TRADING VOLATILITY

Trading Volatility, Correlation,

Term Structure and Skew

Colin Bennett
 Copyright © 2014 Colin Bennett

All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means – electronic, mechanical, photocopying,
recording or otherwise without written permission of the publisher or author.

Trading Volatility: Trading Volatility, Correlation, Term Structure and Skew

ISBN-13: 978-1499206074

Cover Design by Gareth Allen

 ACKNOWLEDGEMENTS
I would like to thank James Peattie for persuading me to work at Merrill Lynch, and starting
my career within research which I still enjoy to this very day.

Iain Clamp aka “The Guru” deserves special recognition, for explaining the intricacies of
volatility trading.

I will always be grateful to Tom Dauterman and Irene Ferrero for their many months of
effort proofing this publication. I would also like to thank Mariano Colmenar, without
whose support this book would never have been published.

And finally, and most importantly, I would like to thank my wife Claire for her endless
patience and understanding with me while I researched and wrote this book over the past 5
years.
 NOTE ON CONTENTS
While there are many different aspects to volatility trading, not all of them are suitable for all
investors. In order to allow easy navigation, the sections are combined into seven chapters
that are likely to appeal to different parts of the equity derivatives client base. The earlier
chapters are most suited to equity investors, while later chapters are aimed at hedge funds
and proprietary trading desks. The Appendix contains reference material and mathematical
detail that has been removed from earlier chapters to enhance readability.
 CONTENTS

PREFACE ........................................................................................................... i

CHAPTER 1
OPTIONS................................................................................................................. 1
1.1: Option basics ........................................................................................................................ 2
1.2: Option trading in practice .................................................................................................... 4
1.3: Maintenance of option positions ..................................................................................... 11
1.4: Call overwriting ................................................................................................................... 15
1.5: Protection strategies using options .................................................................................. 23
1.6: Option structures trading .................................................................................................. 31

CHAPTER 2
VOLATILITY TRADING ..................................................................................... 35
2.1: Volatility trading using options ........................................................................................ 36
2.2: Variance is the key, not volatility ..................................................................................... 49
2.3: Volatility, variance and gamma swaps ............................................................................. 52
2.4: Options on variance ........................................................................................................... 68

CHAPTER 3
WHY EVERYTHING YOU THINK YOU KNOW ABOUT VOLATILITY
IS WRONG ............................................................................................................. 73
3.1: Implied volatility should be above realized volatility .................................................... 74
3.2: Long volatility is a poor hedge ......................................................................................... 77
3.3: Variable annuity hedging lifts long-term vol .................................................................... 81
3.4: Structured products vicious circle.................................................................................... 83
3.5: Stretching black-scholes assumptions ............................................................................. 89

CHAPTER 4
FORWARD STARTING PRODUCTS AND VOLATILITY INDICES .............. 99
4.1: Forward starting products ............................................................................................... 100
4.2: Volatility indices................................................................................................................ 109
4.3: Futures on volatility indices ............................................................................................ 115
4.4: Volatility future ETN/ETF ............................................................................................ 122
4.5: Options on volatility futures ........................................................................................ 129
CHAPTER 5
LIGHT EXOTICS................................................................................................. 133
5.1: Barrier options .................................................................................................................. 134
5.2: Worst-of/best-of options................................................................................................ 140
5.3: Outperformance options ................................................................................................. 143
5.4: Look-back options ........................................................................................................... 146
5.5: Contingent premium options ......................................................................................... 148
5.6: Composite and quanto options ...................................................................................... 149

CHAPTER 6
RELATIVE VALUE AND CORRELATION TRADING ................................... 153
6.1: Relative value trading ....................................................................................................... 154
6.2: Relative value volatility trading ....................................................................................... 158
6.3: Correlation trading ........................................................................................................... 161
6.4: Trading earnings announcements/jumps ..................................................................... 178

CHAPTER 7
SKEW AND TERM STRUCTURE TRADING................................................... 183
7.1: Skew and term structure are linked ............................................................................... 184
7.2: Square root of time rule can compare term structures and skews............................ 193
7.3: Term structure trading ..................................................................................................... 198
7.4: How to measure skew and smile.................................................................................... 201
7.5: Skew trading ...................................................................................................................... 210

APPENDIX
A.1: Local volatility .................................................................................................................. 230
A.2: Measuring historical volatility ........................................................................................ 232
A.3: Proof implied jump formula .......................................................................................... 245
A.4: Proof var swaps can be hedged by log contract (=1/k2) .......................................... 248
A.5: Proof variance swap notional = vega/2 σ ................................................................ 250
A.6: Modelling volatility surfaces........................................................................................... 251
A.7: Black-scholes formula ..................................................................................................... 255
A.8: Greeks and their meaning .............................................................................................. 257
A.9: Advanced (practical or shadow) greeks ......................................................................... 262
A.10: Shorting stock by borrowing shares ........................................................................... 266
A.11: Sortino ratio.................................................................................................................... 269
A.12: Capital structure arbitrage ............................................................................................ 270
INDEX .................................................................................................................... 286
 PREFACE

One of the main reasons I decided to write this book, was due to the lack of other
publications that deal with the practical issues of using derivatives. This publication aims to
fill the void between books providing an introduction to derivatives, and advanced books
whose target audience are members of quantitative modelling community.

