Slides Aymeric KALIFE Derivatives As Hedge Instruments-2022
Slides Aymeric KALIFE Derivatives As Hedge Instruments-2022
1
Map
◼ Hedging principles
Asset vs. Liability Management principles
From linear to non linear risks
The short convex liability issues in practice
In this example, we assume the guarantee liability variation is +0.2 for -1%
(Delta = -20%) in asset variation
Option Value = +2
After 10% drop
100 100 in asset
(fund (fund
units) units) 90 90
(fund (fund
units) units)
After a 10% drop, assets and liabilities are matched ➔ hedging assets is equal to the Option Value,
the Insurance company P&L = 0 (-8 for Asset vs. -8 for Liab)
3
From linear to non linear risks
◼ Dynamic hedging covers equity and interest rate risk but leaves an exposure to equity and interest
rate volatility
◼ Dynamic hedging protects AXA by providing funds to pay for Guaranteed Benefits whenmarkets
underperform
◼ Dynamic hedging involves owning and rebalancing short futures positions
◼ Futures position gain is relative to risk free rate -> The lower the rate, the lower our gain in
declining interest rate market.
4
From Hedging to the “Delta” Greek
f (S,t)
A Taylor Expansion:
f = ft t + f S S + 1 f (t) 2 + 1 f (S) 2 + f tS +...
2 tt 2 SS tS
5
Delta hedged call option:
with = −cS
ct + rScs + 12 2 S 2cSS = rc
Black-Scholes for the final time!
Black&Scholes formula by martingale theory
Proof
8
Black&Scholes formula by martingale theory
9
Black&Scholes formula by martingale theory
10
Appendix: from real-world to risk-neutral
Black Scholes formula proxy
ATM
Brenner and
Subrahmanyam
(1988)
12
Black Scholes formula proxy
OTM
13
Black Scholes formula proxy
OTM
14
Delta as hedge factor
◼ The Delta is also the hedge factor of the options
This represents what spot position we need to take to offset any
change in the option price as spot moves
◼ if we buy an option for which delta is +32%, then we need to
sell 32% of notional
◼ If spot changes, options with high Delta have a larger absolute value
change than options with low Delta
15
Delta hedge:
Current Price: S = 10, Risk Free Rate: r = 0.05
Delta Hedge
6
call
delta
4
2
Call Option Price
-2 Slope =
-4
Position in
-6
Bonds 0 5 10 15
Stock Price
16
Delta Hedging of a Call
Dynamic hedging is the updating of delta in hedging toward maturity.
Positive gamma is desirable on a long option position. Negative gamma
Stock call is not desirable on a short option position.
option price Call price at t before T
Underlying
0 Exercise Price X stock price St
17
How to compute a delta
-100,000
2 -250,000
Delta
global
in EUR
25,000
20,000
OV(x)
15,000
10,000
5,000
AV shock (in %)
-
-5% -4% -3% -2% -1% 0% 1% 2% 3% 4% 5%
❑ Delta splitting:
global
i (n%) = i global (n%)
i
i
1
18
0
Example – Mechanics of delta hedging
19
Effects of delta hedging
20
Effects of delta hedging
21
Effects of delta hedging
22
From Delta hedge to Delta-Gamma hedge
When you are deep out or deep in, Delta is flat and asymptotic as shown above. But when are
you not, a large move can result in significant trading loss despite being Delta hedged.
As long as prices move in small increments and do not jump dramatically, Delta hedge will
cover you. The underlying jumps, you are exposed.
Delta is the first order rate of change and works well within a narrow band. Within and outside
that band Gamma tracks not just the error but also the magnitude of your gain/loss in case of
a large move (up/down).
The magnitude of the error shifts dramatically as the option gets closer to the At/Near money
state. When options are deep in or deep out, similar to Delta, Gamma also flattens out.
23
Non linear risks
24
From Delta hedging to Convexity Hedging
For small moves in the underlying, the replicating hedge is accurate. For larger moves in the
underlying, the long options position is always outperforming the replicating hedge in both
directions. For a delta-hedged option, the gamma P&L will be the outperformance of the option
over the replicating hedge. The more the delta changes, the more the replicating delta-hedge
will underperform the long options position.
As we can see in the above figure, a long position in option is convex and there has a positive
gamma. To delta hedge, the trader will need to sell stocks if the stock price goes up and buy
stocks if the stock price goes down (sell high – buy low). The trader can sit on the bid and offer.
A short position in option is negative gamma. In this case, the trader will need to sell stocks if
the stock price goes down and buy stocks if the stock price goes up to be delta hedged (sell
low – buy high). The trader has to cross the bid-offer spread. CONFIDENTIAL
25
From Delta hedging to Convexity Hedging
◼ About 70% of the variation of the liability price is explained by the first order delta
◼ Most of the remaining 30% are explained by the second order gamma
26
How much the delta hedge will underperform ?
The actual amount depends on the difference between the initial and final delta of the
option, which is gamma time the change in the spot price.
Let us assume that gamma remains constant and see how the gamma P&L can be
calculated.
27
How much the delta hedge will underperform ?
28
Delta-Gamma hedge & P&L
The large variation in P&L linked to jump’s in the underlying price is the un-hedged Gamma at
work; By daily rebalancing you update prices and delta on a daily basis. In the monthly model
your delta model is supposedly correct for the entire month, which we know is not true. The
error creeps in and is responsible for the divergence in results you see.
To make the gamma correction we need to adjust our estimate of delta and plug in a revised
delta (and hopefully improved delta) in our simulation model - > the results with gamma
correction are more tightly centered around zero
29
Second order hedging - Mechanics
30
A delta-gamma hedge (1)
We have a derivative which is a function of a factor: c(S,t)
We will hedge the stock and another derivative on it:
S and c (2) (S,t)
A Taylor Expansion:
c = c t + c S + 1 c (t) 2 + 1 c (S) 2 + c tS +...
t S 2 tt 2 SS tS
+ 12 (cSS + 2 cSS
(2)
)(S) 2 + (ctS + 2 ctS(2) )tS +...
31
A delta-gamma hedge (2)
We have a derivative which is a function of a factor: c(S,t)
We will hedge the stock and another derivative on it:
S and c (2) (S,t)
P = (ct + 2 ct(2) )t + (cS + 1 + 2 cS(2) )S + 12 (ctt + 2 ctt(2) )(t) 2
+ 12 (cSS + 2 cSS
(2)
)(S) 2 + (ctS + 2 ctS(2) )tS +...
