Citi Cdo G

Synthetic CDO Risk-Return Dynamics
Collaborators:
Vivek Kapoor
vivek2.kapoor@citi.com
Andrea Petrelli Credit Suisse

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
CDX.NA.IG.4 6/20/2010 49bps

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
Quote on CDX.NA.IG.4 for 3/31/2005
CDX.NA.IG.4 6/20/2010 78 bps

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
Quote on CDX.NA.IG.4 for 5/19/2005

1
SYNTHETIC CDO TRADING - WHY?
Chase opportunity sloshing across the capital structure
Standard CDO model does not directly address cost of hedging

Replication of a CDO tranche by CDS is difficult*:
• Jumps in spread and jumps to default
• Uncertain recovery
• Random realized spread-spread correlation
Evade credit spread delta radar and make carry#

Tradeoff systematic and idiosyncratic spread convexity
Hunger for yield
*theoretically and practically

#convexity tradeoffs not well captured in risk systems
2
LONG CREDIT INDEX POSITION
An elementary portfolio credit trade of

selling protection on a pool of CDS
The premium cash-flow stream is

denoted by the solid line, and the
trading book CDS index
default contingent cash flows are
denoted by the dotted line
Synthetic CDO trades are compared

with this elementary long credit
position
3
SPREAD DELTA-NEUTRAL SELL EQUITY PROTECTION
A popular CDO trade consists of

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)
Net 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
single name Value on Default

-1% Delta Hedged Senior Mez Tranche
The horizontal axis is the
(% tranche notional)
(7-10%), Carry = 116 bps/yr (left) credit spread level of the
(% 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
P&L (% of tranche notional)

Buy CDS Protection 75 Sell CDS Protection
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)
Sell Equity Protection (0-3%)

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
(100) The maximum loss is proportional

1 3 5 7 9 11 13 15 17 19
number of defaults to initial trade carry

(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%)
Delta Hedged Senior Mez Tranche (7-10%)

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)
at net delta (zero!) and net

carry (largest for the
straddle) Default risk per unit carry
(CDX. NA.IG.4, March 31, 2005)

10
ONE YEAR DEFAULT RISK PER UNIT CARRY
0
3/31/05 (a) LONG INDEX TRADE

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)
Carry per unit default

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

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)
spread shift (bps)

250
1 name 2 names
These positive carry trades have
200 5 names 10 names
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
systematic convexity at spreads

spread shift (bps)
higher than spot
12
(CDX. NA.IG.4, March 31, 2005)
TRANCHE PRICING CORRELATION
30
25
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)
mtm (% of tranche notional)

35%
-30%
34% -35%
-40%
33%
(a) cash -45%
-50%
32%
-55%
31% -60% (b) mtm

-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%
-20% Cash, mtm, & total P&L of

-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
No Hedging Static Delta Hedging

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%
-5% 140
-10% 180%
-10% 120
-15% 160%
-15% 100
-20% 140%
-20% 80 -25% 120%
-25% 60 -30% 100%
-30% 40 -35% normalized cross-sectional spread dispersion (right) 80%

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%
implied equity correlation

0%
-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
No Delta Hedging Static Delta Hedging
Delta Hedging Daily Delta Hedging Every 2 Months

15
REALIZED SPREAD MOVE CORRELATION
40%
35%
realized correlation
(a) Realized correlation versus
30%
time interval (CDX.NA.IG.4)
25%
20% Data: 3/22/05-11/28/05

The standard model 15%
Data: 3/22/05-3/2/07
has not addressed 10%

1 day 2 weeks 2 months 4 months
the cost of hedging time interval
5%
spread dispersion 0%
total P&L (% of tranche notional)

-5%
-10%
-15%
(b) Impact of spread delta- -20%
hedging interval on sell equity -25%
-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
No Hedging Static Delta Hedging

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%
-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
equity tranche implied correlation

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 implied correlation (%)

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
May 05 Sept-Oct 05 junior tranche protection seller discovers his

idiosyncratic negative credit convexity
event event
19
1000
CDX.IG.9 Maturity: 2012
100
10
index (bps/yr) 0_3(% upfront + 500 bps/yr)

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
CDX.IG.9 2012
index 0_3(% upfront +

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)
9/20/2007 52 36.5 139 48.5 26.5 16.75 8

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
Systematic negative convexity to credit - for

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
Opportunity sloshes across the capital structure!