In order to appeal to the widest audience, this publication tries to assume the least amount
of prior knowledge. The content quickly moves onto more advanced subjects in order to
concentrate on more practical and advanced topics.
CHAPTER 1
OPTIONS

This chapter is focused on real life uses of options for directional investors, for example
using options to replace a long position in the underlying, to enhance the yield of a
position through call overwriting, or to provide protection from declines. In addition to
explaining these strategies, a methodology to choose an appropriate strike and expiry
is shown. Answers to the most common questions are given, such as when an investor
should convert an option before maturity, and the difference between delta and the
probability that an option expires in the money.
2 CHAPTER 1: OPTIONS

1.1: OPTION BASICS

This section introduces options and the history of options trading. The definition of
call and put options, and how they can be used to gain long or short equity exposure,
is explained. Key definitions and terminology are given, including strike, expiry,
intrinsic value, time value, ATM, OTM and ITM.

HISTORY OF VOLATILITY TRADING

While standardised exchange traded options only started trading in 1973 when the CBOE
(Chicago Board Options Exchange) opened, options were first traded in London from 1690.
Pricing was made easier by the Black-Scholes-Merton formula (usually shortened to Black-
Scholes), which was invented in 1970 by Fischer Black, Myron Scholes and Robert Merton.

Option trading exploded in the 1990s

The derivatives explosion in the 1990s was partly due to the increasing popularity of hedge
funds, which led to volatility becoming an asset class in its own right. New volatility
products such as volatility swaps and variance swaps were created, and a decade later futures
on volatility indices gave investors listed instruments to trade volatility. In this chapter we
shall concentrate on option trading.

CALL OPTIONS GIVE RIGHT TO BUY, PUTS RIGHT TO SELL

A European call is a contract that gives the investor the right (but not the obligation) to buy
a security at a certain strike price on a certain expiry date (American options can be exercised
before expiry). A put is identical except it is the right to sell the security.

Call option gives long exposure, put options give short exposure
A call option profits when markets rise (as exercising the call means the investor can buy the
underlying security cheaper than it is trading, and then sell it at a profit). A put option profits
when markets fall (as you can buy the underlying security for less, exercise the put and sell
the security for a profit). Options therefore allow investors to put on long (profit when
prices rise) or short (profit when prices fall) strategies.
1.1: Option Basics 3

SELLING OPTIONS GIVES OPPOSITE EXPOSURE

As a call option gives long exposure to the underlying security, selling a call option results in
short exposure to the underlying security. Similarly while a put option is a bearish (profits
from decline in the underlying) strategy, selling a put option is a bullish strategy (profits from
a rise in the underlying). While the direction of the underlying is the primary driver of profits
and losses from buying or selling options, the volatility of the underlying is also a driver.

OPTIONS TRADING GIVES VOLATILITY EXPOSURE

If the volatility of an underlying is zero, then the price will not move and an option’s payout.
is equal to the intrinsic value. Intrinsic value is the greater of zero and the ‘spot – strike price’
for a call and is the greater of zero and ‘strike price – spot’ for a put. Assuming that stock
prices can move, the value of a call and put will be greater than intrinsic due to the time
value (price of option = intrinsic value + time value). If an option strike is equal to spot (or
is the nearest listed strike to spot) it is called at-the-money (ATM).

As volatility increases so does the price of call and put options

If volatility is zero, an ATM option has a price of zero (as intrinsic is zero). However, if we
assume a stock is €50 and has a 50% chance of falling to €40 and 50% chance of rising to
€60, it has a volatility above zero. In this example, an ATM call option with strike €50 has a
50% chance of making €10 (if the price rises to €60 the call can be exercised to buy the stock
at €50, which can be sold for €10 profit). The fair value of the ATM option is therefore €5
(50% × €10); hence, as volatility rises the value of a call rises (a similar argument can be used
for puts).

Options have greatest time value when strike is similar to spot (i.e. ATM)
An ATM option has the greatest time value (the amount the option price is above the
intrinsic value). This can be seen in the same example by looking at an out-the-money
(OTM) call option of strike €60 (an OTM option has strike far away from spot and zero
intrinsic value). This OTM €60 call option would be worth zero, as the stock in this example
cannot rise above €60.

ITM options trade less than OTM options as they are more expensive
An in-the-money (ITM) option is one which has a strike far away from spot and positive
intrinsic value. Due to the positive intrinsic value ITM options are relatively expensive,
hence tend to trade less than their cheaper OTM counterparts.
4 CHAPTER 1: OPTIONS

1.2: OPTION TRADING IN PRACTICE

Using options to invest has many advantages over investing in cash equity. Options
provide leverage and an ability to take a view on volatility as well as equity direction.
However, investing in options is more complicated than investing in equity, as a strike
and expiry need to be chosen. This section explains hidden risks, (e.g. dividends) and
other practical aspects of option trading such as how to choose the strike, and the
difference between delta and the probability an option ends up ITM.

CHOOSING EXPIRY IS THE MOST DIFFICULT DECISION

The biggest difference between using options and cash equities (or delta 1 products) to gain
equity exposure is the fact a suitable expiry has to be chosen. Determining if an equity is cheap
or expensive is often easier than determining the driver and timing of the likely increase /
decrease.

Choosing a far dated expiry gives most opportunity for the expected correction, however far
dated options are very expensive. Conversely if a cheaper near dated expiry is chosen, there is
little time for the anticipated movement to occur. Usually key dates such as quarterly reporting
or elections help determine a suitable expiry.