32
Delta / Gamma / Vega Hedging – Mechanics 1/2
33
Delta / Gamma / Vega Hedging – Mechanics 2/2
34
Delta-Gamma hedge (3)
Current Price: S = 10, Risk Free Rate: r = 0.05
Delta vs. Delta-Gamma Hedge
6
call
delta
delta-gamma
4
2
Call Option Price
-2
-4
-6
0 5 10 15
Stock Price
1.5
Call Option Price
0.5
-0.5
-1
8 8.5 9 9.5 10 10.5 11 11.5 12
Stock Price
37
From Greeks to PnL
A Taylor Expansion:
38
Higher Orders Hedging
Theta x dt Rho x dr
Impact Impact
Vega x Vol
AV ' 10 Impact
OV (AV ', r',t', ') = OV ( AV, r,t, )+ (t'−t )+ − 1 + i (r'i −ri )+ ('− )
AV i=1
AV' 1
2
AV ' −1
−1 + rho i ( r'i −ri ) + DeltaRho i (r' i −r i )
Gamma x dS² 1
+ Equity
10
2
10
39
Vega hedging
40
Risk Management improves Hedge Effectiveness
41
Map
◼ Hedging principles
Asset vs. Liability Management principles
From linear to non linear risks
The short convex liability issues in practice
43
The Delta Greek
44
Delta
◼ The Delta is the change (in pct) of the option price for a
1% change in the underlying asset price - the hedge ratio
0.250
0
0.200
0
0.150
Pric
0 e
Intrinsic
0.050 Value
0
0.100 Delta
0
0.000
1.700
1.730
1.760
1.790
1.820
1.850
1.880
1.910
1.940
1.970
2.000
0
0
0
46
Delta hedging with futures
◼ The delta of a futures contract is exp(r−q)(T−t)
47
Delta of a call vs. Days to Expiry
100
%
90%
80%
70%
1
Delt
60%
3
a
50% 0
4
40%
5
30%
K =
20% 0.95
10% =
1.700
1.650
1.660
1.670
1.680
1.690
1.710
1.720
1.730
1.740
1.750
1.760
1.770
1.780
0% 10%
0
0
Forwa
rd
48
Delta of a call Volatility
100
%
90%
80%
70%
5%
Delt
60%
15
a
50% %
20
40% %
30%
20% K=
10% 0.95 T
0% = 6m
Forward
49
Delta of a call vs. Moneyness
50
Delta as a moneyness measure
52
Aymeric KALIFE - 2010 53
54
◼ As an example, for a typical annual volatility of 0.2 (20%) and an
expiration of one year, we have Δ ≈ 0.5 + 0.04 = 0.54
◼ Suppose J = 0.01, a 1% move away from the at the money. And also
assume T = 1 year and Σ = 0.2
◼ Thus, Δ decreases by two basis points for every 1% that the strike
moves out of the money for a one-year 20%-volatility call option.
◼ Thus the strike of the 25-delta call is about 115. Actually it’s about 117 if
you use the exact Black-Scholes formula to compute deltas.
55
◼ The key variable here is the fractional shift J
divided by the square root of the annual variance.
For a greater volatility or time to expiration, the
distribution of the stock is broader, and you need a
bigger move in the strike to get to the same Δ.
57
Gamma
◼ When you are deep out or deep in, Delta is flat and asymptotic as
shown above. But when are you not, as long as prices move in small
increments and do not jump dramatically, Delta hedge will still cover
you.
◼ However a large move can result in significant trading loss despite
being Delta hedged. The underlying jumps, you are exposed.
◼ Delta is the first order rate of change and works well within a narrow
band. Within and outside that band Gamma tracks not just the error but
also the magnitude of your gain/loss in case of a large move (up/down).
◼ The magnitude of the error shifts dramatically as the option gets closer
to the At/Near money state.
58
Gamma
◼ The Gamma is the change in the Delta with
respect to spot
◼ Gamma is to Delta as Acceleration is to Speed
◼ Gamma = δ2C/δS2
When you buy an option, you have a positive Gamma
position
When you sell an option, you have a negative Gamma
position
◼ Options with high Gamma require more frequent
hedge adjustments than options with low Gamma
Note
Gamma decreases with time
Gamma decreases with higher volatility
59
Gamma of a call vs. Moneyness
60
Weighted Gamma for portfolio
What is important is ability to compare Gamma of different options that are in a
portfolio to know the sign of the net gamma exposure.
58
Up Gamma vs. Down Gamma
◼ When options are deep in or deep out, similar to Delta, Gamma also
flattens out. However given the convex nature of the 2nd derivative in this
case, the impact of a large up move or a large down move is not symmetric.
62
Gamma Bleed: Days to expiry
63
Gamma vs. Implied Volatility
◼ We have seen that volatility attacked the delta, so it is not surprising it weakened the
gamma too. Since a higher volatility induces a less pronounced S-shape delta curve, it
also induces a less pronounced/wider more stable bell-curved gamma. A higher
volatility lowers the gamma in the ATM region and increases the gamma in the
ITM/OTM regions.
◼ It is very easy to think about this effect in terms of time value of options: for low levels
of volatility, deep ITM/OTM options have little time value and can only gain time value
if the underlying asset moves closer to strike. For high levels of volatility, both
ITM/OTM options have time value and the gamma near strike should not be too
different from away from strike. Therefore, gammas tend to be more stable across all
strike prices.
64
Shadow (vol adjusted) Gamma
◼ A change in the asset price suggests a change in volatility. However the
gamma calculation assumes that volatility remains unchanged.
65
Theta or time decay
◼ Theta is the sensitivity of the value of an option to changes in time (usually on a 1 day
change basis)
◼ It indicates the rate at which the value of the option’s time premium declines overtime.
For a call option the theta is always negative, which means that the value of the option’s
time premium will always dissipate as the options approaches maturity.
◼ It is not a monotonic (i.e. only increasing or only decreasing) function in return volatility
nor in Moneyness
For deep out of the money options there is no dissipation in the time premium and theta approaches zero as the
stock approaches zero
As the underlying asset value approaches the strike price, the rate of dissipation increases, with maximum at S =
Strike*exp(Risk free rate * time to maturity + (Vol^2)*time to maturity/2)
When the option is in the money, the rate of decay falls and then levels out for deep in the money options,
approaching the value: Risk free rate * Strike * exp(-Risk free rate * Time to expiry)
66
European Call Option Theta
-0.2
-0.4
-0.6
Theta
-0.8
-1
-1.2
-1.4
8 8.5 9 9.5 10 10.5 11 11.5 12
S
6,000.0
a
4,000.0
0
2,000.0
0
1.690
1.990
1.600
1.630
1.660
1.720
1.750
1.780
1.810
1.840
1.870
1.900
1.930
1.960
0
0
-
Spo
t
68
Time premium & Theta vs. rates
69
Theta analytics (1)
◼ The first term is negative again because the variance of the stock
price at maturity decreases as time-to maturity decreases.
