someone charged with

managing the risk-return profile of a Cliquet trading book. As mentioned in the introductory

sections, we think that valuation modeling should not be divorced from assessment of residual

risks inherent to any attempted replication strategy. OHMC provides a readily implementable

avenue to fix the schism created by formal risk neutral models that do not address replication-

hedging explicitly.

As a part of the OHMC algorithm we looked at the P&L between time step k and k+1 discounted

to time step k; i.e.,

∆Wt (t k , t k +1 )

k

` `

K −2

∆W0 (0, T ) = ∑ ∆Wt (tk , tk +1 )df (0, tk )

k

k =0

The cumulative P&L from trade initiation to tk (present valued to time tk) follows

k −1

∆Wt (0, tk ) = ∑ ∆Wt (t j , t j +1 )[df (t j , t k )]−1

k j

j =0

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 14

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

The total change in wealth present valued to trade execution time is particularly important

because it provides a metric that is directly pertinent at inception. Imperfect replication

can/will/should make a Cliquet protection seller ask for a greater spread than that which simply

results in a zero average change in wealth. We will examine the distribution of the total change

in wealth from the point of view of solvency and associated risk capital, as discussed in the next

sub-section. To translate the total hedge slippage into an error term around the fair spread, we

normalize total P&L by the product of payout reference notional and average premium payment

time:

∆W0 (0, T )

Ψ χ (0, T ) (7)

The hedge slippage measure in (7) is pertinent to judge the risk-return of the Cliquet trade from

inception to finish – by summing all the hedge slippage and translating it into a running spread

by the normalization in (7). In addition to quantifying the risk-return over the transaction, a

trader also wants to control the P&L volatility over smaller time-intervals – say over the hedging

interval. We employ both time aggregated hedge slippage and local hedge slippage in assessing

risk capital, as described in the next section.

Much of formal risk neutral valuation modeling literature altogether ignores the question

of residual risks of replication attempts – invoking the formalism that as long as expectations are

being taken under a de-trended underlying, somehow replicating strategies exists in a risk-

neutral world. Such risk-neutral expectations are often taken under descriptions of the

underlying that are easily shown to thwart perfect replication for even simple derivative contracts

(jump-diffusion, GARCH(1,1), etc). Formal risk neutral expectations are also purported to

provide a valuation model for contracts for which a replicating strategy is hard to conceive (e.g.,

CDO tranches whose payouts occur only when jumps-to-default occur). Such valuation models

are mainly a parameter fitting exercise and are lacking the information needed by the person

charged with attempting to replicate the derivative payoff or responsible for trading-risk

management.

relegated as a risk management topic, whereas valuation modeling focuses on absolutes of

arbitrage-free pricing and is performed by individuals who are not responsible for managing

trading positions and their risks. Therefore hedging analysis has become distant form valuation

modeling, and one often hears of models for valuation of derivatives that are not models for

analyzing hedging! This divorcing of valuation modeling from hedge performance analysis and

risk management can account for the poor state of affairs in all these departments, and their

marginal role in helping making informed trading decisions. Consequently, poor risk

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 15

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

management and lack of understanding of risk return profiles of derivative trading books has

often been concomitant with the proliferation of derivative valuation models.

Risk Capital

The losses incurred by a derivative trading book can jeopardize the solvency of a financial

institution – or certainly the job of a trader or the existence of a trading desk or that of a hedge

fund. While ultimately a firm may be interested in its global risk profile, losses at any sub-unit

that are disproportionately larger than its size indicate that either extreme odds have been

realized and/or that the institution does not understand and can’t control the risks of its parts.

Reputational damage resulting from financial losses in a subset of a firm can have a detrimental

effect on the firm at a global level that go beyond the immediate financial risks. Also, if a clear

methodology of understanding risk-return is not expounded at a trade level or a trading desk

level, it is unlikely (and dangerous to assume) that risks are understood at a global portfolio

level. In this backdrop the unchallenged invocation of replication and/or complete

diversification of residual risks inside a valuation model is dangerous and misleading –

especially for new or exotic options where historical observations of option behavior are lacking.

For a sell Cliquet protection trade we address risk capital using the residual risk of the attempted

replication strategy found within OHMC. Appendix-II provides examples for sell vanilla option

positions.

The expected P&L from a derivative trade should be compared with tail losses to ensure

solvency, and profitability. While great trades may come from market insights that are not

modeled routinely, it should be possible to weed out poor derivative trades quantitatively. To do

so we define a specific solvency target and assess the risk capital associated with the derivative

trade. To compute risk capital over different time intervals (derivative tenor, hedging interval,

etc), we employ a target hazard rate to consistently assess the target survival probability over

different time horizons.

− ln ( p s (τ ))

τ interval hazard rate λτ =

τ

∆Wt (t , t + h)

h interval expected P&L

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 16

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

q(t; ps (h)) ∋ Probability{∆Wt (t , t + h) < q(t; ps (h))} = ps (h)

Q(t; p s (h )) = ∆Wt (t , t + h ) − q(t; p s (h ))

∆Wt (t , t + h)

h interval ps(h) quintile expected return on risk capital

Q(t ; p s (h) )

Our attempted replication via OHMC analysis imposes a zero mean change in wealth constraint.

In the face of residual risks, due to inherently imperfect hedging, (which is the driver of risk

capital) an option seller will need to add an additional charge over the average hedging cost to

create a positive expected change in wealth and try to obtain a certain pre-expense expected

return on risk capital. That expected return of risk capital can express an absolute solvency-

profitability criteria – i.e., this trade’s loss at confidence level x should not exceed y multiplied

by its expected P&L. Alternatively, the solvency-profitability criteria for a derivative trade can

be formulated relative to another trade – for instance simply being long the asset underlying the

derivative. For the Cliquet (and other vanilla derivative trades analyzed via OHMC in the

Appendix-II) we add a risk premium to the average cost of hedging so that they have an

expected return on risk-capital equal to that of a delta-one long only position in the underlying.

The wealth change of the delta-one trader between t and t+h is determined by the change

in asset values, the funding rates, and the discount rates:

s(t )

∆Wt (t , t + h) delta−one−long−trader = s(t + h) − df (t , t + h)

DF(t, t + h)

We assess the expected return on risk-capital of the delta-one trader and assess the bounding sell

price of the derivative contract as one which results in identical expected return on risk capital

for the derivative and the delta-one trader:

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 17

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

∆Wt ( t , t + h ) ∆Wt (t , t + h)

=

Q (t ; p s (h) ) derivative −trader Q (t ; p s (h) ) delta −one −long −trader

The resultant price of the derivative can be described as the delta-one bad deal bound pricing.

At any lower price, the delta-one long only trader has a higher expected return per unit risk

capital than the derivatives trader.

Different market operators can have different solvency targets. For the sake of

illustration in this paper we use the specific choice of 1 year 99.9% confidence level. Such

levels of confidence and even higher (say 1 year 99.97%) are pertinent to regulated financial

institutions that rely on a perception of solvency at a high degree of confidence to maintain

investor confidence and associated competitive funding costs. There can be derivative trades

that look attractive (relative to delta-one long only trade) at extreme confidence levels – but less

attractive at lower confidence levels, and vice-versa. From a relative value perspective one can

try to price the derivative such that over a range of confidence levels it is more profitable than a

delta one long position gamble on the underlying.

We examine the bounding calculation for the change in wealth over the tenor of the

derivative, and over individual hedging intervals. These are two different risk-preference

expressions. In employing the first one the market agent may have the capacity to take longer

terms risks – possibly due to locked in funding terms and a capital base. In the latter, the market

agent wants to keep score over shorter time horizons. In the case where we look at the wealth

change over the derivative tenor the addition of the risk premium is straightforward - it is derived

from the appropriate quintile of the total wealth change which is the sum of the wealth change

over the different hedging intervals:

∆W long (0, T )

Cbound 1 = C0 (s0 ) − qOHMC (0; p s (T )) × long0

Q (0; p s (T ) ) (8)

When we focus on the individual hedge intervals we piece together all the temporally local risk

premiums rendering the trade at least as profitable as a delta-one trader locally. We discount

these local premiums to trade inception, and report it as day one price difference, be it in the

form of upfront or a running premium:

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 18

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

K −2 ∆Wt long (t k , t k +1 )

C bound 2 = C 0 (s0 ) − ∑ qOHMC (t k ; p s (t k +1 − t k )) × long k × df (t 0 , t k )

k =0 Q (t k ; p s (t k +1 − t k ) )

(9)

We report sensitivities of average hedging costs and the sum of average hedging costs and the

above described risk premiums.

Of course we can’t claim to have exhausted all interesting relative value arguments to

establish bounds on a derivative price! The market is made by agents with possibly different

views on the underlying, and different utilities and risk preferences. We focus on providing an

exposition that OHMC can be used to systematically integrate replication based pricing ideas and

risk-preference or utility based ideas, without discarding the key tenets of either of these

approaches – i.e., a derivative trader can try to replicate what she can and express a risk

preference based on the residual risks inherent to attempted replication.