Expiry choice enforces investor discipline

Having to choose an expiry can be seen as a disadvantage of option trading, but some
investors see it as an advantage as it enforces investor discipline. The process of choosing an
expiry focuses attention on the likely dates a stock will converge with a forecast target price.

If the stock performs as expected the process of expiration forces the profits on the option
position to be taken, and ensures a position is not held longer than it should be. Additionally if
the anticipated return has not occurred by the expected date, the position expires worthless
and forces the investor to make a decision if another position should be initiated. Using
options to gain equity exposure therefore prevents “inertia” in a portfolio.

Both an equity and volatility view is needed to trade options

Option trading allows a view on equity. and volatility markets to be taken. If implied volatility
is seen to be expensive then a short volatility strategy is best (short put for a bullish strategy,
short call for a bearish strategy). However if implied volatility is seen to be cheap then a long
volatility strategy is best (long call for a bullish strategy, long put for a bearish strategy). The
appropriate strategy for a one leg option trade is shown in Figure 1 below. For multiple leg
strategies see the section 1.6 Option Structures Trading.
1.2: Option Trading in Practice 5

Figure 1. Option Strategy for Different Market and Volatility Views

MARKET VIEW

VOLATILITY Bearish Bullish

VIEW
Short call Short put

5 5
Volatility high 0 0
90 100 110 90 100 110
-5 -5

Long put Long call

5 5
Volatility low
0 0
90 100 110 90 100 110
-5 -5

Long vol strategies should have expiry just after key date
Typically if a key date is likely to be volatile then a long volatility strategy (long call or long
put) should have an expiry just after this date. Conversely a short volatility strategy (short call
or short put) should have an expiry just before the key date. For investors who wish to trade
the implied jump of a key date, details of how to trade this implied jump is dealt with in the
section 6.4: Trading Earnings Announcements/Jumps.

CHOOSING STRIKE OF STRATEGY IS NOT TRIVIAL

While choosing the strike of a strategy is not as difficult as choosing the expiry, it is not
trivial. Investors could choose ATM to benefit from greatest liquidity. Alternatively, they
could look at the highest expected return (option payout less the premium paid, as a
percentage of the premium paid).
1.2: Option Trading in Practice 7

OTM options have highest return for “abnormal” moves

Only if the expected return is relatively high (or abnormal) do ATM or OTM options have
the highest return. This is because for exceptional returns their low cost and high leverage
more than compensates for their lower delta.

LIQUIDITY CAN BE A FACTOR IN CHOOSING STRIKE

If an underlying is relatively illiquid, or if the size of the trade is large, an investor should take
into account the liquidity of the maturity and strike of the option. Typically, OTM options
are more liquid than ITM options as ITM options tie up a lot of capital. This means that for
strikes less than spot, puts are more liquid than calls and vice versa.

Low strike puts are usually more liquid than high strike calls
We note that as low-strike puts have a higher implied than high-strike calls, their value is
greater and, hence, traders are more willing to use them. Low strike put options are therefore
usually more liquid than high-strike call options. In addition, demand for protection lifts
liquidity for low strikes compared with high strikes.

Single stock liquidity is limited for maturities up to two years

For single stock options, liquidity starts to fade after one year and options rarely trade over
two years. For indices, longer maturities are liquid, partly due to the demand for long-dated
hedges and their use in structured products. While structured products can have a maturity
of five to ten years, investors typically lose interest after a few years and sell the product
back. The hedging of a structured product, therefore, tends to be focused on more liquid
maturities of around three years.

Hedge funds and structured product flow can overlap

Hedge funds tend to focus around the one-year maturity, with two to three years being the
longest maturity they will consider. The two-to-three year maturity is where there is greatest
overlap between hedge funds and structured desks.

DELTA MEASURES DIVIDEND RISK AND EQUITY RISK

The delta of the option is the amount of equity market exposure an option has. As a stock
price falls by the dividend amount on its ex-date, delta is equal to the exposure to dividends
that go ex before expiry. The dividend risk is equal to the negative of the delta. For example,
if you have a call of positive delta, if (expected or actual) dividends rise, the call is worth less
(as the stock falls by the dividend amount).

If a dividend is substantial, it could be in an investor’s interest to exercise early. For more

details, see the section 1.3 Maintenance of Option Positions.
8 CHAPTER 1: OPTIONS

DELTA IS NOT THE PROBABILITY OPTION EXPIRES ITM

A digital call option is an option that pays 100% if spot expires above the strike price (a
digital put pays 100% if spot is below the strike price). The probability of such an option
expiring ITM is equal to its delta, as the payoff only depends on it being ITM or not (the size
of the payment does not change with how much ITM spot is). For a vanilla option this is not
the case; hence, there is a difference between the delta and the probability of being ITM.
This difference is typically small unless the maturity of the option is very long.