◼ The second term is positive.
The PV of the strike grows over time. i.e., with less time-to-
maturity.
Put receive the strike, so this tends to make the put more
valuable as time passes. 72
Theta analytics (4)
There are 2 situations where the Theta turns positive:
• From ATM to ITM: as the stock moves down and the put price rises, BUT the
time value is less than for ATM option. Clearly, if the time value is less altogether,
there is less value to lose when time passes. Remember the time value of ITM
put options can be negative, so nothing really surprising here.
• From ATM to ITM then back to ATM: During the first week, the stock does not
move so that both time value and option value decrease; during the 2nd week,
the stock drops by 10% so that option price increases (due to intrinsic value) and
time value decreases. We know that if the time value is smaller, the daily
decrease in time value is lower than before the actual move. If the stock moves
back up to the ATM level, the option value goes down but the time value
increases. As of then, the daily decrease in time value is more substantial than
before this move back up.
2.For ITM call options, the theta can turn positive as well, in the case where the
dividend yield is larger than interest rates so that the forward level is below
the current spot level. 73
Theta vs. Gamma
◼ When Theta is positive, Gamma is negative and vice-versa: when you buy an option, you
have negative Theta position, i.e., you pay time decay. When you sell an option, you
have positive Theta position, i.e., you earn time decay
◼ When the spot moves, the gamma and theta move the same way
The ratio gamma / theta is stable
Theta is rather linear with gamma
74
The Vega Greek
1.8
1.6
1.4
1.2
1
Vega
0.8
0.6
0.4
0.2
0
8 8.5 9 9.5 10 10.5 11 11.5 12
S
78
European Call Option Vega vs. residual maturity
79
Weighted Vega
59
Vega hedging
0.40%
0.35%
0.30%
0.25%
vol = 8%
Vega
0.10%
0.05%
0.00%
90 100 110 120 130
Strike 61
Shadow Delta / Vanna
◼ A change in the asset price suggests a change in volatility-> adjustment in the delta
measure that considers that the underlying asset price volatility has changed -> shadow
delta measure
◼ Vanna has negative values when the underlying price is higher than strike (in our case
S>$100) and it has positive values when the underlying moves just below it (S<$100):
◼ The graph highlights the fact that vega moves much more when the underlying asset
approaches the ATM strike ($100 in our case) but it tends to approximate 0 for OTM
options. Consequently, the delta is very sensitive to changes in implied volatility when
the ATM area is approached.
◼ However, it is important to point out that delta will not always increase if the underlying
moves from, say, $80 to $100 because in many risky assets (stocks, equity indices,
some currencies and commodities) the implied volatility is inversely correlated to the
stock price, would decrease vanna which, in turn, would diminish the value of delta.
63
Vega skew - dVega/dSpot
■ Model risk
What other derivatives can we price correctly using flat volatility? The reason why
we can apply such a model in this case is for an option that does not have skew
sensitivity or any hidden convexities other than that to the underlying’s price e.g.
Asian options).
An other example, which involves skew, is the call spread. Although the call spread is
sensitive to skew, it is only sensitive to two specific points on the skew: the
volatilities at the correct strikes. This payoff can be broken down into two call
options that can be correctly priced if we have the right implied volatilities for
each: for the first we use the implied volatility of the underlying for that particular
strike K1, plug into Black–Scholes and obtain the first price. Similarly, for the second
all inputs are the same except the strike and the implied volatility of the K2 strike call
option. The difference between these two prices is the price of the call spread.
The reason we were able to do this is because the payoff of the call spread is a linear
combination of two vanillas. It is possible to extend the case of a constant volatility
across maturities to having a time dependent but deterministic volatility, thus
allowing for a term structure of volatilities.
62
Vega skew - dVega/dSpot
■ Local Volatility
Local volatility models offer a way of capturing the implied skew without introducing additional
sources of randomness; the only source of which is the underlying asset’s price that is modelled
as a random variable. In a local volatility model, the volatility of the underlying asset’s price is a
deterministic function of the asset’s price.
Local volatility model extends beyond skew and can also capture term structure. It can therefore
theoretically supply us with a model that gives the exact same prices for vanillas taken from a
whole implied volatility surface.
Given the set of implied volatilities of vanilla options, calibration is the process where we search
for these volatilities σ(S(t), t) so that the model matches these prices. These models can then be
used to price more exotic payouts, knowing that the model correctly prices the liquid vanillas.
There are computational difficulties, practical drawbacks. In particular, there may be more than
one (often unlimited) local volatility model that fits a set of vanillas, so one must lay down a set
of criteria to follow when choosing the model to use.
Recall that the surface is two-dimensional, one in time and one in strike. Depending on the
payoff, we may want to favour one set of functions σ(S(t), t). If the payoff involves only one date
– for example only on the value S(T ) at maturity but not before –there is no path dependency. In
this case, we only need to fit the distribution to that maturity by taking the vanilla skew at that
date only, and the focus is on the strike regions in which the payoff is sensitive.
62
Vega skew - dVega/dSpot
Local volatilitymodel
■ Computing Forward implied volatility :
𝜎,6,,7 is the volatility between two points 𝑡6and 𝑡7, using n unequal time slice betwwen
𝑡3 𝑎𝑛𝑑 𝑡5, we have :
It is possible to infer the local volatility between 𝑡6and 𝑡7 while knowing the volatility between
𝑡3and 𝑡6 and and that between 𝑡3and 𝑡7 ∶
Example :
The volatility 90 days and 180 days in the market are 17% and 15,5%respectively. So the
volatility 3months in 3 months is:
62
Vega skew - dVega/dSpot
■ Local Volatility
As an example, consider a call option that has the usual payoff while additionally paying nothing if
the underlying goes above a certain level: these are known as barrier options. As we will see in
detail in barrier options, we must capture the skew if they are to be correctly priced; if this barrier
is only monitored at maturity, this has no path dependency and we require only that the model
fits as best as possible the vanillas maturing on the same date as this option.
Assuming that the call option is ATM (strike is at 100%), and the barrier is at 150% for example, we
must make sure that the model correctly fits the skew up to the point 150%. The focus here is on
getting the calibration to correctly fit the places where the derivative has skew sensitivity.
European options with these strikes can serve as hedging instruments and the model must be
calibrated to them to show risk against them. In the described example, the option is sensitive to
skew, however its sensitivity to the volatilities in the 70% strike region for example is minimal.
Getting the calibration right in that region should not be done at the expense of less accuracy in
the skew-sensitive region.