The OHMC methodology for analyzing hedging a Cliquet is completely independent of the

dynamics of the underlying asset. The ultimate practical application of the OHMC approach

may even involve employing a proprietary model of the asset returns that combines empirically

observed features as well as beliefs about the asset return nature. Such models also involve

conditioning on observations of underlying and possibly other explanatory factors. A synthesis

of econometric methods and attempted replication with quantification of average hedging costs

and hedge error distributions is possible within the OHMC framework.

A well known model of asset returns that sidesteps the perfect-hedge contrivance even

for vanilla options is afforded by a GARCH(1,1) description of asset returns (Bollersev [1986],

Engle[1994]). The conditioning variable is starting volatility, and the jumpiness of returns

associated with the return kurtosis thwarts the theoretically argued perfect hedge even under

continuous hedging (see Figure 1 and Appendix-II). We employ the GARCH(1,1) description

to provide examples of the OHMC method as well as to describe the risk-return of the delta-one

long only trader.

(

∆s k = s k µ∆t + σ k ∆t ε k ) (10)

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 19

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Standard Normal random-variates generated to create the return stochastic process in (10) are

denoted by ε k , and the volatility evolves per (11).

The empirical return statistics can be used to analytically specify the parameters of the

GARCH(1,1) model for the evolution of the reference asset. Here are details of the

unconditional moments used to infer the parameters (Carnero et al [2004]):

r ≡ ∆s / s; r = µ∆t (12)

2 2 2

r ′ ≡ r − r , σ r ≡ E[(r − r ) ] = ∆tσ (13)

The empirical mean and variance of the returns provide direct inferences of µ and σ through (12)

& (13). These could be purely historical or they could be based on ones views on the reference

asset looking forward. In addition to the long-term volatility the GARCH(1,1) model also has

the starting volatility as an input. This can be based on the volatility estimated from a smaller

window of data trailing the date of analysis.

Kurtosis

E[r ′4 ] 3

κ≡ = (14)

σ 4

r 2α 2

1 − 2

1 − (α + β )

(κ − 3)(κ (3 − 2β 2 ) − 3) − β (κ − 3)

α= (15)

3(κ − 1)

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 20

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

[

α 1 − (α + β )2 + α (α + β )

]

, h =1

1 − (α + β ) + α 2

2

ρ r′ (h ) =

[( )(

E r ′ 2 (t + h∆t ) − σ r2 r ′ 2 (t ) − σ r2 )]

= (16)

2

[(

E r ′ (t ) − σ

2 2 2

r )] (α + β )ρ (h − 1) , h >1

2

n

Γ (n) = ∑ ρ r ' ( j ) 2

j =0

(17)

Finding the value of parameter β that best reproduces the empirical auto-covariance of the

squared returns results in a complete inference of parameters. We chose the sum of the

autocorrelations to a maximum lag (17) as the composite correlation target. This sum is referred

to as the volatility clustering time as it quantifies a characteristic time-scale over which the

volatility is correlated. This can also be described as a characteristic time-scale over which

volatility fluctuations result in excess kurtosis being manifest in the realized return time series.

We iterate over β from 0 to 1 to find the GARCH(1,1) value of the sum of the autocorrelations

that is nearest to the empirical observation.

There are many facets to a Cliquet trading problem, which depend on tenor, strike, hedge

interval, look-back period, in addition to the characteristics of the underlying. We will attempt to

highlight the key role of the OHMC analysis for Cliquets by focusing on hedge performance and

by comparing the hedged P&L distribution with that of a long market agent, to delineate relative

value metrics. This is best accomplished through specific examples – for which we adopt 3

stylized descriptions of assets on which Cliquet contracts are written. These stylized asset

descriptions are in Table 2 and Figure 3. Asset 1 is representative of a single-stock. Asset 3 has

a generic broad market index profile. Asset 2 has the profile of a diversified alternative beta

product, be it basket of sample trades or a portfolio/index of hedge funds.

The characteristics of the stylized assets shown in Figure 3 are: (1) correlation between

the squared return residuals; (2) sample path simulation of volatility; (3) sample path simulation

of return; (4) risk-capital for delta-one long only trade; (5) expected return on risk capital for

delta-one long-only trade. The risk capital is assessed with a 1year 99.9% equivalent confidence

level statistical target.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 21

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Stylized Asset 1

Stylized Asset 2

Monthly Statistics r = 0.008333; σr = 0.023094; Γ (24) = 3 ; κ = 8

Stylized Asset 3

Daily Statistics r = 0.000476; σr = 0.010079; Γ (252 ) = 15 ; κ = 15

*Sensitivities to GARCH(1,1) parameters are shown in the next section.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 22

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

100% 100% 100%

90% 90% 90%

80% 80% 80%

autocorrelation

autocorrelation

autocorrelation

70% 70% 70%

60% 60% 60%

50% 50% 50%

40% 40% 40%

30% 30% 30%

20% 20% 20%

10% 10% 10%

0% 0% 0%

lag (business days) lag (months) lag (business days)

180% 25% 60%

160%

50%

140% 20%

120% 40%

15%

volatility

volatility

100%

volatility

30%

80%

10%

60% 20%

40% 5% 10%

20%

0% 0% 0%

0 252 504 756 1008 1260 1512 1764 2016 2268 2520 0 24 48 72 96 120 144 168 192 216 240 0 252 504 756 1008 1260 1512 1764 2016 2268 2520

business days months business days

20% 8% 8%

15% 6% 6%

10% 4% 4%

monthly return

daily return

daily return

5% 2% 2%

0% 0% 0%

-5% -2% -2%

-10% -4% -4%

-15% -6% -6%

-20% -8% -8%

0 252 504 756 1008 1260 1512 1764 2016 2268 2520 0 24 48 72 96 120 144 168 192 216 240 0 252 504 756 1008 1260 1512 1764 2016 2268 2520

business days months business days

delta-one long position delta-one long positiion delta-one long position

risk-capital (1 year 99.9%)

50% 50%

risk capital (1 yr 99.9%)

80%

40% 40%

60%

30% 30%

40%

kurtosis = 20 20% kurtosis = 8 20%

kurtosis = 15

20% kurtosis = 3 kurtosis = 3

10% 10% kurtosis = 3

0% 0% 0%

0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5 0 0.2 0.4 0.6 0.8 1

tenor (years) tenor (years) tenor (years)

delta-one long position delta-one long position

expected return on risk capital

expected return on risk capital

expected return on risk capital

60%

20% 15%

50%

15% 40%

10%

10% 30%

kurtosis = 20 kurtosis = 8 kurtosis = 15

20% 5%

5%

kurtosis = 3 10% kurtosis = 3 kurtosis = 3

0% 0% 0%

0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5 0 0.2 0.4 0.6 0.8 1

tenor (years) tenor (years) tenor (years)

Figure 3. For the three stylized assets analyzed here the auto-covariance of squared residuals,

sample path simulation of the volatility and asset return, the risk capital (1 year 99.9%

confidence level) and the expected change in wealth per unit risk capital for a delta-one long

only trader are shown above.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 23

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

The multiplicity of time-scales characteristic to a Cliquet contract is generally perceived to be the

main complexity over and above more vanilla contracts. It is shown in this section that in the

OHMC approach developed here these additional time-scales do not create an order of

magnitude additional complexity beyond the ones endemic to a vanilla contract. If one is ready

to explicitly deal with fat-tails and hedging errors endemic to vanilla contracts, then the

methodology of dealing with the exotic contract is not a world apart from the vanilla contract.

The OHMC framework is subject to being customized for a trader to express her risk

preference. That risk preference – subject to revision, criticism, and fitting - can be readily

expressed for the Cliquet contract too in OHMC. This is quite different from the presumed

perfect replication risk-neutral paradigm where the volatility surface or stochastic volatility

parameters fitted to vanilla contracts are often sought to be enforced on the exotic contract. The

fitting exercise (volatility surface or stochastic volatility parameters) to vanillas in the risk-

neutral frame-work is an unacceptable starting point to dealing with exotics because of two

reasons: (1) it is based on the perfect hedge paradigm which is inconsistent with the reality of the

asset behavior that gives rise to implied volatility smile-skew (2) the exotic contract can have

distinct drivers of unhedgable risks compared to the vanilla contracts used to fit parameters.

It is surprising that the presumed perfect replication (i.e., zero-risk) model is treated as

the fundamental building block to dealing with exotics in the risk neutral approach and that is

widely used in accounting of P&L. The continued sponsorship of such models seems to be

driven by the motivation of executing exotics that create upfront P&L when marked to market

using models “calibrated” to vanillas, often with little understanding of their hedging and

residual risk characteristics.

We are not offering a magical volatility surface or stochastic volatility parameters that

address Cliquet trading, hedging, and pricing (see Appendix-II for OHMC implied volatility

results). We are demonstrating a framework that requires developing an objective measure

description of the asset, delineating a hedging strategy, and assessing the performance of the

hedging strategy. The average cost of hedging and the hedge slippage distribution are central

results of the OHMC exercise – these are needed for responsibly designing and trading exotics.