Delta takes into account the amount an option can be ITM

While a call can have an infinite payoff, a put’s maximum value is the strike (as spot cannot
go below zero). The delta hedge for the option has to take this into account, so a call delta
must be greater than the probability of being ITM. Similarly, the absolute value (as put deltas
are negative) of the put delta must be less than the probability of expiring ITM. A more
mathematical explanation (for European options) is given below:

Call delta > Probability call ends up ITM

Abs (Put delta < Probability put ends up ITM

Mathematical proof option delta is different from probability of being ITM at

expiry
Call delta = N(d 1 ) Put delta = N(d 1 ) - 1

Call probability ITM = N(d 2 ) Put probability ITM = 1 - N(d 2 )

where:

Definition of d 1 is the standard Black-Scholes formula for d 1. For more details, see the
section A.7 Black-Scholes Formula.

d2 = d1 - σ T

σ = implied volatility

T = time to expiry

N(z) = cumulative normal distribution

10 CHAPTER 1: OPTIONS

Put underwriting benefits from selling expensive implied

Typically the implied volatility of options trades slightly above the expected realised volatility
of the underlying over the life of the option (due to a mismatch between supply and
demand). Stock replacement via put selling therefore benefits from selling (on average)
expensive volatility. Selling a naked put is known as put underwriting, as the investor has
effectively underwritten the stock (in the same way investment banks underwrite a rights
issue).

Put underwriting pays investors for work that otherwise might be wasted
The strike of put underwriting should be chosen at the highest level at which the investor
would wish to purchase the stock, which allows an investor to earn a premium from taking
this view (whereas normally the work done to establish an attractive entry point would be
wasted if the stock did not fall to that level).

Asset allocators use put underwriting to rebalance portfolios

This strategy has been used significantly recently by asset allocators who are underweight
equities and are waiting for a better entry point to re-enter the equity market (earning the
premium provides a buffer should equities rally). If an investor does not wish to own the
stock and only wants to earn the premium, then an OTM strike should be chosen at a
support level that is likely to remain firm.

Put underwriting benefits from selling skew

Put underwriting gives a similar profile to a long stock, short call profile, otherwise known as
call overwriting. One difference between call overwriting and put underwriting is that if
OTM options are used, then put underwriting benefits from selling skew (which is normally
overpriced). For more details on the benefits of selling volatility, see the section 1.4 Call
Overwriting.

STOCK REPLACEMENT ALTERS DIVIDEND EXPOSURE

It is important to note that the dividend exposure is not the same, as only the owner of a
stock receives dividends. While the option owner does not benefit directly, the expected
dividend will be used to price the option fairly (hence investors only suffer/benefit if
dividends are different from expectations).
1.3: Maintenance of Option Positions 11

1.3: MAINTENANCE OF OPTION POSITIONS

During the life of an American option many events can occur in which it might be
preferable to own the underlying shares (rather than the option) and exercise early.
In addition to dividends, an investor might want the voting rights, or alternatively
might want to sell the option to purchase another option (rolling the option). We
investigate these life cycle events and explain when it is in an investor’s interest to
exercise, or roll, an option before expiry.

CONVERTING OPTIONS EARLY IS RARE

Options on indices are usually European, which means they can only be exercised at
maturity. The inclusion of automatic exercise, and the fact it is impossible to exercise before
maturity, means European options require only minimal maintenance. Single stock options,
however, are typically American (apart from emerging market underlyings). While American
options are rarely exercised early, there are circumstances when it is in an investor’s interest
to exercise an ITM option early. For both calls and puts the correct decision for early
exercise depends on the net benefit of doing so (ie, the difference between earning the
interest on the strike and net present value of dividends) versus the time value of the option.

 Calls should be exercised just before the ex-date of a large unadjusted dividend.
In order to exercise a call, the strike price needs to be paid. The interest on this strike
price normally makes it unattractive to exercise early. However, if there is a large
unadjusted dividend that goes ex before expiry, it might be in an investor’s interest to
exercise an ITM option early (see Figure 4 above). In this case, the time value should be
less than the dividend NPV (net present value) less total interest r (=erfr×T-1) earned on
the strike price K. In order to maximise ‘dividend NPV– Kr’, it is best to exercise just
before an ex-date (as this maximises ‘dividend NPV’ and minimises the total interest r).
 Puts should be exercised early (preferably just after ex-date) if interest rates are
high. If interest rates are high, then the interest r from putting the stock back at a high
strike price K (less dividend NPV) might be greater than the time value. In this case, a
put should be exercised early. In order to maximise ‘Kr – dividend NPV’, a put should
preferably be exercised just after an ex-date.
1.3: Maintenance of Option Positions 13

Puts should only be exercised if interest earned (less dividends) exceeds time
value
For puts, it is simplest to assume an investor is long stock and long an American put. This
payout is similar to a long call of the same strike. An American put should only be exercised
against the long stock in the same portfolio if it is in an investor’s interest to exercise the
option and buy a European call of the same strike.

 Choice A: Do not exercise. In this case the portfolio of long stock and long put benefits
from the dividend NPV.
 Choice B: Exercise put against long stock, receiving strike K, which can earn interest r
(=erfr×T-1). The position has to be hedged with the purchase of a European call (of cost
equal to the time value of a European put).
An investor will only exercise early if choice B > choice A

 Kr – time value > dividend NPV

 Kr – dividend NPV > time value for American put to be exercised
Selling ITM options that should be exercised early can be profitable
There have been occasions when traders deliberately sell ITM options that should be
exercised early, hoping that some investors will forget. Even if the original counterparty is
aware of this fact, exchanges randomly assign the counterparty to exercised options. As it is
unlikely that 100% of investors will realise in time, such a strategy can be profitable.