Take, on the other hand, a derivative that has a huge amount of path dependency and again look
at the above example of a knock-out call option that pays like a call option unless the asset goes
above the barrier, in which case it pays nothing. Now allow the barrier to be monitored on a daily
basis. At the close each day we see if the underlying went above this barrier level, and, if so, the
option is then immediately rendered worthless.
We cannot in fact find liquid vanillas for all dates, and therefore we have to work with a finite
set of maturities to which we can calibrate. In this instance we will only need vanillas of
maturities up to the maturity of the knock-out call we are pricing, specifically because these can
serve as hedging instruments for Vega or Gamma risk and must therefore be correctly priced in
the model. What we do need to know is that the local volatility model does imply a reasonable
volatility for those dates in between the dates where we have vanilla data; that is, we want to
know that, through time, the calibration is smooth. We also want to make sure that the strikes are
calibrated as best as possible, but must balance between these and a smooth calibration through
time. This case we refer to as smooth surface calibration. When implied volatility data is not
available, one must resort to interpolation or extrapolation of the surface. Doing this in62 an
arbitrage-free manner.
Vega skew - dVega/dSpot
■ Local Volatility model arbitrage-free conditions
Firstly, for all maturities T in the above set, there cannot be any negative call spreads. If there was
a negative call spread this would imply an obvious arbitrage.
Secondly, one can approximate a binary payoff using a call spread, and the use of the two
closest strikes must yield a value less than 1 for these call prices to be arbitrage free
Any interpolation between the implied volatilities of two consecutive strikes in the above set
must also observe these conditions to be arbitrage free.
62
Vega skew - dVega/dSpot
■ Stochastic Volatility
As we saw in the example of the call and put spreads, options that can be broken down into
vanillas can be priced using Black–Scholes as long as one uses the right implied volatility for each
option. Now consider options such as the aforementioned barrier option. These have skew
dependency yet such payoffs cannot be broken down into vanillas. In such cases, we can use
local volatility assuming it is correctly calibrated to the skew (or surface) in a manner consistent
with the skew sensitivity of the option.
In more complex payoffs, as we will see as we progress, almost all the payoffs will exhibit some
form of Vega convexity, although in many cases this is captured in the skew and can be correctly
priced by getting the skew right (for example, with a local volatility model). Other payoffs exhibit
such convexities that are not captured in the skew and we must in these cases use stochastic
volatility within which volatility is taken to be random, with its own volatility, known as the
volatility of volatility, or vol-of-vol.
Although local volatility models can capture the market’s consensus on the prices of vanilla options
by matching the volatility surface, the evolution of future volatility implied by these models is not
realistic. In the case of forward skews we are faced with the problem that the forward skews
generated by local volatility models flatten out as we go forward in time. The local volatility
model, therefore, does not provide the correct dynamics for products with sensitivities such as
these.
In stochastic volatility models, the asset price and its volatility are both assumed to be random
processes -> give rise to forward implied volatility skews and term structures. Owing to the
randomness of volatility – they generate forward skews that do not fade
Once such a model is specified, the skews generated by the model are a function of its parameters, and
finding the parameters that fit a certain skew (or a surface) is again the act of calibration.
62
Vega skew - dVega/dSpot
Heston’s stochastic volatility model
In its
basic form, the Heston model describes the evolution of the two processes according to the following set
of SDEs:
where 𝑣 , is the instantaneous variance, and 𝑑 𝑊 , and 𝑑𝑍 , are Brownian motions with correlation ρ.
The first feature to note is that the variance is modelled by a mean-reverting process. The term κ (θ − 𝑣 , )dt in
the drift of the variance governs how it reverts: θ is the long-term variance to which this process reverts, and κ
is the rate at which it reverts to this mean.
The term σ is the vol-of-vol term
The correlation ρ is the coefficient that governs joint movements in the stock and its variance
The time t price of a call option 𝐶(𝑆3, 𝐾, 𝑣3, 𝑡, 𝑇), with maturity T , under Heston’s model satisfies the partial
differential equation (Heston, 1993):
62
Vega skew - dVega/dSpot
Heston’s stochastic volatility model
Additional terms to the Black–Scholes PDE due to the variance being taken as a
random process :
■ Vega convexity term :
In equities, we know that the skew is downward sloping, and this will be reflected
inthe calibration as we will find a suitable calibrated value for ρ to benegative.
62
Shadow Greeks
57
Weighted Vega
What is important is ability to compare vega of different options that are in a portfolio
to know the sign of the net vega exposure.
If you sold $100mm of 1m straddles and bought $40mm of 6m straddles, your
vega would be zero - is your risk really zero?
1m bucket short 110,000 of vega, 6m bucket long
110,000
Clearly you have spread risk (if the spread moves against you by
1% you lose $110,000)
Is there a way of representing the risk in one number?
45.00%
40.00% 38.58%
35.00% 33.71%
30.00% 29.04%
24.91%
Volatility %
25.00% 20.81%
20.00%
13.3% 13.7%
15.00% 12.5% 12.7% 12.9%
10.00%
12.50%
8.77% 11.13%
5.00% 7.15% 8.51%
0.00%
0 50 100 150 200 250 300 350 400
AymericDays
KALIFE - 2010 58
Weighted Vega post a vol shock
◼ Weighted Vega Measures sensitivity to
nonparallel changes in volatility curve
◼ Theoretical Weighting
Square root time rule
Start with e.g. 30-day exposure
Weight exposures at other maturities by
(30/t)1/2
Example: 120 day Vega is weighted by
(30/120)^0.5 = 0.5
So $1M Vega risk in one month is equivalent to
$2M exposure at 120 days 59
Weighted Vega post a vol shock
Weighted vega is the sum product of vega per bucket
multiplied by bucket weightings
Bucket weightings
theoretical is proportional to 1/sqrt(t)
in practice slightly different to reflect damping of longer dated vol
moves
sqrt(t) Market
1w 2.08 2.05
1m 1.00 1.00
2m 0.71 0.75
In the example: 3m 0.58 0.58
weighted 1m vega = 1.0 * (-110,000) 6m 0.41 0.38
weighted 6m vega = 0.38 * 110,000 12m 0.29 0.23
portfolio weighted vega = -68,200
If 1 month vols go up 1% we expect to lose $68,200 if the
weightings hold
Weightings unreliable if the whole level of volatility rises (the
‘flat curve’) - very unreliable for Emerging markets 60
The Rho Greek
A trader sells a call option and delta hedges by buying delta shares. To buy those
shares, he must borrow money at the bank. He will have to pay interest on his loan.
The higher the interest rates, the more interests to pay, the higher the cost of his
hedge, the higher the call price. This is the reason why rho is positive for call
options.