Sensitivity Analysis

The different Cliquet Contracts analyzed here are summarized in Tables 3, 4, & 5 (Figures 4, 5,

& 6). The starting asset value and the Cliquet payout notional are both set to $100 for the results

presented here. The hedging costs and hedge slippage measures are calculated assuming upfront

payments, but reported in terms of running premiums for simplicity and to facilitate comparison.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 24

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

item case: 1 2 3 4 5 6

Tenor months 2 2 2 2 2 2

Look-back interval weeks 1 1 1 1 1 1

Roll interval days 1 1 1 1 1 1

Hedge interval days 1 1 1 1 1 1

K1 % 50 75 85 95 75 75

K2 % 0 0 0 0 0 0

mu (1/yr) 0.2 0.2 0.2 0.2 0.2 0.2

vol (1/yr^0.5) 0.4 0.4 0.4 0.4 0.4 0.4

kurtosis - 20 20 20 20 30 3

vol clustering time weeks 2 2 2 2 2 2

initial hedge notional %spot -0.025 -0.634 -3.052 -12.688 -0.6876 -0.0002

avg duration months 1.99 1.98 1.90 0.90 1.98 1.99

avg hedging cost bps/yr 0.75543 28 184 2327 30 0.00493

std dev residual P&L bps/yr 54 292 659 2211 309 2.31

bps/yr 26 351 577 3124 361 0.077

bad deal bound 1 multiple of avg

hedging cost

34.2 12.5 3.1 1.3 11.8 15.7

bad deal bound 2 multiple of avg

hedging cost

87.3 17.0 4.2 1.6 15.4 9.5

Strike

The more out of money that the Cliquet contract is, the greater is the hedge slippage compared to

the average hedging cost. This is shown in the strike dependence of Cliquet on Stylized Asset 1

in Table 3 and in the Appendix-II. As we look at more OTM strikes, the average hedging cost

decreases, the bounds on the pricing decreases, and a greater fraction of the bounding price is

based on hedge slippage. This interpretation of hedging error as a part of option value is

different from the risk-neutral practice where prices are simple averages of option payoffs under

de-trended underlying descriptions. In that risk-neutral approach there is no concession made for

replication errors (instantaneous hedging zero kurtosis case) and the whole distribution of the

underlying is distorted to fit an observed price – and that distribution is labeled risk-neutral.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 25

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

120 asset-1: case 3 0%

110 Sample path and distributional

100 -5% behavior for Cliquet on Stylized

asset value ($)

hedge-ratio

32% drop

Asset 1 case 3

90

-10%

80 The sample path shown here corresponds to

70 the 1 year 99.9% confidence level

-15%

asset-value ($) equivalent total wealth change over the

60

hedge ratio (%) options life – i.e., the left tail of the total

50 -20% change in wealth of the Cliquet put seller-

0 5 10 15 20 25 30 35 40 hedger. The P&L plots (b) do not show the

business days

(a) risk premium the option seller will add to

2 10 get compensated for residual risks.

asset-1: case 3

0 As shown in (a), the payout is triggered on

5

day 18, with the asset showing a drop of

-2

32% over the 1 week look-back period.

0

-4 While the P&L variance optimal hedging

Cliquet and hedge P&L ($)

-5

-6 days preceding the look-back, a loss of $15

-8 -10

(b). If a generally higher hedge ratio were

dw-hedge used, then the P&L volatility over the time

-10

-15 leading up to the trigger would be higher.

dw-total

-12 While OHMC can be tailored to minimize

-20 tail losses alone, there is no perfect hedge in

-14

the face of return kurtosis which is driven

-25 from volatility fluctuations in the

-16

GARCH(1,1) model employed here.

-18 -30

0 2 4 6 8 10 12 14 16 18 Note that the underlying asset returns to

hedge-interval (business days) (b) $100 in 12 days after the Cliquet is

triggered. This indicates the inherent

asset-1: case 3 riskiness of a Cliquet put relative to a

standard European option, especially for the

1 high kurtosis of asset 1 (kurtosis = 20,

cdf

similar to that of a single stock).

0.1

The asymmetry of the total wealth change

0.01 distribution (c) is important to recognize in

deciding to sell a Cliquet put. The bounding

0.001

prices (Tables 3-5) ensure that the expected

wealth change of the Cliquet seller per unit

0.0001

risk capital (at 1yr 99.9% confidence) is no

-10000 -7500 -5000 -2500 0 2500

worse than a delta-1 long position (Fig. 3).

hedge slip (bps/year) (c) OHMC analysis makes available all this

information en-route to valuation.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 26

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

item case: 1 2 3 4 5 6 7 8 9 10

Tenor months 36 36 36 36 12 24 36 36 36 36

Look-back interval months 1 2 3 4 3 3 2 3 4 3

Roll interval months 1 1 1 1 1 1 2 3 4 1

Hedge interval months 1 1 1 1 1 1 2 3 4 1

K1 % 85 85 85 85 85 85 85 85 85 85

K2 % 0 0 0 0 0 0 0 0 0 0

mu (1/yr) 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1

vol (1/yr^0.5) 0.08 0.08 0.08 0.08 0.08 0.08 0.08 0.08 0.08 0.08

kurtosis - 8 8 8 8 8 8 8 8 8 8

vol clustering time months 3 3 3 3 3 3 3 3 3 1.5

initial hedge notional %spot -0.16 -1.25 -3.01 -3.99 -1.51 -2.48 -0.33 -0.47 -0.55 -3.74

avg duration months 33.9 33.7 33.6 33.5 11.7 22.9 33.8 33.8 33.7 33.6

avg hedging cost bps/yr 0.65 4.74 9.57 12.27 6.15 8.51 3.03 5.14 6.59 11.36

std dev residual P&L bps/yr 5.97 16.64 21.21 22.18 31.92 25.47 15.48 20.83 24.54 21.95

bps/yr 2.9 64 95 101 102 101 50 79 97 104

bad deal bound 1 multiple of avg

hedging cost

4.4 13.6 9.9 8.3 16.6 11.9 16.5 15.3 14.7 9.2

bps/yr 8.3 142 202 209 173 195 89 126 147 215

bad deal bound 2 multiple of avg

hedging cost

12.6 29.9 21.1 17.0 28.0 22.9 29.5 24.5 22.4 19.0

Tenor

For Cliquets we do not expect to find strong sensitivity of the running premium with tenor – as

the risk of asset falling over the look-back interval is not directly influenced by the tenor.

However we need to consider the volatility clustering time-scale in judging the sensitivity of the

deal tenor. If the volatility clustering time-scale is a significant fraction of tenor, then the choice

of starting volatility becomes important. We remind the reader that in all our examples we are

setting starting volatility to the long-term historical average. So, if the volatility clustering time-

scale is a significant fraction of tenor then the full range of volatility fluctuations are potentially

not experienced. We witnessed that for the vanilla option in Appendix-II.

The other consideration in judging the import of tenor is the way a risk-premium is

assessed. A global risk premium adds the wealth change over all the hedge intervals. The global

risk premium (bound 1) is based on the sum of wealth changes over the deal tenor. The

summation results in some cancellations among gains and losses and results in a tighter wealth

change distribution as the averaging interval increases. In contrast, the temporally local risk

premium (bound 2) assessment does not benefit from cancellations over longer tenors and

therefore is expected to be less dependent on tenor, and can also become larger if the transaction

tenor becomes longer than the volatility clustering time such that the full range of asset volatility

and kurtosis is felt. For the Cliquet examples on stylized asset 2 we find that the global risk

premium based bound decreases slightly with tenor, however the local risk premium based

measure actually increases as the volatility forgets its starting value and bounces around.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 27

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

115 -2% Sample path and distributional

asset-value ($)

110

22 % drop

behavior for Cliquet on Stylized

hedge-ratio

-4%

105

Asset 2 case 3

-6%

100 As shown in (a), the payout is triggered on

asset-value ($) month 16, with the asset showing a drop of

95 -8%

hedge-ratio (%) 22% over the 3 month look-back period.

While the P&L variance optimal hedging

90 -10%

amount increases from 0.5% to 5.5% over

0 12 24 36

month the days preceding the look-back, a loss of

(a)

$4.6 is experienced on the day payout is

triggered (b).

asset-2: case 3 independent of the absolute level of the

1

asset. This asset climbed up rapidly (from

0 $100 to $115 in one month) and then

monthly P&L ($)

-1

dw-cliquet

dropped 22 % triggering the Cliquet.