24
OMISSION OF “STANDARD MODEL”
Need to directly address hedging costs and residual risks to discern
relative value and express risk-preference
Beware of derivative “valuation” models that do not address

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
Need to establish the cost of hedging and also provide description

of unhedgeable risks!
25
EYES OPEN ASSESSMENT OF VALUE & RISKS
How to assess hedging cost and hedging error en-

route to valuation while accounting for spread
diffusion, spread jumps, defaults with dependent
uncertain recovery?
Start with objective measure model of credit spreads & default

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
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.
MULTIDIMENSIONAL CONSTRAINED MINIMIZATION PROBLEM
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.
MULTIDIMENSIONAL VARIATIONAL CALCULUS PROBLEM
29
OPTIMAL STATIC HEDGING RESULTS
80
hedge error (% tranche)

Sample Problem: Setup 70
ESF0.95
60
5 yr transaction, 125 issuers of par spread 78 bps 50
& recovery fixed to 30% 40
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)
Reduced form Normal Copula (PNC: solid line)
break-even upfront (% tranche)

100
90
80
Structural Variance Gamma (SVG: dashed line) 70
60
Results for equity tranche with 500 bps running 50

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
no-default carry (% tranche/ yr)

12
8
Description of un-hedged risks 4
-4
-8
-12
0 10 20 30 40 50
total hedge notional (x tranche)
30
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
break-even upfront (% tranche)

100
90
As the hedge notional is increased from zero the 80
70
P&L uncertainty is reduced – it achieves its minima 60
– and increases after that 50

40
30
20
10
0 10 20 30 40 50
The no-default carry is significantly positive at the total hedge notional (x tranche)
hedge notional that minimizes the variance
no-default carry (% tranche/ yr)

12
8
Minimizing the loss tail risk measure requires 4
hedge notionals that eliminate no-default carry and 0
can result in negative no-default carry -4
-8
-12
0 10 20 30 40 50
31
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)
Sell equity (0%-3%) tranche protection wealth change distribution (SVG

model, bin size = 1% tranche notional, zero mean constraint) 32
Sample Problem: Optimal Hedging Given Tranche Pricing
• Given the equity tranche upfront what is the optimal hedge ratio?:
W mean change in wealth : W

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
0.5 mean/std dev 0.0001
0
0 5 10 15 20 25 0.00001
-50 -25 0 25 50 75 100
wealth change (% tranche)
Sell equity (0%-3%) tranche protection for 60% upfront & 500 bps running
(solid line: PNC; dashed line: SVG)
33
Sample problem: Optimal Hedging Implied Correlation Skew
50%
45% Variance optimal PNC
ESF0.8 optimal PNC
40%
Variance optimal SVG
implied correlation
35% ESF0.8 optimal SVG
30%
25%
20%
15%
10%
5%
0%
equity mezzanine senior
Optimal static default hedge break-even price implied correlation for

sample problem (0%-3% equity; 3%-7% mezzanine; 7%-10% senior)
34
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
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
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
Historical daily GM 5 yr CDS par spread

37
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
We seek to find C ( sk | t d t k ) and ( sk | t d tk )

2
such that W tk , tk 1 0 and E[ W W ] is minimized
38
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
If perfect replication is possible the results of optimal hedging analysis under

sufficiently small time-intervals agree with the risk-neutral framework
When perfect replication is fundamentally unattainable, the optimal hedging

framework will delineate costs and errors of the hedging strategy and help express a
risk preference
39
1
probability density function ($)
probability density function

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
total change in wealth ($) total change in wealth ($)
(a) =0 (b) = 0.3