The discounting effect slightly offsets this delta hedge impact though
98
European Put Option Rho
A trader sells a put option and delta hedges by selling delta shares. By selling
shares, he receives money. He can put this money in the bank and receives interest
on it. The higher the interest rates, the more interests he receives, the lower the cost
of his hedge, the less expensive the put price. This is the reason why rho is negative
for put options.
The discounting effect works in the same direction and reinforces this impact.
If you make more money on your delta hedge, you are going to pay more (receive
less) for it. 99
Rho & Maturity
Rho increases as time to expiration increases. Long-dated options are far more
sensitive to changes in interest rates than short-dated options. Though rho is a
primary input in the Black-Scholes model, a change in interest rates generally has a
minor overall impact on the pricing of options. Because of this, rho is usually
considered to be the least important of all the option Greeks.
10
0
The Greeks over
time
◼ v
10
2
◼
Greeks over time (2)
Away-from-the-money options experience less time value decay as in and
out-of-the-money options have less time value than do comparable at- or
near-the-money options. Moreover, the theta associated with moderately in- or
out-of-the-money options may be relatively constant signifying linear decay
characteristics.
◼ "Deep" in- or out-of-the-money options will have very little or perhaps no time
value. Thus, the theta associated with an option whose strike is very much away
from the money may "bottom-out" or reach zero well before expiration.
◼ Time value decay works to the benefit of the short but to the detriment of the
long. Note that the same options that have high thetas likewise exhibit have
high gammas. Convexity as measured by gamma works to the detriment of the
short and to the benefit of the long. Near-themoney options will have high thetas
and high gammas. As expiration approaches, both theta (measuring time
value decay) and gamma (measuring convexity) increase. It is apparent that
you "can't have your cake and eat it too." In other words, it is difficult, if not
impossible, to benefit from both time value decay and convexity
simultaneously.
◼ When rising volatility and convexity work in your favor, time value decay
generally works against your position … and vice versa.
◼ Delta bleed achieves its highest absolute values when the options are around the ATM
area. Therefore, slightly in-the-money or out-of-the-money options will have the highest delta
bleed values. This makes sense because the greatest impact of time decay is precisely on
options “floating” around the ATM zone:
◼ In fact, deep ITM options will behave almost like the underlying asset while OTM
options with the passage of time will approach 0.
◼ Consequently, the deltas of slightly ITAMymoerricOKTAMLIFoEp-t2io01n0swill be the most eroded by time. 69
Delta bleed (2)
Warning on Friday & Week-Ends
Mathematically:
delta bleed = δDelta / δT = δ2C / δSδT
change in decay with spot = δTheta / δS = δ2C / δTδS
10
7
Delta & Gamma vs. time passing
◼ ATM
As time passes, the option
loses more of its time value,
OTM options see their delta
approach zero and ITM
options see their delta
become closer to 1.
Gamma Bleed
◼ Change in Gamma per day
10
9
◼ Delta and Gamma Bleed (2)
The bleed is the change in the delta and the gamma of an option
position with the passage of time.
Delta bleed = Delta today – Delta next day
Gamma bleed = Gamma today – Gamma next day
11
0
Vega Bleed
◼ DvegaDtime measures how fast vega is going to change with respect to the time decay
◼ Negative because time decay is clearly a price that every options holder has to pay
◼ The influence of time decay on volatility exposure measured by vega is mostly felt in
the ATM area especially for options with short time to maturity
11
1
Theta Bleed
◼ Depending on the money-ness of the option the theta may be an increasing or a
decreasing function of time:
For in and out of the money options the rate at which the option’s time premium declines
remaining fairly constant over the life of the option and then decreases as the option
approaches expiry.
For at the money options the rate of decay is much higher increasing as the option
approaches maturity.
11
2
11
3
Trading the Greeks
11
5
Delta Trade & Time spreads (1)
◼ Time spreads are also sometimes called calendar spreads or
horizontal spreads.
◼ Time spreads are identified by the ratio of options written and
purchased.
◼ To create a 1:1 time spread, the investor buys one option
and sells one option.
The two options must either be both calls, or both puts,
and share the same strike price on the same underlying
asset.
However, unlike most of the option portfolios discussed
previously, the time spread portfolio of options has
options with different maturity dates.
11
6
Delta Trade & Time spreads (2)
◼ The maximum profit on a time spread occurs
when the stock price equals the strike price on the
expiration date of the nearby call.
n1 − 0.5415
= = −1.22
n2 0.4439
83
Long Gamma Trade (1)
◼ Start with a position that is initially market neutral but that gets long if
the market rises and gets short if the market falls: Long 100 options
and simultaneously short 50 underlyings
◼ Small Underlying Price Moves: Market neutral or delta neutral
◼ Large Underlying Price Moves: Whichever way the underlying price
moves, we always make a profit
The trade works simply because of price curvature and the price profile is
only curved because of the kink on expiry
If the underlying price rises, the exposure of the option increases above
the constant exposure of the short underlying position and so the total
portfolio automatically becomes long
12
3
From Gamma to Theta trades (2)
◼ Long Put and delta hedge
12
4
From Gamma to Theta trades (3)
◼ Long Call and Put
12
5
From Gamma to Theta trades (4)
◼ When Gamma is high…
12
6
Theta trades (1)
◼ Short Call and delta-hedge
◼ Short Call and Put
12
8
Theta trades (3)
◼ At times we believe the market volatility is getting lower.
So, the time value we received can be higher than the
hedging loss in the holding period.
135
◼ f
137
Trading the Volatility Skew
Trading the Slope: Long a put spread is “short the skew”
By selling a 90% strike put, and buying the ATM put, giving the long position in the 90–100%
put spread, the investor has sold the 90% strike implied volatility and bought the ATM
volatility. If in fact the skew was steeper than it should have been and the market begins to
imply a flatter skew, the volatility of the 90% strike put will be lower and thus its price is
lower, making the price of the put spread higher. The holder of the put spread is said to be
short skew, in the sense that if skew increases, the put spread’s value decreases.
Alternatively, the seller of the put spread is long skew.
Again this has to do with supply and demand, since, in the short term, people may be less
keen to sell OTM puts (and thus the OTM put volatility). A jump in the underlying’s price in
the immediate future would have a large impact on the price of the put; for the short term
this is more severe as the market may not have time to recover. There is also an increase in
the OTM call implied volatility compared to the longer maturities, again for similar reasons.
139
Map
◼ Hedging principles
Asset vs. Liability Management principles
From linear to non linear risks
The short convex liability issues in practice
A Taylor Expansion:
142
Monitoring Option Risks
◼ f
144
Delta & Dynamic hedging
◼ Delta-hedging is the most commonly used strategy on option trading desks
to replicate the payoff of options they sell (or buy).