-2 dw-hedge

The hedge performance associated with a

-3 dw-total tail loss event (1 year 99.9% confidence

-4 level) is shown in (b). Over the period

-5 leading to the trigger event the hedging

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 works – albeit not perfectly. The P&L

hedge-interval (month) (b) variance optimal hedging fails to prevent the

large loss in the event of trigger in the

shown sample path.

asset-2: case 3 losses in the event of trigger, but increase

1 pdf the volatility otherwise. Ways to improve

the hedge performance by conditioning on

cdf

0.1 additional information are briefly mentioned

in the discussion section later. In no case do

0.01 we expect a perfect hedge. Hence the

importance of the residual P&L distribution,

0.001 shown in (c). At the time of trade

execution, a delineation of hedging strategy

0.0001 and residual risks is the key result of OHMC

-600 -500 -400 -300 -200 -100 0 100

analysis.

hedge slip (bps/year) (c)

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 28

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

item case: 1 2 3 4 5 6 7

Tenor yr 0.5 0.5 0.5 0.5 0.5 0.5 0.5

Look-back interval day/week/month 1 day 1 week 1 month 1 month 1 month 1 month 1 month

Roll interval day/week/month 1 day 1 day 1 day 1 day 1 day 1 day 1 day

Hedge interval days 1 day 1 day 1 day 2 day 3 day 4 day 5 day

K1 % 90 90 90 90 90 90 90

K2 % 80 80 80 80 80 80 80

mu (1/yr) 0.12 0.12 0.12 0.12 0.12 0.12 0.12

vol (1/yr^0.5) 0.16 0.16 0.16 0.16 0.16 0.16 0.16

kurtosis - 15 15 15 15 15 15 15

vol clustering time days 15 15 15 15 15 15 15

initial hedge notional %spot -0.0073 -1.04 -4.90 -4.68 -4.51 -4.32 -4.21

avg duration years 0.495 0.485 0.454 0.454 0.454 0.454 0.454

avg hedging cost bps/yr 0.47 16 58 58 58 58 58

std dev residual P&L bps/yr 15 84 122 124 126 127 128

bps/yr 26 136 174 176 179 180 183

bad deal bound 1 multiple of avg

54.5 8.5 3.0 3.0 3.1 3.1 3.2

hedging cost

bps/yr 80 350 328 297 281 271 265

bad deal bound 2 multiple of avg

170.0 21.8 5.7 5.1 4.9 4.7 4.6

hedging cost

Look-Back Interval

By comparing case 1, 2, and 3 of Table 5 on asset 3 the role of look-back interval is illustrated.

The average hedging cost, the hedge notional, and the residual risks increase with look-back

interval. However the ratio of the residual risk to average hedging cost decreases with look-back

interval. This is similar to the behavior observed for a vanilla put with respect to tenor, as shown

in Appendix-II. This decay of the hedge slippage as a fraction of average hedging cost is likely

associated with the increasing efficacy of hedging as multiple hedge adjustments are made in

between the look-back interval, and in the case of bound-1, it is likely due to the effect of

temporal aggregation in shrinking the width of the wealth change distribution due to

cancellations of hedge slips of opposite signs.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 29

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

130 asset-3: case 3 0%

120 -2%

-4%

Sample path and distributional

110

behavior for Cliquet on Stylized

asset-value ($)

-6%

hedge-ratio

15.9% drop

100 -8%

Asset 3 case 3

90 asset-value ($) -10%

hedge-ratio (%) -12%

80 Like shown in Figures 4 & 5, the sample

-14%

70 path shown here corresponds to the 1 year

-16%

60 -18% 99.9% confidence level equivalent total

0 10 20 30 40 50 60 70 80 90 100 110 120 wealth change over the options life – i.e.,

business days (a) the left tail of the total change in wealth of

the Cliquet put seller-hedger. The P&L

plots do not show the risk premium the

1 asset-3: case 3 4

option seller will add to get compensated for

Cliquet & hedge daily P&L ($)

0 2 residual risks.

-1 0

As shown in (a), the payout is triggered on

-2 -2 day 18, with the asset showing a drop of

-3 dw-cliquet -4 15.9% over the look-back period. While the

dw-hedge P&L variance optimal hedging amount

-4 -6

increases from 0.5% to 4% over the days

dw-total

-5 -8 preceding the look-back, a loss of $5.5 is

-6 -10 experienced on the day payout is triggered

0 5 10 15 20 25 30 35 40 45 50 55 (b). If a generally higher hedge ratio were

used then P&L volatility over the time

hedge-interval (business days)

(b) leading up to the trigger would be higher.

There is no perfect hedge in the face of

asset-3: case 3 return kurtosis which is driven from

volatility fluctuations in the GARCH(1,1)

pdf model employed here.

1

cdf The asymmetry of the total wealth change

0.1 distribution (c) is important to recognize in

deciding to sell a Cliquet put. The bounding

0.01 price presented in this work ensures that the

expected wealth change of the Cliquet seller

0.001

per unit risk capital is no worse than a

0.0001

simple delta-1 long position.

-2000 -1500 -1000 -500 0 500

hedge slip (bps/year) (c)

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 30

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Roll Interval

If the roll-interval is increased then the instances that the payout condition is checked decreases

and one expects average hedging costs to go down as well as the hedge slippage measures. In

Table 4 we see that happens (despite an increasing hedging interval) by comparing case 2 and 7

and case 3 & 8. As shown in Figure 4, 5, a payout trigger can be associated with a rapid decline

of the asset which may be followed by an increase – a larger roll interval makes the protection

seller immune from some of these reversals, hence decreasing the value of the Cliquet put.

Kurtosis

The return kurtosis mediates the pricing dynamics in two ways: (1) higher kurtosis makes the

asset values have greater probability of being realized away from the spot, hence shifting average

hedging costs away from at-the-money (ATM) strikes to out-of-the-money (OTM) strikes; (2)

the residual risks are controlled by kurtosis – as in the limit of continuous hedging a kurtosis

value set to 3 enables the perfect hedging limit, whereas for kurtosis > 3 the residual risks can be

significant. So, for some strikes that are ATM or not too OTM the pricing can become

insensitive to kurtosis as it become large as the mean hedging costs shift to more OTM strikes

but the hedge slippage relative to average hedging costs gets worse.

kurtosis going from 3 (Geometric Brownian Motion) to 20-30 which is characteristic of a typical

daily return single stock time-series. For the OTM strike considered in those examples, we see a

dramatic increase in average hedging cost and hedge slippage error measures in going from

kurtosis of 3 to 20, and a relatively muted difference in going from 20 to 30.

The volatility clustering time-scale is the time-scale over which the volatility tends to forget its

starting value and bounces around over the range of volatilities characteristic to the GARCH(1,1)

description. Note that the realized volatility of volatility is linked with realized kurtosis in the

GARCH(1,1) model employed here. If the volatility clustering time is smaller than the look-

back interval, then the associated return jumpiness is seen by the Cliquet contract and one

expects higher hedging costs on the average and higher hedge slippage measures. Table 4

presenting results on stylized asset 2 (case 2 compared with case 10) shows the impact of

decreasing the volatility time-scale from the look-back interval (3 months) to half the look-back

interval. We see an increase in hedging costs and hedge slippage standard deviation associated

with the halving of the volatility clustering time. The tail hedge slippage measures react much

more than the mean and standard deviation of hedging costs– as witnessed in the larger

differences in the bounds on the Cliquet sell price. If the volatility clustering time were to

become larger than the look-back interval then the fat-tails of the underlying asset (excess

kurtosis) are not manifest strongly in the hedge analysis of a Cliquet.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 31

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Hedging Frequency

As shown in Figure 1 in the introduction, for a vanilla put, the total hedging error decreases as

the hedging frequency increases, albeit to a significant irreducible value for realistic asset models

that exhibit excess kurtosis. We see the same feature in the Cliquet contract, as illustrated in

Table 5 case 3 through 7. In those examples the mean hedging cost is not visibly impacted, but

the hedge slippage standard deviation increases with hedging interval. Bound-1 on the Cliquet

price – which is based on a time-aggregated hedge slippage measure - increases with hedging

interval. This is due to the widening of the loss-tail of the total change in wealth distribution

with an increase in the hedging interval. However bound-2, that assesses risk premiums based

on hedge error slippage over individual hedge intervals - albeit much larger than bound-1,

decreases with infrequent hedging for the example shown here. The rate of decrease of bound-2

with the hedging interval decreases with an increase in the hedging interval – indicating

competing influences that balance out over larger intervals. As the hedging interval increases,

the 1 yr 99.9 equivalent confidence level over the hedging interval becomes less deep into the

left tail of the hedge interval based wealth change distribution. So as the wealth change

distribution itself is likely to be getting wider due to increasing hedging errors, by looking less

deep into the left tail one can actually expect a local risk premium to be smaller. One must also

remember that as the hedging interval increases, the bounding return on risk capital associated

with a delta-one position also increases. These competing influences can explain the

computationally simulated behavior of bound 2 as a function of the hedge interval.