ITM OPTIONS USUALLY EXERCISED AUTOMATICALLY

In order to prevent situations where an investor might suffer a loss if they do not give notice
to exercise an ITM option in time, most exchanges have some form of automatic exercise. If
an investor (for whatever reason) does not want the option to be automatically exercised, he
must give instructions to that effect. The hurdle for automatic exercise is usually above ATM
in order to account for any trading fees that might be incurred in selling the underlying post
exercise.
Eurex automatic exercise has a higher hurdle than CBOE
For the CBOE, options are automatically exercised if they are US$0.01 or more ITM
(reduced in June 2008 from US$0.05 or more), which is in line with Euronext-Liffe rules of a
€0.01 or GBP0.01 minimum ITM hurdle. Eurex has a higher automatic hurdle, as a contract
price has to be ITM by 99.99 or more (eg, for a euro-denominated stock with a contract size
of 100 shares this means it needs to be at €0.9999 or more). Eurex does allow an investor to
specify an automatic exercise level lower than the automatic hurdle, or a percentage of
exercise price up to 9.99%.
14 CHAPTER 1: OPTIONS

CORPORATE ACTIONS CAN ADJUST STRIKE

While options do not adjust for ordinary dividends 1, they do adjust for special dividends.
Different exchanges have different definitions of what is a special dividend, but typically it is
considered special if it is declared as a special dividend, or is larger than a certain threshold
(eg, 10% of the stock price). In addition, options are adjusted in the event of a corporate
action, for example, a stock split or rights issue.

Equities and indices can treat bonus share issues differently

Options on equities and indices can treat bonus share issues differently. A stock dividend in
lieu of an ordinary dividend is considered an ordinary dividend for options on an equity
(hence is not adjusted) but is normally adjusted by the index provider.

Adjustment negates impact of dividend or corporate action

For both special dividends and corporate actions, the adjustment negates the impact of the
event (principal of unchanged contract values), so the theoretical price of the options should
be able to ignore the event. As the strike post adjustment will be non-standard, typically
exchanges create a new set of options with the normal strikes. While older options can still
trade, liquidity generally passes to the new standard strike options (particularly for longer
maturities which do not have much open interest).

M&A AND SPINOFFS CAN CAUSE PROBLEMS

If a company spins off a subsidiary and gives shareholders shares in the new company, the
underlying for the option turns into a basket of the original equity and the spun-off
company. New options going into the original company are usually created, and the liquidity
of the options into the basket is likely to fade. For a company that is taken over, the existing
options in that company will convert into whatever shareholders were offered. If the
acquisition was for stock, then the options convert into shares, but if the offer is partly in
cash, then options can lose a lot of value as the volatility of cash is zero.

OPTIONS OFTEN ROLLED BEFORE EXPIRY

The time value of an option decays quicker for short-dated options than for far-dated
options. To reduce the effect of time decay, investors often roll before expiry. For example,
an investor could buy a one-year option and roll it after six months to a new one-year
option.

1 Some option markets adjust for all dividends.

1.4: Call Overwriting 15

1.4: CALL OVERWRITING

For a directional investor who owns a stock (or index), call overwriting by selling an
OTM call is one of the most popular methods of yield enhancement. Historically,
call overwriting (otherwise known as buy-write, as the stock is bought but a call is
written against it) has been a profitable strategy due to implied volatility usually
being overpriced. However, call overwriting does underperform in volatile, strongly
rising equity markets. Overwriting with the shortest maturity is best, and the strike
should be slightly OTM for optimum returns.

IMPLIED VOLATILITY IS USUALLY OVERPRICED

The implied volatility of options is on average 1-2pts above the volatility realised over the
life of the option. This ‘implied volatility premium’ is usually greater for indices than for
single stocks. As we can see no reason why these imbalances will fade, we expect call
overwriting to continue to outperform on average. The key imbalances are:

 Option buying for protection. In the same way that no one buys car insurance because
they think it is a good investment, investors are happy buying expensive protection to
protect against downside risks.
 Unwillingness to sell low premium options causes market makers to raise their
prices (selling low premium options, like selling lottery tickets, has to be done on a large
scale to be attractive).
 High gamma of near-dated options has a gap risk premium (risk of stock jumping,
either intraday or between closing and opening prices).
 Index implieds lifted by structured products. Structured products are often based on
an index, and can offer downside protection. This lifts index implied relative to single
stock implied. Also protection is usually bought on an index to protect against macros
risks. It is rare to protect a single stock position (if an investor is worried about
downside in a stock they usually do not buy it to begin with).

BUY-WRITE BENEFITS FROM SELLING EXPENSIVE VOL

Short-dated implied volatility has historically been overpriced 2 due to the above supply and
demand imbalances. In order to profit from this characteristic, a long investor can sell a call
against a long position in the underlying of the option. Should the underlying perform well
and the call be exercised, the underlying can be used to satisfy the exercise of the call. As
equities should be assumed to have, on average, a positive return, it is best to overwrite with
a slightly OTM option to reduce the probability of the option sold expiring ITM.

2 We note that implied volatility is not necessarily as overpriced as would first appear. For more detail,
see the section 3.1 Implied Volatility Should Be Above Realized Volatility.
1.4: Call Overwriting 21

at a period of time where the SX5E had a positive return shows that for one-month options
a strike between 103%-104% is best (see Figure 10 above).