◼ An option’s delta is the change in the option’s value due to a unit change in
the underlying price. Mathematically it is simply the first derivative with
respect to the underlying price:
147
Gamma (2)
◼ The two graphs are identical: delta-hedged calls and puts have the same
P&L profile.
◼ The P&L is always positive: a delta-hedged long call or put will always
generate profits as the underlying price moves away from its initial price.
Unsurprisingly, there is a downside which we investigate later.
◼ The Gamma prediction of P&L is quite accurate. Thus, the gamma is a
good measure for the P&L of a delta-hedged option position caused by
movements in the underlying price: Positive gamma means profits &
negative gamma means losses. The larger the gamma and the price
movement, the larger the profit or loss.
Typically Gamma and Theta have opposite signs: For a long call or put
position, Gamma is positive and Theta is negative, i.e. the trader is long
shocks/volatility (she makes money as the stock price moves) and short
time (she loses money as maturity approaches.). For a short call or put
position, the situation is reversed.
◼ This equation tells us that the daily option P&L on a delta-hedged option
position is driven by two factors:
Dollar Gamma, which has the role of a scaling factor and does not determine
the sign of the P&L;
Variance Spread (realised vs. implied), which determines the sign of the P&L.
◼ Thus, a trader who is long dollar gamma will make money if realised
variance is higher than implied, break even if they are the same, and lose
money if realised is below implied.
◼ Summing all daily option trading P&L’s until maturity, we obtain the path-
dependency equation
i.e. sum of the daily Variance Spread weighted by the Dollar Gamma. Thus,
days when Dollar Gamma is high will tend to dominate the final P&L.
Aymeric KALIFE - 2010 152
Gamma-Theta P&L (3)
Suppose a market maker sells and
delta-hedges a vanilla option.
Assuming that whatever volatility is
realized is constant and option is
delta-hedged over infinitesimally small
time step, then the market maker will
profit if and only if the realized
volatility is larger than the implied
volatility purchased.
However, volatility is not
constant! Furthermore the
magnitude of the P&L depends where
that volatility is realized in relation to
the option strike.
If the volatility realizes close to
maturity when the underlying trades
near the strike then the gamma would
be very high and the absolute value of
the P&L will be larger.
-> For non-constant volatility, it is
possible to buy and delta-hedge an
option at an implied volatility smaller
than the subsequent realized volatility
and still losing money from delta-
hedging! Aymeric KALIFE - 2010 153
Gamma-Theta P&L (4)
The SX5E index is initially quoting at 3500 with an ATM implied volatility of 28.5%. During the
first seven months, the index ranged between 3500 and 3800 points, realizing a volatility of
20%. After that period, the index fell rapidly to 2500 points, realizing approximately 50%
volatility on the way. Over the whole year, the realized volatility was 36%.
Conclusion
When delta-hedging daily an option, the P&L is a daily accrual that depends on the difference
between the realized volatility and the implied volatility. The magnitude of the contribution of
this daily accrual is weighted by the current dollar gamma, which is unpredictable since path-
dependent Aymeric KALIFE - 2010 154
Gamma-Theta P&L (5)
◼ Break-even volatility
158
Market-Making and Delta-Hedging
159
Market-Making and Delta-Hedging
160
Market-Making and Delta-Hedging
161
Market-Making and Delta-Hedging
162
Market-Making and Delta-Hedging
163
Market-Making and Delta-Hedging
164
Market-Making and Delta-Hedging
165
Market-Making and Delta-Hedging
166
Market-Making and Delta-Hedging
167
Market-Making and Delta-Hedging
168
Market-Making and Delta-Hedging
169
Market-Making and Delta-Hedging
170
Market-Making and Delta-Hedging
171
Market-Making and Delta-Hedging
172
Market-Making and Delta-Hedging
173
Market-Making and Delta-Hedging
174
Market-Making and Delta-Hedging
175
Market-Making and Delta-Hedging
176
Delta-Gamma-Theta Hedging (1)
Suppose is 0.5824, when S = $40
A $0.75 increase in stock price would be expected to increase option value by $0.4368 ($0.75 x
0.5824)
The actual increase in the option’s value is higher: $0.4548. This is because increases as stock price
increases. Using the smaller at the lower stock price understates the the actual change. Similarly,
using the original overstates the change in the option value as a response to a stock price decline
Can estimate new option value using alone when stock price moved up (down) by e.
(e = St+h – St)
Using the − the accuracy can be improved a lot
1
C ( St+h ) = C ( St ) + e ( St ) + e 2 ( St )
2
178
Delta-Gamma-Theta Hedging (3)
−− approximation
: If a day passes with no change in the stock price, the option becomes
cheaper. Since the option position is short, this time decay increases the
profits of the market-maker.
Interest cost: In creating the hedge, the market-maker purchases the stock
with borrowed funds. The carrying cost of the stock position decreases the
profits of the market-maker.
C ( St+h , T − t − h)
1 2
= C ( St , T − t ) + e ( St , T − t ) + e ( St , T − t) + h( St , T − t)
2
179
PnL uncertainty :
variance and drivers
of PnL
180
Hedging with future realized (known)
Volatility (1)
181
Hedging with future realized (known)
Volatility (2)
182
Hedging with future realized (known)
Volatility (3)
183
Hedging with implied Volatility (1)
185
Hedging with arbitrary constant Volatility
187
Hedging error from discrete hedging (2)
189
Hedging error from discrete hedging (4)
190
Hedging error from discrete hedging (5)
194
Rebalancing frequency & efficiency of the hedge
195
Volatility empirics
50%
Realized Volatility (S&P 500, 6M)
As markets calmed post the March 2003
45% rally we asked ourselves:
Have we transitioned out of a “high- A “high-vol
40% volatility regime” into a “lower-volatility regime?”
regime?”
35%
30%
Or is this a
S&P 500 6-Month ATM lull in a
25% Implied Volatility as of longer-term
09-Dec-05: 13.3%
“high-
20%
volatility”
environment?
15%
10%
A “low-vol
5%
regime?”
0%
1941 1951 1961 1971 1981 1991 2001
Sources: Deutsche Bank estimates and calculations, Bloomberg, Reuters.
197
Volatility: Which regime is it anyway?
Realized Volatility (S&P 500, 6M)
50%
45%
15%
10%
S&P 500 6-Month
5% Realised Volatility:
Median 11.9%
0%
1941 1951 1961 1971 1981 1991 2001
Sources: Deutsche Bank estimates and calculations, Bloomberg, Reuters.
198
Volatility: Even Longer-term historical context
Realized Volatility (Dow Jones Industrial Average, 6M & 60M)
Looking back even longer
on the Dow we see that
60%
6M Realised the recent spate of high
60M Realised volatility was nowhere
near so high nor
50% prolonged as after the
1929 crash ...