6. Discussion

Relationship With Prior Work on Cliquets

Cliquet type derivatives appear in all type of exotic flavors and colors (see Gatheral [2006]):

Reverse Cliquets, locally capped-globally floored Cliquets, Napoleons (a distinct French tilt in

this market). Many of the early structures were fixed income in nature where the periodic

coupon of the note was determined by a call or put Cliquet. A review of the market up to early

2004 has been given by Jeffery [2004]. One of the main difficulties in these derivatives is the

exposure to the forward smile-skew of the underlying. Wilmott [2002] pointed this out and

assessed that it results in a rather high sensitivity of the underlying risk model used in pricing

such structures. As discovered by leading dealers in the late 1990’s, using a local volatility

model can seriously underestimate the volatility-smile sensitivity of these derivatives (a couple

of dealers are rumored to have lost several 100’s of millions of dollars). As mentioned by

Jeffery [2004], “dealers…failed to fully factor in hedging costs in deals written in 2002 and early

2003”. The crux of the problem was stated as follows: “While sensitivity towards the volatility

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 32

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

net change over the life of the equity option is often described as a ‘second order’ parameter in

classical options theory [Black Scholes] , effectively modeling the so-called vega convexity in

reverse Cliquets and Napoleon options is probably the most critical component in pricing.” In a

nutshell, the sensitivity to the “volatility of volatility” is extremely high. Therefore, any model

that purports to accurately value such structures must take into account changing volatility

regimes and dynamic hedging with a view to address large sudden moves in the underlying asset.

As Wilmott [2006] succinctly states, “The way in which the volatility impacts the price of this

contract [Cliquet] is subtle to say the least, so it makes the perfect subject for an in-depth study

which I hope will reveal how important it can be to get your volatility model right”.

After the debacle of the local volatility model in dealing with Cliquets, the knee-jerk

reaction of a mainstream valuation modeling is to introduce stochastic volatility models.

However, fitting parameters to expectations of payoffs with de-trended descriptions of

underlying does not address replication directly – so one should not expect insights into hedging

just because the underlying comes from a stochastic volatility model. To add to that, even the

most popular such stochastic volatility model, the Heston Stochastic Volatility Model, have

highly variable parameters while fitting to Vanilla option prices within the standard risk-neutral

fitting framework. A comparison of stochastic versus local volatility models is given in

Gatheral, [2006]. Our view is that Cliquet prices must be driven from a synthesis of average

hedging costs plus a premium for unhedgeable risks associated with the treacheries of repeated

vega-convexity flare-ups associated with the reset dates.

This work has addressed both the issues of a changing volatility and hedging analysis.

The sensitivity of average hedging costs and hedge slippage to kurtosis and the volatility

clustering time-scale described in the previous section squarely address points raised by Jeffrey

[2004] and Wilmott [2002] & [2006]. The GARCH(1,1) parameters have been calibrated to the

objective measure of the underlying. That parameter fit tends to be more stable than fitting a risk

neutral model to option prices. The ability of GARCH(1,1) to represent the richness of the

dataset and the return of the underlying depends on the length of the data, and one can always

argue about extreme events not captured within a finite dataset, and empirical statistical

characteristics not captured in GARCH(1,1). Those criticisms can be objectively adjudicated by

examining datasets of different lengths and/or out-rightly specifying statistical characteristics

that are desired and refining the objective measure description (including adding a death state).

We are not adverse to the Heston model. In fact we use the Heston model (among others) within

the OHMC framework - but calibrated to the objective measure of the underlying (using a

similar method of moment matching as in the GARCH methodology in section 4). We do not

find any value in the calibration of risk-neutral models to vanilla option prices (without

analyzing hedging and residual risks) and making assessments of valuation of Cliquets or other

exotics, presuming perfect replication. That does not address the risk-return dynamics of the

Cliquet contract or any other exotic derivative.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 33

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

The recognition of the idea of replication and the evolution of quantitative approaches from the

Bachelier price (average of future cash-flows) to a replication cost price (cost of mitigating risks)

is widely reflected in current quantitative finance practice. The idea of replication is powerful

and convenient. Indeed, when perfect replication is feasible, one can simply take a statistical

average of the option payoff under a de-trended description of the underlying asset (relative to

carry cost) and the user of the model does not have to deal with any statistical risk measure

(other than the average of a hypothetical distribution). As the marketplace gives birth to

complex derivatives at a rapid pace, implied parameters fitted to observed prices of vanilla

options are often used in the valuation of complex derivatives by taking statistical averages of its

payoff with a de-trended underlying. The upfront P&L and/or carry for an exotic derivative,

found by using a valuation model employing parameters fitted to vanillas, are of great interest to

businesses.

inconvenience. If the vanilla derivative contract cannot be practically replicated, then its price

should/could reflect a mix of some average hedging cost plus possibly a risk premium for

unhedgable risks. In fitting a parameter of a presumed perfect replication model to the observed

prices of a far from perfectly replicable option one may end up believing that the prevailing risk

premiums are being correctly or adequately represented, and will be effectively propagated to

the exotic derivative valuation. This is how valuation models create an appearance of an

intelligent exercise of risk aversion. However at no stage is risk actually being assessed in this

cascading use of averages of payoffs under de-trended descriptions. The asymmetries of

residual risks and the market signals about them in the bid-offer of vanilla contracts are not

explicitly propagated into the value of the exotic derivative by taking averages under de-trended

descriptions. The increase in complexity of the derivative contracts is therefore not accompanied

by a better understanding of the hedging strategy or a quantification of residual risks in this mode

of valuation modeling that has taken hold of accounting practices for simple and complex

derivatives. The purported unique price assessed using such valuation models and the ensuing

implications for upfront P&L and/or carry beclouds the substantive business risk management

issues of residual risks inherent in any attempted replication and assessment of risk-return. Thus

we find ourselves surrounded by a plethora of “risk-neutral” valuation models whose main

claims are convenience and rapidity of fitting – with no analysis of hedging, attempted

replication and residual risks being done en-route to valuation.

Alternatively, as pursued in this paper, one can start with a description of the underlying

and explicitly analyze hedging and try to find the best hedging strategy as done in the OHMC

approach. If the hedging strategy is perfect, then the cost of hedging is the value of the

derivative. On the other hand if the hedge performance is not perfect, then the probability

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 34

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

distribution of the residual P&L is useful. The OHMC method is mindful of the power of

replication, but puts the burden of invoking it on the designer of the hedging strategy and

tempers it with an estimate of the hedge slippage, accounting for characteristics of the derivative

contract when the underlying asset exhibits jumpy returns. While the routine water cooler talk

with a savvy trader or risk manager could reveal the catalogue of unhedgable risks endemic to an

attempted replication strategy, the mainstream quantitative valuation models have found it

difficult to resist the allure of presuming perfect replication. This aversion to attempting to deal

with the reality of imperfect replication is perhaps due to the steep gradient in going from taking

averages under de-trended underlyings, to actually articulating a hedging strategy and elucidating

its limitations. Another reason mainstream quantitative valuation modeling has avoided dealing

with imperfect replication is, perhaps, that imperfect replication challenges the notion of an

unassailable or unique model price, which has become the main accounting goal of mainstream

valuation models. Rather than abandon addressing the imperfect replication situation, OHMC

provides assessments of average hedging costs and residual hedging errors that can be used by a

market agent in judging the price at which a market opportunity presents itself. While OHMC in

itself does not address the diversifiability of residual risks, or the risk preferences of distinct

market agents, the information it provides can be used by a market agent to express her risk

preference in the contexts of her trading book, and in effecting a hedging strategy.

We believe that the derivatives trader and his business and risk managers benefit from the

in-depth hedge performance analysis that is provided by OHMC, prior to pricing any derivative

trade. In using OHMC, results on hedge performance and residual risks are available along with

any assessment of valuation and the ensuing P&L (or trade carry). As a result, the upfront P&L

and carry of the hedged derivative position can be readily compared with irreducible hedging

errors. Hence a relative value metric for the trade is available as a part and parcel of valuation

using OHMC. Thus, the OHMC framework is a suitable tool for providing a consistent view of

derivatives trades to trading, risk-control, and product-control.

Future Work

For problems with one major risk factor, the model developed here demonstrates the practical

feasibility of OHMC for any option problem, including path dependent problems. The key to

efficient implementation is recognizing the limited support of the basis functions (Appendix-III)

while assembling the set of equations to effect the numerical solution to the variational problem

(Appendix-I). Employing total wealth change quintiles as calibration targets, rapid fitting (by

post-processing OHMC algorithm output) to observables can also be achieved – but the main

benefit being the availability of hedge slippage measures at the time of pricing. So the OHMC

based valuation model is a tool of trading strategy and risk management while it is being used to

mark-to-market a trading book. Extensions to other hedging error measures – say expected

shortfall – are also feasible.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 35

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Based on this work, we also believe that OHMC can be applied to practical applications

involving multidimensional problems, including multi-name credit derivatives. For example,

results on static hedge optimization for synthetic CDOs have been reported by Petrelli et al,

[2006] and implications of uncertainty of realized correlation of spreads on hedging trades that

are long correlation (and long carry) are assessed in Petrelli et al [2007]. Any sort of

‘correlation-trading’ (best-of-baskets, multi-asset options, CDOs) should benefit from

understanding the role of uncertainty in realized correlation on hedge slippage, in addition to the

hedging error driven by marginal fat tail distributions. Providing a practical primer on hedging

while ‘correlation-trading’ and the ensuing implications for risk-return should be a fruitful target

of multidimensional applications for OHMC.