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
2.2 1
pdf

2
pdf(conditioned on no-default)
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
Probability density of wealth change of variance optimal hedged ATM no-

knockout payer swaption under 1-day hedging interval
41
400 90
spread 800 90
380
hedge notional ($)

85
hedge notional ($)

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
time (yrs) time (yr)

2 2.75
40
option & hedge P&L ($)
daily P&L ($) (total)
option & hedge P&L ($)

50 2.5
0
daily P&L ($) (total)

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)
Defaulted sample path behavior of variance optimal hedged ATM no-

knockout payer swaption under 1 day hedging interval with = 0.3 42
Sample Problem: ATM No-Knockout & Knockout Payer Swaption
No-Knockout Knockout

1 1
pdf pdf
pdf(conditioned on no-default) pdf (conditioned on no-default)

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
E W no default 0 E W no default 0
Probability density of wealth change of variance optimal hedged ATM no-

knockout and knockout payer swaption under 1-day hedging interval with =
0.3
Carry chases its own tail!

43
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
The plethora of irreducible hedging errors makes CDOs far from a

redundant derivative!
As the standard model has not addressed any aspect of replication of a

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.
Trading opportunities arise around the ever evolving partitioning of value

across the capital structure – market agents experience in real time the
challenges of hedging and tend to adjust valuations in response.
45
SUMMARY
Extensive scenario analysis (ad-hoc and with realistic temporal

structure) is needed to understand risk-return tradeoffs.
Optimal hedging analysis formalizes “extensive scenario analysis” as a

precursor to finding value by addressing the hedge performance over the
ensemble of credit outcomes and minimizing a statistical hedge error
measure.
Coupled spread-default stochastic models are needed for scenario

analysis as well as optimal hedging analysis.
Spread volatility and covariance in addition to default dependence need

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
Petrelli, A., O. Siu, J. Zhang, V. Kapoor, Optimal Static Hedging of Defaults

in CDOs, DefaultRisk.com, April 2006
Petrelli, A., J. Zhang, N. Jobst, V. Kapoor, Practical Guide to CDO Risk

Management, The Handbook of Structured Finance, by Arnaud
de Servigny and Norbert Jobst, McGraw-Hill 2007
Zhang, J., O. Siu, A. Petrelli, V. Kapoor, Optimal Dynamic Hedging of CDS

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.
Keywords: CDO, Hedging, Default, Carry, Expected Shortfall
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.
II. Change in Wealth of a CDO Trader

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:
k1* (t ) = min[max[k1 , cpL(t )], k2 ] ; k2* (t ) = max[min[k2 , N − cpR(t )], k1 ] (3)
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.
Portfolio of Reference Pool Bonds

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 = ΔWbond + ΔWtranche (7)
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.
Hedge Error Measures

We employ two hedge error measures:
Standard Deviation ( (
Θ = σ ΔW ≡ E[ ΔW − ΔW ] ) )
2 1/ 2
(10)
Expected Shortfall Θ = ESFα ≡ − E[ΔW | ΔW ≤ −VaRα ] (11)
Prob{ΔW < −VaRα } = 1 − α (12)
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:
ΔWno−default = ΔWbond ,no−default + ΔWtranche,no−default (13)
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.
Reduced Form Poisson-Normal-Copula (PNC) Model of Default Time

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.
Structural Variance Gamma (SVG) Model of Default Time

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.
IV. Sample Results
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
Reduced form Poisson-Normal Copula (PNC) (Li, 1999)

Hazard rate λi = 0.65 %/yr ∀ i
Asset correlation 25%
Structural Variance Gamma (SVG) structural (see Appendix-B)

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.
Hedge Notional, Hedging Error, Break-Even Pricing & Carry
Sell Equity Tranche Protection

80 100
break-even upfront (%tranche)

90
70
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
0.35 (0 x hedge) (0 x hedge)

(12 x hedge) (12 x hedge)
0.3 0.1
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
(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
break-even spread (bps/yr)