Typically call overwriting with c25% delta call options is best

For three-month options, the optimal strike is a higher 107%-108%, but the outperformance
is approximately half as good as for one-month options. These optimal strikes for
overwriting could be seen to be arguably high, as recently there have been instances of
severe declines (TMT bubble bursting, Lehman bankruptcy), which were followed by
significant price rises afterwards. For single-stock call overwriting, these strikes could seem
to be low, as single stocks are more volatile. For this reason, many investors use the current
level of volatility to determine the strike or choose a fixed delta option (eg, 25%).

OVERWRITING REDUCES VOLATILITY

While selling an option could be considered risky, the volatility of returns from overwriting a
long equity position is reduced by overwriting. This is because the payout profile is capped
for equity prices above the strike. An alternative way of looking at this is that the delta of the
portfolio is reduced from 100% (solely invested in equity) to 100% less the delta of the call
(c50% depending on strike). The reduced delta suppresses the volatility of the portfolio.

Risk reduction less impressive if Sortino ratios are used

We note that the low call overwriting volatility is due to the lack of volatility to the upside, as
call overwriting has the same downside risk as a long position. For this reason, using the
Sortino ratio (for more details, see the section A11 Sortino Ratio in the Appendix) is likely to be
a fairer measure of call overwriting risk than standard deviation, as standard deviation is not
a good measure of risk for skewed distributions. Sortino ratios show that the call overwriting
downside risk is identical to a long position; hence, call overwriting should primarily be done
to enhance returns and is not a viable strategy for risk reduction.

Optimal strike is similar for single stocks and indices

While this analysis is focused on the SX5E, the analysis can be used to guide single-stock call
overwriting (although the strike could be adjusted higher by the single-stock implied divided
by SX5E implied).
22 CHAPTER 1: OPTIONS

ENHANCED CALL OVERWRITING IS DIFFICULT

Enhanced call overwriting is the term given when call overwriting is only done
opportunistically or the parameters (strike or expiry) are varied according to market
conditions. On the index level, the returns from call overwriting are so high that enhanced
call overwriting is difficult, as the opportunity cost from not always overwriting is too high.
For single stocks, the returns for call overwriting are less impressive; hence, enhanced call
overwriting could be more successful. An example of single-stock enhanced call overwriting
is to only overwrite when an implied is high compared to peers in the same sector. We note
that even with enhanced single-stock call overwriting, the wider bid-offer cost and smaller
implied volatility premium to realised means returns can be lower than call overwriting at the
index level.

Enhanced call overwriting returns likely to be arbitraged away

Should a systematic way to enhance call overwriting be viable, this method could be applied
to volatility trading without needing an existing long position in the underlying. Given the
presence of statistical arbitrage funds and high frequency traders, we believe it is unlikely that
a simple automated enhanced call overwriting strategy on equity or volatility markets is likely
to outperform vanilla call overwriting on an ongoing basis.
26 CHAPTER 1: OPTIONS

Implied volatility is far more important than skew for put-spread pricing
A rule of thumb is that the value of the OTM put sold should be approximately one-third
the value of the long put (if it were significantly less, the cost saving in moving from a put to
a put spread would not compensate for giving up complete protection). While selling an
OTM put against a near-ATM put does benefit from selling skew (as the implied volatility of
the OTM put sold is higher than the volatility of the near ATM long put bought), the effect
of skew on put spread pricing is not normally that significant (far more significant is the level
of implied volatility).

Collars are more sensitive to skew than implied volatility

Selling a call against a long put suffers from buying skew. The effect of skew is greater for a
collar than for a put spread, as skew affects both legs of the structure the same way (whereas
the effect of skew on the long and short put of a put spread partly cancels). If skew was flat,
the cost of a collar typically reduces by 1% of spot. The level of volatility for near-zero cost
collars is not normally significant, as the long volatility of the put cancels the short volatility
of the call.

Capping performance should only be used when a long rally is unlikely

A collar or put spread collar caps the performance of the portfolio at the strike of the OTM
call sold. They should only therefore be used when the likelihood of a strong, long-lasting
rally (or significant bounce) is perceived to be relatively small.

Bullish investors could sell two puts against long put

If an investor is bullish on the equity market, then a protection strategy that caps
performance is unsuitable. Additionally, as the likelihood of substantial declines is seen to be
small, the cost of protection via a put or put spread is too high. In this scenario, a zero cost
1×2 put spread could be used as a pseudo-protection strategy. The long put is normally
ATM, which means the portfolio is 100% protected against falls up to the lower strike, and
gives partial protection below that until the breakeven. A loss is only suffered if the equity
market falls below the breakeven.

1×2 put spreads only give pseudo-protection

We do not consider 1x2 put spreads to offer true protection, as during severe declines it will
suffer a loss when the underlying portfolio is also heavily loss making. The payout of 1×2
put spreads for maturities of around three months or more is initially similar to a short put,
so we consider it to be a bullish strategy. However, for the SX5E a roughly six-month zero-
cost 1×2 put spread, whose upper strike is 95%, has historically had a breakeven below 80%
and declines of more than 20% in six months are very rare. As 1×2 put spreads do not
provide protection when you need it most, they could be seen as a separate long position
rather than a protection strategy.
1.5: Protection Strategies Using Options 29

Short-dated puts offer greatest protection but highest cost

If equity markets fall 20% in the first three months of the year and recover to the earlier
level by the end of the year, then a rolling three-month put strategy will have a positive
payout in the first quarter but a one-year put will be worth nothing at expiry. While rolling
near-dated puts will give greater protection than a long-dated put, the cost is higher
(see Figure 17 above).