Even in seemingly single asset problems, such as the Cliquet analysis pursued here,

conditioning on explanatory variables could make the hedging strategy more effective. For

instance, if there is temporal persistence of volatility, then treating the squared residuals of return

as a second variable (not necessarily traded) that is observed and accounted for in the hedge

optimization holds the prospect of improving hedge performance. Such conditioning can be

effected in the multidimensional OHMC framework. We will next report on efficient two and

three dimensional OHMC implementations and present examples employing multi-dimensional

jumpy assets.

Acknowledgements. The authors gratefully acknowledge Joseph Hedberg for reviewing drafts of this

document, and Ram Balachandran for an efficient implementation of the OHMC algorithm described in

Appendix I & III.

____________________________________________________________________________________

JEL Classification

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 36

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Appendix-I

The change in wealth of a Cliquet protection seller-hedger follows

∆Wt (t k , t k +1 ) = C k ( sk ) − Gk + Φ k (sk )H k

k (I-1)

(I-2)

In (I-1) & (I-2) Ck(.) and Фk(.) are unknown functions that we seek to find to accomplish the

following statistical goal:

E[∆Wtk (t k , t k +1 )] = 0

(I-3a)

Minimize:

σ ∆2W tk (tk ,tk +1 ) {

= E[ ∆Wtk (t k , t k +1 ) − ∆Wtk (t k , t k +1 ) ] }

2

(I-3b)

Minimizing the wealth change variance with a zero mean change of wealth is equivalent

minimizing the mean squared change in wealth with a zero mean change of wealth. To render

this variational problem finite-dimensional we represent the option value and hedge notional

functions of spot in a finite dimensional representation

M C −1 MΦ −1

Ck ( sk ) = ∑aj =0

k

j A j (sk ) Φ k (sk ) = ∑b

j =0

k

j B j (sk ) (I-4)

The finite-dimensional representation used for the computations of the main section is detailed in

Appendix-III. Using (I-1) through (I-4) the first two statistical moments of the option-seller-

hedger’s wealth change follow

M C −1 M Φ −1

[

E ∆Wtk (t k , t k +1 ) = ] ∑ a E[A (s )] − E[G ] + ∑ b E[B (s

k

j j k k

k

j j k )H k ] (I-5)

j =0 j =0

M C −1 M Φ −1

2

[ 2

]

E ∆Wtk (t k , t k +1 ) = E ∑ a j A j (s k ) − G k + ∑ b j B j (s k )H k

k k

(I-6)

j =0 j =0

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 37

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

To solve for the unknown coefficients a kj and b kj we employ a Lagrange multiplier technique:

2

(I-7)

dFk

=0 0 ≤ i ≤ MC −1 (I-8)

daik

dFk

=0 0 ≤ i ≤ MΦ − 1 (I-9)

dbik

dFk

=0 (I-10)

dγ

Substituting (I-5) and (I-6) into (I-7)-(I-10) defines the set of linear equations that must be solved

to solve the finite-dimensional approximation of the Cliquet optimal hedge-valuation variational

problem. Further details of these linear equations are provided here. (I-8) through (I-10) can be

expressed as

Gij h j = q i 0 ≤ i, j ≤ M C + M Φ (I-11)

A sum over the repeated index j is implied in the above (I-11). The vector of unknowns is

represented by hj:

h j = a kj 0 ≤ j ≤ M C −1

h j = b kj − M C M C ≤ j ≤ M C + MΦ −1 (I-12)

hM C + M Φ = γ

From (I-8)

for 0 ≤ i ≤ M C − 1

qi = E[ Ai (sk )Gk ] ;

Gij = E[ Ai ( s k ) A j ( s k )] 0 ≤ j ≤ M C −1

Gij = E[ Ai ( s k ) B j − M C ( s k ) H k ] M C ≤ j ≤ M C + MΦ −1 (I-13)

Gij = E[ Ai ( sk )] j = M C + MΦ

From (I-9)

for M C ≤ i ≤ M C + MΦ − 1

q i = E[ Bi − M c ( s k )G k H k ]

Gij = E[ Bi− M C ( sk ) A j ( sk ) H k ] 0 ≤ j ≤ M C −1 (I-14)

2

Gij = E[ Bi − M C ( s k ) B j − M c ( s k ) H k ] M C ≤ j ≤ M C + MΦ −1

Gij = E[ Bi − M c ( s k ) H k ] j = M C + MΦ

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 38

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

From (I-10)

for i = M C + M Φ

qi = E[Gk ]

Gij = E[ A j ( s k )] 0 ≤ j ≤ M C −1

Gij = E[ B j − M c (s k )H k ] M C ≤ j ≤ M C + M Φ −1 (I-15)

Gij = 0 j = MC + MΦ

Solving for the pricing and hedge notional is an ordinary minimization problem at the first time

step:

∆Wt (t 0 , t1 ) = C0 − G0 + Φ 0 H 0

0 (I-16)

C0 and Ф0 are the unknown quantities in (I-16). We define perturbed quantities as deviations

around ensemble averages

G0' = G0 − G0 ; H 0' = H 0 − H 0

The solution for C0 and Ф0 that enforce zero mean change in wealth and that minimize the

wealth change variance are

G0' H 0'

Φ0 = C0 = G0 − Φ 0 H 0

2 ; (I-17)

H 0'

The OHMC algorithm to solve the variational problem for a Cliquet is not too different from that

to solve the vanilla equity option hedging problem. We start from the option maturity and work

backwards, solving for the option spread value and optimal hedge notional. Payout trigger

events and the hedging interval are explicitly accounted for in the wealth balance. For the

Cliquet hedging problem, all the statistical averages are conditioned on knockout not having

occurred at the starting time step. The optimal hedge ratio and average hedging cost value are

conditioned on there being a live Cliquet contract.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 39

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

The similarity of OHMC algorithm between vanilla options and exotics is attractive because it

enables consistency in the analysis of vanilla options and more exotic options. This consistency

is further reinforced if one uses identical descriptions of the process underlying the derivative

contract. If the exotic option is sensitive to a particular time scale description of the process, or,

extreme tail behavior, then one needs to refine the description of the underlying based on

available empirical information or ones view. That view can be a proprietary view and developed

by comparison with other assets for which more empirical information is available. The vanilla

option analysis may also benefit (or at least not hurt) from such a refined stochastic description

of the underlying. The hedge ratios resulting from the OHMC analysis have a concrete real-

world objective– and there is no need to switch measures and descriptions of underlying between

valuation analysis and assessment of hedge ratios. The model parameters fitted to vanilla prices

(without addressing irreducible hedging errors) do not have to be imposed on the exotic

derivative under the superficial guise of consistency.

It is our experience that exotics can have distinct sensitivities that control the hedging

errors in attempting to replicate them, and the option trader needs to directly focus on them,

rather than being bound by say a volatility surface that originates from fitting volatilities to

vanillas. This is practically important because often upfront P&L resulting from the imposition

of volatility surfaces that fit vanillas onto exotics can end up becoming the motivation for doing

a trade, rather than an argument based on risk-return of the exotic option. In the OHMC

framework, consistency of analysis of vanilla and exotic option requires the following:

1. Employ the same empirically realistic description of the underlying to the vanilla and

exotic derivative problem;

2. Assess the optimal hedging strategy for both the vanilla and exotic along with the average

hedging costs;

3. Assess residual hedging errors – be they driven by fat tails, or by discrete hedging

intervals for both the vanilla and the exotic option;

4. Where market prices for the vanilla option are readily available, interpret them based on

risk return metrics derived from average hedging costs and residual risks, and develop a

picture of market risk aversion that is cognizant of demand and supply for the derivative;

5. Develop a view of exotics pricing based on average hedging costs and deviations around

that average. If there are common elements of risk-return between an exotic and a vanilla

derivative, then the vanilla derivative observable pricing can help guide the exotic option

trader to a competitive pricing point and the associated risk-return of the trade.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 40

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Appendix-II

Examples of variance optimal hedging analysis of sell vanilla option positions are presented

here. The stylized asset descriptions employed in the main text (Section 4) are used here. We

present hedging costs and residual risks arising from a combination of asset return kurtosis and

discrete hedging interval. We also compare the P&L distribution of a trader that sells options

and hedges to that of a delta-one trader that is simply long the underlying asset. Those relative

value metrics are also cast in the form of bonds on the option price (detailed in the main section)

and corresponding bounding implied volatility surfaces.