50
ESF 0.8 ESF 0.95 800

40
700
std dev
30 600
500
20
400
10
300
0 200
0 5 10 15 20 0 5 10 15 20
(a) (b)
3
no-default carry (% tranche/yr)
2
1
0
-1
-2
-3
-4
-5
-6
0 5 10 15 20
(c)
Figure 3. Sell mezzanine (3%-7%) tranche protection (solid-lines: PNC; dashed-lines: SVG)
0.8 1
(0 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.
Sell Senior Tranche Protection
40 600
35
break-even spread ((bps/yr)
500
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
(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
(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)
0.8 (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
(a) (b)
Figure 6. Sell senior (7%-10%) tranche protection wealth change distribution (SVG model,
bin size = 1% tranche notional)
Optimal Static Hedge Breakeven Price & Correlation Skew

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
ESF0.8 optimal SVG
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.
Optimal Hedging Given Tranche Pricing

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
0.5 mean/std dev
0
0 5 10 15 20 25
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
(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
0.1
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
(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)
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
(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
Optimal Static Hedging of Default Risks of a Bond by CDS

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
Probability[τ < time to default (t d ) ≤ τ + dτ ] = λe −λτ dτ (A1)
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
Figure A1. Portfolio of a defaultable bond and a credit default swap
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:
dσ Δ2W (n − pu )ϖ 1 + puϖ 2 + nuϖ 3

= 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.
Par Bond Case (pu = nu)

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
The residual variance corresponding to the parameters (A7) is zero:
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.
Uncertain Recovery and Non Par Bond

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
Figure A2. Variance optimal hedge solution: known recovery case

T = 5 yr, c = 10%/yr, r = 5%/yr, R = 0.5 , σ R / R = 0
24
attribute symbol value
Bond notional nu $100
Bond price pu $90
Bond coupon c 10%/yr
Interest rate r 5%/yr
Bond maturity T 5 years
Default hazard rate λ 5 (%/yr)
Recovery average R 0.5
case 1: 0%
σR
Recovery std dev case 2: 50%
(% of R )
case 3: 75%
Table A1. Sample problem setup
item case1 case2 case3
Optimal CDS hedge notional $87.9 $90.4 $92.3
CDS break-even spread 823 bps/yr 808 bps/yr 796 bps/yr
Residual hedge error std. dev

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
Figure A3. Variance optimal hedge solution: uncertain recovery case

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
sr* (1/yr) 0.082

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
Figure A4. Variance optimal hedge solution: uncertain recovery case

T = 5 yr, pu/nu = 0.9, c = 10%/yr, r = 5%/yr, R = 0.5
27
attribute symbol value
Bond notional nu $100
Bond price pu $98
Bond coupon c 5.5%/yr
Interest rate r 5%/yr
Bond maturity T 5 years
Default hazard rate λ 0.5 %/yr
Recovery average R 0.5
case1 0%
σR
Recovery std dev case 2 50%
(% of R )
case 3 75%
Table A3. Sample problem setup
item case1 case2 case3
Optimal CDS hedge notional $97.7 $98.2 $98.55
CDS break-even spread 0.974%/yr 0.971%/yr 0.968%/yr
Residual hedge error std. dev

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
CDS hedge notional (% bond notional)
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
Multi-Asset Variance Gamma Structural Default Model

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
g i (Δt ;1,ν ) = g market (Δt ; κ ,νκ ) + ui (Δt ;1 − κ ,ν (1 − κ )) (B3)
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:
μi = 0.0 (1/yr); σ i = 0.20 (1/yr1/2 );
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
different colors using different symbols for the different issuers.
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.
Joshi, M., A. Stacey, Intensity Gamma, Risk, July 2006.
Luenberger, D. G., Investment Science, Oxford University Press, 1998.
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
Optimal Dynamic Hedging of Cliquets

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.
OHMC Versus “Risk-Neutral” Expectations

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.
*
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).
• OHMC is applicable to general stochastic descriptions of the underlying. Risk-neutral