SHORT VOL AGAINST LONG PUT PERFORMS WELL

All protection strategies that combine a long and a short aim to offset the overpriced cost of
protection by selling the same overpriced implied volatility for a different maturity and
strike. Hence such strategies tend to back-test well as the overall exposure to expensive
implied volatility is near zero. As the net cost of such strategies is near zero, while at the
same time (usually) decreasing the volatility of the portfolio, their risk adjusted performance
can be impressive. Their performance can often be further improved by selling near dated
volatility against the long far dated protection.

MULTIPLE EXPIRY PROTECTION STRATEGIES

Typically, a protection strategy involving multiple options has the same maturity for all of
the options. However, some investors choose a nearer maturity for the options they are
short, as more premium can be earned selling a near-dated option multiple times (as near-
dated options have higher theta). These strategies are most successful when term structure is
inverted, as the volatility for the near-dated option sold is higher. Having a nearer maturity
for the long put option and longer maturity for the short options makes less sense, as this
increases the cost (assuming the nearer-dated put is rolled at expiry).

Calendar collar effectively overlays call overwriting on a long put position

If the maturity of the short call of a collar is closer than the maturity of the long put, then
this is effectively the combination of a long put and call overwriting. For example, the cost
of a three-month put can be recovered by selling one-month calls. This strategy outperforms
in a downturn and also has a lower volatility (see Figure 18 below).
1.6: Option Structures Trading 31

1.6: OPTION STRUCTURES TRADING

While a simple view on both volatility and equity market direction can be
implemented via a long or short position in a call or put, a far wider set of payoffs is
possible if two or three different options are used. We investigate strategies using
option structures (or option combos) that can be used to meet different investor
needs.

BULLISH COMBOS ARE REVERSE OF BEARISH

Using option structures to implement a bearish strategy has already been discussed in the
section 1.5 Protection Strategies Using Options. In the same way a long put protection can be
cheapened by selling an OTM put against the put protection (to create a put spread giving
only partial protection), a call can be cheapened by selling an OTM call (to create a call
spread offering only partial upside). Similarly, the upside exposure of the call (or call spread)
can be funded by put underwriting (just as put or put spread protection can be funded by
call overwriting). The four option structures for bullish strategies are given below.

 Calls give complete upside exposure and floored downside. Calls are the ideal
instrument for bullish investors as they offer full upside exposure and the maximum loss
is only the premium paid. Unless the call is short dated or is purchased in a period of
low volatility, the cost is likely to be high.
 Call spreads give partial upside but are cheaper. If an underlying is seen as unlikely
to rise significantly, or if a call is too expensive, then selling an OTM call against the
long call (to create a call spread) could be the best bullish strategy. The strike of the call
sold could be chosen to be in line with a target price or technical resistance level. While
the upside is limited to the difference between the two strikes, the cost of the strategy is
normally one-third cheaper than the cost of the call.
 Risk reversals (short put, long call of different strikes) benefit from selling skew.
If a long call position is funded by selling a put (to create a risk reversal), the volatility of
the put sold is normally higher than the volatility of the call bought. The higher skew is,
the larger this difference and the more attractive this strategy is. Similarly, if interest rates
are low, then the lower the forward (which lifts the value of the put and decreases the
value of the call) and the more attractive the strategy is. The profile of this risk reversal
is similar to being long the underlying.
 Call spread vs put is most attractive when volatility is high. A long call can be
funded by selling an OTM call and OTM put. This strategy is best when implied
volatility is high, as two options are sold.
CHAPTER 2
VOLATILITY TRADING

We investigate the benefits and disadvantages of volatility trading via options,

volatility swaps, variance swaps and gamma swaps. We also show how these products,
correlation swaps, basket options and covariance swaps can give correlation exposure.
Recently, options on alternative underlyings have been created, such as options on
variance and options on volatility futures. We show how the distribution and skew
for options on variance is different from those for equities (options on volatility futures
have many similarities to options on variance but are slightly different, and are covered
in the section 4.5: Options On Volatility Futures).
2.1: Volatility Trading Using Options 41

Long gamma position can sit on the bid and offer

As shown above, a long gamma (long volatility) position has to buy shares if they fall, and
sell them if they rise. Buying low and selling high earns the investor a profit. Additionally, as
a gamma scalper can enter bids and offers away from current spot, there is no need to cross
the spread (as a long gamma position can be delta hedged by sitting on the bid and offer).

Short gamma position have to cross the bid-offer spread

A short gamma position represents the reverse situation, and requires crossing the spread to
delta hedge. While this hidden cost is small, it could be substantial over the long term for
underlyings with relatively wide bid-offer spreads.

Best to delta hedge on key dates or on turn of market

If markets have a clear direction (ie, they are trending), it is best to delta hedge less
frequently. However, in choppy markets that are range bound it is best to delta hedge very
frequently. For more detail on how hedging frequency affects returns and the path
dependency of returns, see the section 3.5 Stretching Black-Scholes Assumptions. If there is a key
announcement (either economic or earnings-related to affect the underlying), it is best to
delta hedge just before the announcement to ensure that profit is earned from any jump (up
or down) that occurs.