In applying risk-neutral models to value exotics, often the starting point is fitting their

parameters to the observed prices of vanilla calls and puts. Unlike the risk-neutral formalism,

OHMC method does not presume perfect replication and makes available to the user the residual

risks in attempted replication. OHMC enables interpreting the observed prices as a combination

of average hedging costs and the distribution around that average. From a pure accounting mark-

to-market perspective if the objective is merely to fit a specific price, then one can calibrate to

the different quintiles around the average hedging cost in OHMC. The convenience of fitting the

risk neutral formalism based models to derivative price data should not be an excuse for not

knowing hedging errors endemic to attempted option replication. OHMC offers an ability to fit

prices too, although that is not its main goals. More importantly, OHMC develops and back-tests

the hedging strategy and makes available average hedging costs and hedging errors.

OHMC Methodology

Denoting the optional value and hedge ratio at time tk by Ck and Фk, the change in wealth of the

option seller-hedger follows:

k k k

Using the symbols that are convenient in the Cliquet OHMC algorithm, we rewrite (II-1) as

k

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 41

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

The OHMC solution is propagated from maturity to the option starting point, where it becomes

an ordinary minimization problem, as formulated in Appendix-I.

Example 1. Sell Put and Delta Hedge Stylized Asset 3

The disparity in hedge slippage error between a realistically fat-tailed distribution and an

idealized return distribution with no kurtosis is visible in Figure 1 (main section 1). Kurtosis >

3 renders the hedging error irreducible as the hedging frequency increases. For Geometric

Brownian Motion the hedging error decreases much more rapidly with hedging frequency, and is

headed to zero in the limit of continuous hedging. This result shows how much of the

mathematical machinery to deal with continuous hedging while ignoring realistic return fat tails

is of marginal practical importance because in the limit of continuous hedging the transaction

costs would become unboundedly large, and in practically realistic returns, after a point, hedging

more often does not reduce the hedging errors. The OHMC framework handles discrete hedging

without having to assume zero excess kurtosis for the underlying.

1 1

kurtosis=15 kurtosis = 15

0.1 0.1

pdf

cdf

0.01 0.01

0.001 0.001

0.0001 0.0001

-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2

total change in wealth of put seller-optimal hedger ($) total change in wealth of put seller-optimal hedger ($)

Figure II-1. Probability and cumulative density functions of the total wealth change ∆W0 (0, T )

of the put seller-daily variance optimal hedger. The OHMC algorithm imposes a zero mean

change in wealth E[∆W0 (0, T )] = 0 and minimizes E[ ∆Wtk (t k , t k +1 )2 ] = 0 over every hedge

interval. The average cost of hedging, C0, and the residual P&L distribution is provided by

OHMC by “back-testing” the trading strategy as a part of the MC simulation. In this work we

report multiple measures of value: (1) average hedging cost; (2) average hedging cost plus risk

premium that renders the expected return on risk capital identical to a trade that is long the

underlying asset. The risk premium is assessed over the whole deal life (giving rise to bound 1),

and alternatively, over each hedging interval (bound 2), as detailed in the main section (see

section 3 for a full description of these bounds).

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 42

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Example 2. Sell Put and Daily Delta Hedge on Stylized Asset 1

hedge average hedging cost 1 year 99.9% equivalent quintile of Sum of 1 year 99.9% equivalent quintile

strike K sum of hedging errors based of hedging errors based

notional (%

($) bound (1) bound (2)

spot)

Black-Scholes multiple of avg Black-Scholes multiple of avg Black-Scholes

% spot % spot % spot

implied vol (%) hedge cost implied vol (%) hedge cost implied vol (%)

90 -17.5 1.002 39.8% 1.580 1.58 47.5% 1.440 1.44 45.7%

100 -46.4 4.176 37.8% 4.858 1.16 43.7% 4.685 1.12 42.2%

110 -75.4 10.987 38.8% 11.706 1.07 46.7% 11.513 1.05 44.6%

120 -90.2 19.997 41.7% 20.733 1.04 55.1% 20.501 1.03 51.5%

60%

Black-Scholes implied volatility

55%

50%

45%

40%

35%

historical avg hedge cost

30%

bad deal bound1 bad deal bound2

25%

80 90 100 110 120

strike ($)

Table & Figure II-2. OHMC results for a sell 1 month put with daily delta hedging (Example 2).

OHMC evaluates the P&L variance optimal hedge ratio and the residual hedge errors. The bad deal

bound for the seller is the put sell price below which the put sellers expected return on risk capital (at 1 yr

99.9% confidence level) is below that of a simple long position in the asset. The risk capital is assessed

based on the hedging errors that are residual in the variance optimal hedging strategy. The bad-deal

bound-1 is based on equating the option seller-hedgers return on risk capital to a delta-1 trade over its life

– whereas bad-deal bound-2 simply focuses on the discrete hedging interval risk-return.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 43

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Kurtosis Sensitivity

The central role of return kurtosis in thwarting perfect replication and creating a dependence of

implied volatility with strike is demonstrated in the results presented in this Appendix (Figure 1

main section & Figure II-1 and II-2). Now we examine the impact of changing the kurtosis,

keeping all else equal. Below are shown results for Example 2 with κ = 30. The increase in

kurtosis results in a greater chance that the asset values will end up further away from its average

value – hence the average hedging around the spot decreases, but increases away from spot.

However as shown in the first example, the residual risks are also an increasing function of

kurtosis. The option seller’s bounds on pricing, found by adding residual risks to the average

hedging cost, increased for all the strikes for Example 2.

average hedging 1 year 99.9% equivalent quintile of Sum of 1 year 99.9% equivalent

hedge cost

strike K sum of hedging errors based quintile of hedging errors based

notional (%

($) bound (1) bound (2)

spot)

Black-Scholes multiple of avg Black-Scholes multiple of avg Black-Scholes

% spot % spot % spot

implied vol (%) hedge cost implied vol (%) hedge cost implied vol (%)

90 -17.58 1.000 39.8% 1.596 1.60 47.7% 1.442 1.44 45.7%

100 -46.39 4.164 37.7% 4.866 1.17 43.8% 4.681 1.12 42.2%

110 -75.38 10.981 38.7% 11.715 1.07 46.8% 11.511 1.05 44.6%

120 -90.06 19.999 41.7% 20.754 1.04 55.4% 20.506 1.03 51.5%

Table II-3 OHMC results for a sell 1 month put with daily delta hedging (Example 2 with κ = 30).

In this example the bad deal bound-1 is larger than bad deal bound-2. In comparing bound 1 and

bound 2, two factors need to be kept in mind:

• In bound 1 the residual hedging errors are summed up over the option tenor and the total

hedging error tail is employed to find the derivative sell price that makes the expected

return on risk capital over the option tenor identical to a delta-one long position

• In bound 2 the residual hedging errors over each hedging interval is analyzed separately

to assess an addition to the average hedging cost that ensures the return on risk capital

over every hedge interval is equal to a delta-one long position over the corresponding

hedge-interval

In the next example the pricing bad deal bound 1 falls below the bad deal bound 2.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 44

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Example 3. Sell Put & Monthly Delta Hedge Stylized Asset 2

hedge average hedging cost 1 year 99.9% equivalent quintile of sum Sum of 1 year 99.9% equivalent quintile

T of hedging errors based of hedging errors based

notional

(months) (bound 1) (bound 2)

(% spot)

Black-Scholes multiple of avg Black-Scholes multiple of avg Black-Scholes

% spot % spot % spot

implied vol (%) hedge cost implied vol (%) hedge cost implied vol (%)

6 -1.17 0.02366 11.06 0.42587 18.00 18.33 0.52214 22.07 19.30

9 -2.19 0.05499 10.62 0.68339 12.43 17.62 0.80604 14.66 18.46

12 -3.07 0.09006 10.37 0.90596 10.06 17.17 1.05451 11.71 17.98

18 -4.21 0.15499 10.06 1.20435 7.77 16.38 1.36099 8.78 17.03

24 -4.86 0.20996 9.91 1.32973 6.33 15.57 1.62705 7.75 16.63

Black Scholes implied volatility

18%

16%

14%

12%

10%

8%

6%

3 6 9 12 15 18 21 24

tenor (months)

Table & Figure II-4. OHMC results for a sell 85% strike put with monthly delta hedging (Example

3). OHMC evaluates the P&L variance optimal hedge ratio and the residual hedge errors. The bad deal

bounds for the seller is the put sell price below which the put sellers expected return on risk capital (at 1

yr 99.9% equivalent confidence level) is below that of a simple long position in the asset. The risk capital

is assessed based on the hedging errors that are residual in the variance optimal hedging strategy. The

first bad deal bound assesses risk-capital based on the sum of LIBOR discounted hedging errors over the

deal life. The second bad deal bound is based on assessing risk capital over the hedging interval and

adding them to the upfront price (discounted at LIBOR).