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.
*
100%
90%
hedge slip stdev/avg hedge cost

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.
105 100 option hedge total

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.
*
Reality Versus Perfect Replication & Unique Price

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.
*
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.
*
Characteristics Risk-Neutral Approach OHMC
i. Seeks to minimize hedging error

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)
i. Valuation model based on risk-

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)
*
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.
*
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.
*
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.
Cliquet Option Sellers P&L

The terminology and symbols needed to specify a Cliquet contract and a protection seller’s P&L
are enumerated here.
Contract trigger condition/look-back dates: {tˆ0 = 0, tˆ1 , tˆ2 ,......, tˆN −1 = T }
Hedging interval:
(tk , tk +1 ]
Running premium rate: η
Cliquet value: C(t)
s(tî )
Payout trigger time: tî* = min tî ∋ < K1
s(tî − nlb )
i ∈ {nlb , nlb + 1,......., N − 1}
nlb is the number of observation intervals in a look-back period
Reference notional: Ψ
  s(tî* ) 
Payout amount: P(tî* ) =Ψ  K1 − max , K 2 
  s(tî*− nlb ) 
1 payout triggers at tî* ∈ (tk , tk +1 ] 

Trigger indicator: I (tk , tk +1 ] =  
0 no payout trigger over (tk , tk +1 ] 
*
Risk-free discount factor: df(τ, t)
Premium accrual time:

t*
 tî* 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:
ω (tk , tk +1 ] = I (tk , tk +1 ]P (tî* )df (tk , tî* )
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.
Hedging the Sell Cliquet Protection Position
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
*
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.
3. Optimal Hedge Monte-Carlo Formulation for Cliquet

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.
∆Wt (t k , t k +1 ) = ∆Wt cliquet (t k , t k +1 ) + ∆Wt hedge (t k , t k +1 ) =

k k k
C k ( s k ) + ηΨχ (t k , t k +1 ) − ω (t k , t k +1 ] − (1 − I (t k , t k +1 ])C k +1 ( s k +1 )df (t k , t k +1 )

(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
∆Wt (t k , t k +1 ) = Ck (sk ) − Gk + Φ k (sk )H k

k
Gk = (1 − I (tk , tk +1 ])Ck +1 (sk +1 )df (tk , tk +1 ) + ω(tk , tk +1 ] − ηΨχ (tk , tk +1 )

(4)
H k = (sk +1 − sk / DF(tk , tk +1 ))df (tk , tk +1 )
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:
*
Find Ck(sk) and Фk(sk) so that
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.
Residual Risks While Attempting to Replicate

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.
The OHMC hedging time-grid is specified as
{t0 = 0, t1 , t2 ,.....,tk , tk +1 ,...., t K −1 = T }

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
` `
The total change in wealth discounted to t0 is then given by

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
*
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.
The issue of residual risks in an attempted replication strategy is operationally often

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
*
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.
survival probability over period τ ps(τ)
− ln ( p s (τ ))
τ interval hazard rate λτ =
τ
τ interval hazard rate based h interval survival probability ps(h) = exp[-λτ h]
∆Wt (t , t + h)
h interval expected P&L
*
h interval ps(h) quintile wealth change

q(t; ps (h)) ∋ Probability{∆Wt (t , t + h) < q(t; ps (h))} = ps (h)
h interval ps(h) quintile deviation from average wealth change

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) )
Delta-One Bad Deal Bounds on Derivative Seller-Hedger
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:
*
∆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:
*
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.
4. Reference Asset Description

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.
Under GARCH(1,1) the asset and its volatility evolve as follows:
(
∆s k = s k µ∆t + σ k ∆t ε k ) (10)
σ k2 = (1 − α − β )σ 2 + σ k2−1 (β + αε k2−1 ) (11)
*
Standard Normal random-variates generated to create the return stochastic process in (10) are
denoted by ε k , and the volatility evolves per (11).
Method of Moments Fitting to Empirical Returns