20%
We could be here
10% for a while …
0%
Sources: Deutsche Bank estimates and calculations, Bloomberg, Reuters.
1901 1921 1941 1961 1981 2001
199
Implied Volatility in a Historical Context
S&P 500 and EUROSTOXX 50 Index 3-Month ATM Implied Volatility (mid-market)
55%
Implied Volatility has fallen
50% significantly from
recent highs and despite
45% bouncing remains close to 10
year lows in the U.S. and
40% Europe ...
35%
30%
EURO STOXX
25% 50 3-Month
ATM Implied
Volatility:
20% 13.6%
15%
5%
mai-95 mai-96 mai-97 mai-98 mai-99 mai-00 mai-01 mai-02 mai-03 mai-04 mai-05
Sources: Deutsche Bank estimates and calculations, Bloomberg, Reuters.
200
Market volatility: Reactions to events
Realized Volatility (Euro STOXX 50, S&P 500, 3M)
Credit related Starting in 1997,
sell-offs of Jul- global equity markets
and Oct-02 lurched from one
crisis to the next.
55%
Volatility seemed
50% LTCM/Ruble 9/11 increasingly a part of
crisis equity life.
45% Tech bubble
expands and
40% bursts
Until the last two
35% Asian financial years, when market
crisis volatility has fallen
30% Madrid to levels not seen in
bombings several years
25% London Are we getting used
bombings to crises?
20%
15%
10%
5%
janv-97 janv-98 janv-99 janv-00 janv-01 janv-02 janv-03 janv-04 janv-05
Sources: Deutsche Bank estimates and calculations, Bloomberg, Reuters.
201
Implied and realised volatility as monitors of risk aversion
Ratios of 3M ATM implied to realised volatility on selected global indices
2,00
S&P 500 EURO STOXX 50 Nikkei 225 FTSE 100
1,75
1,50
1,25
1,00
202
What about longer-dated implieds?
S&P 500 and EURO STOXX 50 Index
5-Year ATM Implied Volatility (mid-market)
40%
35%
High volatility, rising interest rates
→ income products EURO STOXX 50
30% Longer-dated implied
volatilities are most
affected by retail … and hedging
25% product supply/demand …
S&P 500
20%
15% Low volatility, falling interest rates Low volatility, low interest rates
→ growth/participation products ?→ growth/participation products
10%
oct-95 oct-96 oct-97 oct-98 oct-99 oct-00 oct-01 oct-02 oct-03 oct-04 oct-05
Sources: Deutsche Bank estimates and calculations, Bloomberg, Reuters.
203
Different forces acting on implied volatility
EURO STOXX 50 ATM Implied Volatility Term Structures over 2 years
(Fear of)
Demand
from
24% 09-déc-05 09-déc-04 09-déc-03 structured
products
22% Falling
realised
volatility Supply from
20%
overwriters, vol
sellers and
18% closed hedges
Longer-dated hedges
16%
(Insurance)
More to come?
14%
12%
Some short-term hedges, event plays
10%
1M 3M 6M 12M 24M 36M 48M 60M
204
What have they done to the Term Structure?
10% EURO STOXX 50 Index
5-Year minus 3-Month ATM Implied Volatility Term Structure (mid-market)
5%
0%
-5%
Since the market rally
of March 2003 the term
structure has steadily
-10% increased to historically
high levels
-15%
The term structure
became steeply
-20% downward-sloping in the
aftermath of 9/11 and
the credit worries of
Jul- and Oct-2002
-25%
mai-99 mai-00 mai-01 mai-02 mai-03 mai-04 mai-05
Sources: Deutsche Bank estimates and calculations, Bloomberg, Reuters.
205
The Smile post 1987 market crash
◼ The smile’s appearance after the 1987 crash
and was clearly connected in some way with
the visceral shock of discovering, for the first
time since 1929, that a giant market could
drop by 20% or more in a day or two.
8 0 .0 0 %
7 0 .0 0 %
6 0 .0 0 %
5 0 .0 0 %
Volatili
4 0 .0 0 % Im p lie d
ty
3 0 .0 0 %
2 0 .0 0 %
1 0 .0 0 %
0 .0 0 %
0 50 Aymeric
100 1 5KALIFE
0 2 0 0 - 2010
250 300 350 167
S t ri k e s
Some Behavioral Reasons for an Implied Volatility Skew
◼ 1. Out-of-the -money puts provide crash protection cheaply, which may have
driven up implied volatilities after 1987. Also, realized volatility will increase in a
crash, therefore this is self-consistent. Equity Risk is on the downside as most
market players are naturally buyers of stocks, leading investors to buy put
protections and selling calls to finance the operation : Skew represented by
difference of implied volatility between OTM puts and OTM calls.
◼ 2. The minimum volatility as a function of strike occurs near atm strikes strikes
corresponding to slightly otm call options. Low strike volatilities are usually higher
than high-strike volatilities, but high strike volatilities can also
◼ 3. The volatility of implied volatility is greatest for short maturities, as with Treasury
rates.The term structure is usually increasing but can change depending on views
of the future. After large sudden market declines, the implied volatility out-ofthe-
money calls may be greater than for atm calls, reflecting an expectation that the
market may rebound.
◼ 7. Shocks across the surface are highly correlated. There are a small number of
principal components or driving factors. We’ll study these effects more closely
later in the course.
Aymeric KALIFE - 2010 169
Implied Volatility empirical features (1)
◼ The empirical distribution of asset returns is leptokurtic
the empirical 4th moment around the mean is greater than for a normal
distribution with the same variance => more extreme returns => fewer
midrange returns than would be expected under a Gaussian distribution =>
higher peak and fatter tails.
211
Implied Volatility empirical features (3)
◼ Volatility clustering and mean reversion
periods of high and low volatility often appear in “blocks”: large moves
follow large moves and small moves follow small moves, implies that
volatility (or variance) is auto-correlated
◼ Leverage effects
falling share prices increases the debt-to-equity ratio of firms, which
leads to higher uncertainty thus volatility of the share price. Hence
price movements and volatility are negatively correlated as
empirically observed below
212
Implied Volatility skew modeling (1)
Implied Volatility skew modeling (2)
Implied Volatility skew modeling (3)
Implied Volatility skew modeling (4)
Skew reflects continued high risk aversion
The State Street confidence indicator shows that investor positioning is still near its lows.