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 45

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

To illustrate the impact of the volatility clustering time-scale we show results with a smaller

volatility clustering time. We keep all the parameters identical, but set the volatility clustering

time to one-and-a-half months: Γ (24) = 1.5 instead of 3 months in the base case. The starting

volatility is the same as the previous case- i.e., set to the long-term volatility. The effect of

lowering the volatility clustering time is that the volatility bounces around over shorter time

intervals. As a result, the hedging errors are expected to become larger over shorter tenors as the

starting volatility is forgotten and the full gamut of volatilities are realized. Due to the shorter

memory of the volatility clusters we also anticipate that the temporal averaging of volatility will

occur more effectively over a given tenor compared to the case of more persistent volatility.

hedge average hedging cost 1 year 99.9% equivalent quintile of sum Sum of 1 year 99.9% equivalent

T

notional (% of hedging errors based quintile of hedging errors based

(months) (bound 1) (bound 2)

spot)

Black-Scholes multiple of avg Black-Scholes multiple of avg Black-Scholes

% spot % spot % spot

implied vol (%) hedge cost implied vol (%) hedge cost implied vol (%)

6 -1.29 0.02573 11.18 0.503 19.57 19.12 0.565 21.95 19.69

9 -2.05 0.04944 10.46 0.695 14.06 17.71 0.815 16.48 18.52

12 -2.70 0.07653 10.10 0.846 11.06 16.83 1.016 13.27 17.77

18 -3.48 0.12522 9.69 1.007 8.04 15.50 1.266 10.11 16.63

24 -3.86 0.16498 9.47 1.053 6.38 14.50 1.492 9.04 16.16

Black Scholes implied volatility

18%

16%

14%

12%

10%

8%

6%

3 6 9 12 15 18 21 24

tenor (months)

Table & Figure II-5. OHMC results for a sell 85% strike put with monthly delta hedging (Example

3). The parameters are identical to the previous example – other than the volatility clustering time is

shorter (1.5 months instead of 3 months). As a result, over smaller tenors the impact of volatility

fluctuations is felt, yielding a higher value of the put. The temporal aggregation of the volatility also

occurs more rapidly, giving rise to a sharper decay of implied volatility versus tenor.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 46

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

Appendix-III

The hedge optimization problem posed here is infinite-dimensional insofar as we seek to find

how the hedge ratio and option value should depend on spot, so that the expected change in

wealth is zero and the hedge error measure is as small as possible. This calculus of variations

problem is rendered numerically tractable by using finite dimensional representations of the

hedge ratio and option value.

increasing sequence of M nodal values, xj, 0 ≤ j ≤ M − 1 , and basis functions are chosen to have

limited support around these nodal locations and provide the desired level of continuity at the

nodal locations. Piecewise cubic Hermite polynomials ensure continuity up to the first derivative

at the nodal locations. An extended node list is defined, with each nodal value repeating itself

once:

x j / 2 even j

xˆ 2 j = xˆ 2 j +1 = x j , 0 ≤ j ≤ M − 1 ⇔ xˆ j = , 0 ≤ j ≤ 2M − 1 (III-1)

x ( j −1) / 2 odd j

2 M −1

f ( x) = ∑ψ Ψ

j =0

j j ( x) (III-2)

− ( x − xˆ j − 2 ) 2 [2( xˆ j − x ) + ( xˆ j − xˆ j − 2 )]

Ψ j ( x ) = , xˆ j −2 ≤ x < xˆ j

( xˆ j − xˆ j −2 ) 3

+ ( x − xˆ j + 2 ) 2 [ 2( x − xˆ j ) + ( xˆ j + 2 − xˆ j )]

Ψ ( x ) = , xˆ j ≤ x < xˆ j + 2

Ψ j ( x) = j ( xˆ j + 2 − xˆ j ) 3 (III-3)

0, x < xˆ j − 2 , x > xˆ j + 2

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 47

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

− ( x − xˆ j −2 ) 2 ( x − xˆ j )

j

Ψ ( x ) = , xˆ j −2 ≤ x < xˆ j

( xˆ j − xˆ j −2 ) 2

2

Ψ + ( x ) = ( x − xˆ j + 2 ) ( x − xˆ j ) , xˆ ≤ x < xˆ

j +2

Ψ j ( x) = j ( xˆ j + 2 − xˆ j ) 2

j

(III-4)

0, x < xˆ j −2 , x > xˆ j + 2

Since x j* < x ≤ x j*+1 ⇔ xˆ 2 j* < x ≤ xˆ 2 j*+3 , at most 4 terms directly contribute to approximating

f(x)

2 j*+3

f ( x) = ∑ψ Ψ

j = 2 j*

j j ( x) =

(III-5)

+ + − −

ψ 2 j*Ψ 2 j* ( x) + ψ 2 j*+1Ψ 2 j*+1 ( x) + ψ 2 j*+2Ψ 2 j*+2 ( x) + ψ 2 j*+3Ψ 2 j*+3 ( x)

Let us express (III-5) in a more compact way. If j • is the smallest (even) nodal value such that

xˆ j • ≤ x < xˆ j •+2 , then

For the optimal hedging application in this paper, a discrete (M in number) set of values of the

reference asset, at time-step k, will be chosen as nodal locations for the finite dimensional

representation for the option value and the hedging parameter.

The basis functions (III-3)-(III-4) are defined such that the values of the parameters multiplying

them provide directly provide the values of f and df/dx:

f ( xˆ j ) even j

ψj = (III-7)

df ( xˆ j ) odd j

dx

This property is useful to examine departures from delta hedging because the derivative of the

option value with respect to the asset price is directly accessible.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 48

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

1 0.4

Hermite cubic basis functions

0.35

0.8

0.3

hermite

0.6 cubic 0.25

approx.

0.4 0.2

f(x) 0.15

0.2

0.1

0 0.05

-0.2 0

-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3

x x

Figure III-1. Piecewise Hermite cubic polynomial representation of f(x) = x2exp[-x2] over [-3,3]

using 7 nodes . The 14 basis function and numerical valuations of the piecewise Hermite

representation and direct evaluations of the function are shown in the plots.

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

Optimal Dynamic Hedging of Cliquets 49

A. Petrelli, J. Zhang, O. Siu, R. Chatterjee, V. Kapoor*

References

Bollersev, T., Generalized Autoregressive Conditional Heteroskedasticity, Journal of Econometrics 31,

307-327, 1986

Bouchaud, J-P, M. Potters, Theory of Financial Risk and Derivative Pricing, From Statistical Physics to Risk

Management, Cambridge Univeristy Press, Cambridge 2003.

Carnero, M. A., D. Pena, E. Ruiz, Persistence and Kurtosis in GARCH and Stochastic Volatility Models,

Journal of Financial Econometrics, Vol. 2, No. 2, pp319-342, 2004

Gatheral, J., The Volatility Surface: A Practitioner’s Guide, John Wiley and Sons, New Jersey, 2006.

Jeffery, C., Reverse cliquets: end of the road?, Risk Magazine, 17, (2004).

Kapoor, V., L. Cheung, C. Howley, Equity securitization; Risk & Value, Special Report, Structured

Finance, Standard & Poor’s, (2003)

Laurent, J.-P., H. Pham, Dynamic Programming and mean-variance hedging, Finance and Stochastics, 3, 83

(1999)

Laurent, J.-P., A. Cousin, J-D Fermanian, Hedging Default Risks of CDOs in Markovian Contagion Models,

DefaultRisk.Com (2008)

Petrelli., A., O. Siu, J. Zhang, V. Kapoor, Optimal Static Hedging of Defaults in CDOs, DefaultRisk.com,

April (2006)

Petrelli, A., J. Zhang, S. Federico, V. Kapoor, Chapter 3: Four Synthetic CDO Trading Strategies, in Credit

Derivative Strategies: New Thinking on Managing Risk & Return, edited by R. Douglas,

Bloomberg Press (2007)

Pham, H., Dynamic Programming and mean-variance hedging, Finance and Stochastics, 3, 83 (1999)

Potters, M., J.-P. Bouchaud, D. Sestovic, Hedged Monte Carlo: low variance derivative pricing with objective

probabilities, Physica A, 289, 517, (2001a)

Potters, M., J.-P. Bouchaud, D. Sestovic, Hedge your Monte Carlo, Risk Magazine, 14, 133 (2001b)

Schweizer, M., Variance-Optimal hedging in discrete time, The Mathematics of Operations Research, 20, 1-

32 (1995)

Wilmott, P., Cliquet Options and Volatility Models, Wilmot Magazine, December 78-83, (2002)

Wilmott, P., Paul Wilmott on Quantitative Finance, 2nd Edition, vol. 3, Chapter 49, Wiley (2006)

Zhang, J., O. Siu, A. Petrelli, V. Kapoor, Optimal Dynamic Hedging of CDS Swaptions, working paper

(2006)

*

This article reflects the opinions and views of the authors and not that of their employers, and no representation as to the accuracy

or completeness of the information is provided.

## Гораздо больше, чем просто документы.

Откройте для себя все, что может предложить Scribd, включая книги и аудиокниги от крупных издательств.

Отменить можно в любой момент.