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]):
Mean and Variance
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 ) 
The empirical kurtosis is employed to express α as a function of β :
(κ − 3)(κ (3 − 2β 2 ) − 3) − β (κ − 3)
α= (15)
3(κ − 1)
*
Volatility Clustering Time
[
 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.
Three Stylized Assets

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.
*
Stylized Asset 1
Daily Statistics r = 0.000793651; σr = 0.025197632; Γ (252 ) = 10 ; κ = 20
Fitted Parameters µ = 0.20 (1/yr); σ = 0.40 (1/yr0.5); α = 0.197468; β = 0.755553
Sensitivity Shown * (a) κ = 30 ; α = 0.206672; β = 0.744686
(b) κ = 3 ; α = 0; β = 0 (i.e., Geometric Brownian Motion)
Stylized Asset 2
Monthly Statistics r = 0.008333; σr = 0.023094; Γ (24) = 3 ; κ = 8
Fitted Parameters µ = 0.10 (1/yr); σ = 0.08 (1/yr0.5); α = 0.354366; β = 0.419041
Sensitivity Shown* (a) Γ (24) = 1.5 ; α = 0.48795; β = 0; κ = 3
Stylized Asset 3
Daily Statistics r = 0.000476; σr = 0.010079; Γ (252 ) = 15 ; κ = 15
Parameters µ = 0.12 (1/yr); σ = 0.16 (1/yr0.5); α = 0.146813; β = 0.825871
Sensitivity Shown * (a) κ = 3 ; α = 0; β = 0 (i.e., Geometric Brownian Motion)
Table 2. Stylized asset descriptions employed to illustrate Cliquet sensitivities.

*Sensitivities to GARCH(1,1) parameters are shown in the next section.
*
squared return autocorrelation squared return autocorrelation squared return autocorrelation

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%
0 21 42 63 84 105 126 0 3 6 9 12 15 18 21 24 0 21 42 63 84 105 126

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
100% 60% 60%

delta-one long position delta-one long positiion delta-one long position
risk-capital (1 year 99.9%)

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)
25% 70% 20% delta-one long position

delta-one long position delta-one long position
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)
(a) Stylized Asset 1 (b) Stylized Asset 2 (c) Stylized Asset 3
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.
*
5. Expected Cost of Hedging & Residual Risks for a Cliquet Put

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.
*
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
bps/yr 66 478 778 3810 468 0.04676

hedging cost
87.3 17.0 4.2 1.6 15.4 9.5
Table 3. Cliquet put contracts on Stylized Asset 1
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.
*
Figure 4 (a), (b), (c)

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 ($)
amount increases from 1% to 7% over the

-5
-6 days preceding the look-back, a loss of $15
dw-cliquet total P&L ($) is experienced on the day payout is triggered

-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
pdf
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.
*
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
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
hedging cost
12.6 29.9 21.1 17.0 28.0 22.9 29.5 24.5 22.4 19.0
Table 4. Cliquet contracts on Stylized Asset 2
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.
*
120 asset-2: case 3 0% Figure 5 (a), (b), (c)

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).
Note that the nature of the Cliquet is

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.
Merely focusing on tail losses may limit the

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)
*
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
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
170.0 21.8 5.7 5.1 4.9 4.7 4.6
hedging cost
Table 5. Cliquet contracts on Stylized Asset 3
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.
*
Figure 6 (a), (b), (c)

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 ($)
0 2 residual risks.
total daily P&L ($)

-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)
*
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.
Comparing cases 2, 5, 6 in Table 3 on the stylized asset 1 documents the impact of

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.
Volatility Clustering Time
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.
*
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
*
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.
*
OHMC Approach & Exotics

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.
In this evolution of modeling of derivative contracts, limits on replication are a major

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
*
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.
*
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
G13, G11, D81
G13 Contingent pricing; Futures Pricing
G11 Portfolio Choice; Investment Decisions
D81 Criteria for Decision-Making under Risk and Uncertainty
*
Appendix-I
Cliquet Hedging Variational Problem