115
State Street Investor Confidence Index
110
105
Investor Confidence
100
95
90
85
80
75
sept- mars- sept- mars- sept- mars- sept- mars- sept- mars- sept- mars- sept- mars- sept- 98
99 99 00 00 01 01 02 02 03 03 04 04 05 05
Source: State Street, Deutsche Bank European Strategy calculations
217
Implied vol skew: Historical context
Steepness of the Skew (S&P 500, Euro STOXX 50, 90% minus 110%, 3M)
16%
14%
12%
After steepening
materially, global index
10% implied volatility skews
retreated rapidly.
8%
6%
4%
2%
juin-95 déc-96 juin-98 déc-99 juin-01 déc-02 juin-04
Sources: Deutsche Bank estimates and calculations, FAME, Bloomberg, Reuters.
218
Implied vol skew: Moves over one year
Steepness of the Skew (S&P 500, FTSE 100, 90% minus 110%, 3M)
8,5%
8,0%
7,5%
7,0%
6,5%
6,0%
5,5% The spread between S&P
500 and FTSE 100 skew
5,0% has flipped several times
this year as hedging
4,5% demand has ebbed and
4,0% flowed in either market.
3,5%
févr-05
mars-05
janv-05
oct-05
avr-05
déc-05
mai-05
juil-05
sept-05
nov-05
août-05
juin-05
219
Comparing index skew and ATM implied vols
Implied Vol Skew Steepness (3M, 90%-110%)
and ATM Implied Vol (3M): EURO STOXX 50 Index
Although ATM
12% 55% implied fell
steadily from
11% 50% March 2003,
skew stayed
10% 45% firm until early
September ‘04.
Implied Volatility
9% Skew 40%
8% 35%
Skew
mai-05
mai-00
mai-01
mai-03
mai-04
sept-02
sept-04
sept-00
sept-01
sept-03
sept-05
janv-02
janv-00
janv-01
janv-03
janv-04
Sources: Deutsche Bank estimates and calculations, FAME, Bloomberg, Reuters. janv-05
220
Options Open Interest, a Shift in Strategy U.S.
S&P 500 Index Option Open Interest
6 000 000
Puts
Open Interest (number of contracts)
5 000 000
Variance
As implied volatilities have receded, the dominant form of hedging
4 000 000 hedging has moved from collars to straight puts also
playing a
3 000 000 part?
2 000 000
Calls
1 000 000
0
mai-97 mai-98 mai-99 mai-00 mai-01 mai-02 mai-03 mai-04 mai-05
Sources: Deutsche Bank estimates and calculations, FAME, Bloomberg, Reuters.
221
Options Open Interest, a Shift in Strategy Europe
14 000 000
DJ EURO STOXX 50 Index Option Open Interest
Open Interest (number of contracts)
12 000 000
In Europe the trend has not been so apparent, but puts started
to lead calls around the beginning of 2003
10 000 000
4 000 000
Puts
2 000 000
août-05
oct-05
juin-02
juin-03
juin-04
juin-05
févr-03
févr-04
févr-05
avr-03
déc-03
déc-02
déc-04
oct-02
oct-03
oct-04
avr-04
avr-05
août-02
août-03
août-04
Sources: Deutsche Bank estimates and calculations, FAME, Bloomberg, Reuters.
222
Map
◼ Hedging principles
Asset vs. Liability Management principles
From linear to non linear risks
The short convex liability issues in practice
• Usually the higher the rebalancing frequency, the lower the replication cost
• However such cost is proportional in the volatility, which may increase on different
business time units depending on market environment thus flows: intra-daily, daily,
weekly…
• Besides, when realized volatility increases, large moves toward the end of day may occur.
The market often “bounced back” the following day, leading to mean reversion thus costly
portfolio rebalancing if based on intra-day or daily time units (vs. weekly)
225
Volatility risk & rebalancing frequency (2)
• When the Volatility is low, the expected hedging cost from delta hedging is relatively low
• When the Volatility shoots up, the expected hedging cost from delta hedging rises significantly
226
Volatility risk & Tail Risks
• Equity market volatility exhibits different trading patterns than the equity market
• Negative correlation to equity markets
• Positive spikes during market downside
• Mean reversion: historically, implied volatility rarely traded below 10%
227
Tail Risks (1)
◼ Example for delta hedging of an annual ratchet GMDB (i.e. look-back options),
especially under real-world conditions, is more costly and difficult than hedging vanilla
puts
◼ Higher-order hedging enables to reduce Tail Risks, through the use of options
◼ However such higher-order hedges are costly and far lower liquid
Bid-ask spreads of options >> bid-ask spreads of Futures (2 bps vs. 20-50 bps), pct within volatile market conditions
Option volumes << Futures volumes
229
Dealing with partial
Gamma / Vega hedging
230
Partial Delta / Gamma / Vega hedging: Generic Approach
• We hedge Gamma and Vega by buying other options (specifically cheaper out of
money options) with similar maturities. Like Delta hedging we need to rebalance but the
rebalance frequency is less frequent than Delta hedging
• objective is to hedge Gamma and Vega exposure using a universe of cheaper options on the
same date
• -> Your final hedge is therefore a mix of exposure to the underlying (partial delta hedge) and
cheaper options with similar maturities
• -> it’s a large universe of options out there, how do we manage multiple constraints including
premium & sensitivities across products, maturities (tenors), Delta, Gamma & Vega
• Should you leave some residual Gamma and Vega exposure or should our hedging
portfolio neutralize it in its entirety?
• If that is not possible we will let the positions run with the Gamma and Vega mismatch but use
exposure limits to track our sensitivity to these two factors
• -> imperfect semi-hedge which would leave some room for benefiting from unanticipated
changes in market volatility as well as large unexpected jumps in the price of the underlying
231
Partial Delta / Gamma / Vega hedging: Implementation
• 1st step: put aside the universe of options available for hedging.
• 3rd Step: configure Solver options to ensure non-negative values. In addition to the non-
negative values we also want to use a quadratic model for estimates and conjugate
method for searching optimal solutions
• 4th step: track Vega and Gamma residual unhedged exposure across maturity bucket
and update hedging portfolio accordingly, by specifying hedging gap limits by maturity
• 5th step: restrict the number of options of the hedging portfolio in order to make it both
manageable and cost efficient
•
232
Partial Delta / Gamma / Vega hedging: Illustrations
• ensure non-negative values (Click on Options and select the check box against
“Assume Non-Negative”).
• use a quadratic model for estimates and a conjugate method for searching for optimal
solutions.
233
Partial Delta / Gamma / Vega hedging: Checks
• Is there a better solution possible? Let us add new conditions to refine the model:
hedge portfolio Gamma and Vega should be greater than the target position by forcing
allocation > 100%
• What if rather than focusing on Gamma and Vega we only focus on our attention on
Vega?
• Drop the multiple position force constraint introduced above ? The cost of the hedge
portfolio however may go up.
234