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)
Gk = (1 − I (tk , tk +1 ])Ck +1 (sk +1 )df (tk , tk +1 ) + ω(tk , tk +1 ] − ηΨχ (tk , tk +1 )

(I-2)
H k = (sk +1 − sk / DF(tk , tk +1 ))df (tk , tk +1 )
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  
*
To solve for the unknown coefficients a kj and b kj we employ a Lagrange multiplier technique:
Fk = E[∆Wtk (tk , tk +1 ) ] + 2γE[∆Wtk (tk , tk +1 )]

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
*
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
Initial Time Step
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.
*
Consistency Between Vanilla and Exotic Options

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.
*
Appendix-II
Optimal Hedge Analysis of Vanilla Options

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:
∆Wt (t k , t k +1 ) = ∆Wt option (t k , t k +1 ) + ∆Wt hedge (t k , t k +1 ) =

k k k
C k ( s k ) − C k +1 ( s k +1 ) df (t k , t k +1 ) + Φ k (s k )(s k +1 − s k / DF (t k , t k +1 ))df (t k , t k +1 ) (II-1)
Using the symbols that are convenient in the Cliquet OHMC algorithm, we rewrite (II-1) as
∆Wt (t k , t k +1 ) = Ck (sk ) − Gk + Φ k (sk )H k

k
Gk = Ck +1 (sk +1 )df (tk , tk +1 ); H k = (sk +1 − sk / DF(tk , tk +1 ))df (tk , tk +1 ) (II-2)
*
The OHMC solution is propagated from maturity to the option starting point, where it becomes
an ordinary minimization problem, as formulated in Appendix-I.
I. Hedging Error Dependence on Hedging Interval & Return Kurtosis

Example 1. Sell Put and Delta Hedge Stylized Asset 3
Underlying characteristics: µ = 0.12 (1/yr); σ = 0.16 (1/yr0.5); Γ (252 ) = 15 ; κ = 15
Put parameters: spot = 100; strike = $100; r = 4%/yr; maturity = 1 month
The disparity in hedge slippage error between a realistically fat-tailed distribution and an
idealized return distribution with no kurtosis is visible in Figure 1 (main section 1). Kurtosis >
3 renders the hedging error irreducible as the hedging frequency increases. For Geometric
Brownian Motion the hedging error decreases much more rapidly with hedging frequency, and is
headed to zero in the limit of continuous hedging. This result shows how much of the
mathematical machinery to deal with continuous hedging while ignoring realistic return fat tails
is of marginal practical importance because in the limit of continuous hedging the transaction
costs would become unboundedly large, and in practically realistic returns, after a point, hedging
more often does not reduce the hedging errors. The OHMC framework handles discrete hedging
without having to assume zero excess kurtosis for the underlying.
1 1
kurtosis=15 kurtosis = 15
0.1 0.1
pdf
cdf
0.01 0.01
0.001 0.001
0.0001 0.0001
-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2
total change in wealth of put seller-optimal hedger ($) 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).
*
II. Strike-Kurtosis Sensitivity

Example 2. Sell Put and Daily Delta Hedge on Stylized Asset 1
Underlying characteristics: µ = 0.20 (1/yr); σ = 0.40 (1/yr0.5); Γ (252 ) = 10 ; κ = 20
Put parameters: spot = $100; strike = K; r = 4%/yr; maturity = 1 month
delta-one risk-return based bad deal bound for option-seller

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 (%)
80 -4.6 0.185 44.7% 0.607 3.29 58.0% 0.513 2.78 55.6%

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.
*
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)
80 -4.70 0.188 44.9% 0.625 3.33 58.4% 0.523 2.78 55.9%

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.
*
III. Tenor & Volatility Clustering Time Sensitivity

Example 3. Sell Put & Monthly Delta Hedge Stylized Asset 2
Underlying characteristics: µ = 0.10 (1/yr); σ = 0.08 (1/yr0.5); Γ (24) = 3 ; κ = 8
Put parameters: spot = $100; strike = 85; r = 4%/yr; maturity = T
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)
3 -0.15 0.00223 11.60 0.09782 43.91 18.35 0.16491 74.03 20.18

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
22% avg hedge cost bad deal bound1

Black Scholes implied volatility
20% bad deal bound2 hist vol
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).
*
Volatility Clustering Time Sensitivity

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)
3 -0.27 0.00412 12.27 0.167 40.53 20.23 0.221 53.54 21.39

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
22% avg hedge cost bad deal bound1

Black Scholes implied volatility
20% bad deal bound2 hist vol
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.
*
Appendix-III
Piecewise Hermite Cubic Basis Functions

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.
Consider a finite dimensional representation of a function f(x). The x space is discretized by an

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
The finite dimensional representation of f(x) is made as
2 M −1
f ( x) = ∑ψ Ψ
j =0
j j ( x) (III-2)
where for even j
 − ( 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



*
and for odd j
 − ( 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
f ( x) = ψ j •Ψ j+• ( x) + ψ j •+1Ψ j+•+1 ( x) + ψ j•+2Ψ j−•+2 ( x) + ψ j •+3Ψ j−•+3 ( x) (III-6)
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.
*
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.
*
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
Engle, R. F., Arch Models, Oxford: Oxford University Press, 1994
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)
*

Citi Cdo G

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Citi Cdo G

Uploaded by

Copyright:

Available Formats

Synthetic CDO Risk-Return Dynamics

Andrea Petrelli Credit Suisse

CDX.NA.IG.4 6/20/2010 49bps

Quote on CDX.NA.IG.4 for 3/31/2005

CDX.NA.IG.4 6/20/2010 78 bps

Quote on CDX.NA.IG.4 for 5/19/2005

Standard CDO model does not directly address cost of hedging

Evade credit spread delta radar and make carry#

*theoretically and practically

An elementary portfolio credit trade of

The premium cash-flow stream is

Synthetic CDO trades are compared

A popular CDO trade consists of

Net P&L (% of tranche notional)

single name Value on Default

(7-10%), Carry = 116 bps/yr (left) credit spread level of the

P&L (% of tranche notional)

Sell Equity Protection (0-3%)

(100) The maximum loss is proportional

number of defaults to initial trade carry

Delta Hedged Senior Mez Tranche (7-10%)

at net delta (zero!) and net

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

3/31/05 (a) LONG INDEX TRADE

Carry per unit default

P&L (% of tranche notional)

spread shift (bps)

200 5 names 10 names

systematic convexity at spreads

mtm (% of tranche notional)

31% -60% (b) mtm

-20% Cash, mtm, & total P&L of

No Hedging Static Delta Hedging

-30% 40 -35% normalized cross-sectional spread dispersion (right) 80%

implied equity correlation

No Delta Hedging Static Delta Hedging

Delta Hedging Daily Delta Hedging Every 2 Months

20% Data: 3/22/05-11/28/05

has not addressed 10%

total P&L (% of tranche notional)

(b) Impact of spread delta- -20%

hedging interval on sell equity -25%

No Hedging Static Delta Hedging

equity tranche implied correlation

equity implied correlation (%)

May 05 Sept-Oct 05 junior tranche protection seller discovers his

index (bps/yr) 0_3(% upfront + 500 bps/yr)

index 0_3(% upfront +

9/20/2007 52 36.5 139 48.5 26.5 16.75 8

Systematic negative convexity to credit - for

Opportunity sloshes across the capital structure!

Beware of derivative “valuation” models that do not address

Need to establish the cost of hedging and also provide description

How to assess hedging cost and hedging error en-

Start with objective measure model of credit spreads & default

MULTIDIMENSIONAL CONSTRAINED MINIMIZATION PROBLEM

MULTIDIMENSIONAL VARIATIONAL CALCULUS PROBLEM

hedge error (% tranche)

5 yr transaction, 125 issuers of par spread 78 bps 50

& recovery fixed to 30% 40

Reduced form Normal Copula (PNC: solid line)

break-even upfront (% tranche)