OPTION TRADING RULES OF THUMB

To calculate option premiums and volatility exactly is typically too difficult to do without the
aid of a tool. However, there are some useful rules of thumb that can be used to give an
estimate. These are a useful sanity check in case an input to a pricing model has been entered
incorrectly.

 Historical volatility roughly equal to 16 × percentage daily move. Historical

volatility can be estimated by multiplying the typical return over a period by the square
root of the number of periods in a year (eg, 52 weeks or 12 months in a year). Hence, if
a security moves 1% a day, it has an annualised volatility of 16% (as there are c252
trading days and 16 ≈ √252).
 ATM option premium in percent is roughly 0.4 × volatility × square root of time.
If one assumes zero interest rates and dividends, then the formula for the premium of
an ATM call or put option simplifies to 0.4 × σ × √T. Therefore, a one-year ATM
option on an underlying with 20% implied is worth c8% (= 0.4 × 20% × √1). OTM
options can be calculated from this estimate using an estimated 50% delta.
 Profit from delta hedging is proportional to square of return. Due to the convexity
of an option, if the volatility is doubled the profits from delta hedging are multiplied by a
factor of four. For this reason, variance (squared returns) is a better measure of
deviation than volatility.
42 CHAPTER 2: VOLATILITY TRADING

 Difference between implied and realised determines carry. While variance is the
driver of profits if the implied volatility of an option is constant, the carry is determined
by the absolute difference between realised and implied (ie, the same carry is earned by
going long a 20% implied option that realises 21% as by going long a 40% implied
option that realises 41%.

ANNUALISED VOL = 16 × PERCENTAGE DAILY MOVE

Volatility is defined as the annualised standard deviation of log returns (where return = P i / P i-
1 ). As returns are normally close to 1 (=100%) the log of returns is very similar to ‘return – 1’
(which is the percentage change of the price). Hence, to calculate the annualised volatility for
a given percentage move, all that is needed is to annualise the percentage change in the price.
This is done by multiplying the percentage move by the square root of the number of
samples in a year (as volatility is the square root of variance). For example, market
convention is to assume there are approximately 252 trading days a year. If a stock moves
1% a day, then its volatility is 1%×√252, which is approximately 1%×16 = 16% volatility.
Similarly, if a stock moves 2% a day it has 32% volatility.

Num of trading days in year = 252 => Multiply daily returns by √252 ≈ 16

Num of weeks in year = 52 => Multiply weekly returns by √52 ≈7

Num of months in year = 12 => Multiply monthly returns by √12 ≈ 3.5

HEDGING CAN ‘PIN’ A STOCK APPROACHING EXPIRY

As an investor who is long gamma can delta hedge by sitting on the bid and offer, this trade
can pin an underlying to the strike. This is a side effect of selling if the stock rises above the
strike, and buying if the stock falls below the strike. The amount of buying and selling has to
be significant compared with the traded volume of the underlying, which is why pinning
normally occurs for relatively illiquid stocks or where the position is particularly sizeable.
Given the high trading volume of indices, it is difficult to pin a major index. Pinning is more
likely to occur in relatively calm markets, where there is no strong trend to drive the stock
away from its pin.
44 CHAPTER 2: VOLATILITY TRADING

ATM OPTION PREMIUM (%) = 0.4 × VOLATILITY × √TIME

Call price = S N(d 1 ) – K N(d 2 ) e-rT

Assuming zero interest rates and dividends (r = 0)

 ATM call price = S N(σ × √T / 2) – S N(-σ × √T / 2) as K=S (as ATM)

 ATM call price = S × σ × √T / √(2π)
 ATM call price = σ × √T / √(2π) in percent
 ATM call price ≈ 0.4 × σ × √T in percent
where:

Definition of d 1 and d 2 is the standard Black-Scholes formula.

σ = implied volatility
S = spot
K = strike
R = interest rate
T = time to expiry
N(z) = cumulative normal distribution

Example 1

A 1 year ATM option on an underlying with 20% implied is worth c.8% (=0.4 × 20% × √1)

Example 2

A 3 month ATM option on an underlying with 20% implied is worth c.4% (=0.4 × 20% ×
√0.25 =0.4 × 20% × 0.5)

OTM options can be calculated by assuming 50% delta

If an index is 3000pts and has a 20% implied then the price of a 1Y ATM option is
approximately 240pts (3000×8% as calculated in Example 1 above). A 3200 call is therefore
approximately 240 – 50% (3200-3000) = 240 – 100 = 140pts assuming a 50% delta.
Similarly, a 3200 put is approximately 340pts (240 + 100).
2.1: Volatility Trading Using Options 45

PROFIT PROPORTIONAL TO PERCENT MOVE SQUARED

Due to the convexity of an option, if the volatility is doubled, the profits from delta hedging
are multiplied by a factor of four. For this reason, variance (which looks at squared returns)
is a better measure of deviation than volatility. Assuming constant volatility, zero interest
rates and dividend, the daily profit and loss (P&L) from delta hedging an option is given
below.

Daily P&L = Delta P&L + Gamma P&L + Theta P&L

 Daily P&L = Sδ + S2γ /2 + tθ

where S is change in Stock
 Daily P&L - Sδ = + S2γ /2 + tθ = Delta hedged P&L
 Delta hedged P&L = S2γ /2 + cost term

(tθ does not depend on stock price)

where:

δ = delta
γ = gamma
t = time
θ = theta

If the effect of theta is ignored (as it is a cost that does not depend on the size of the stock
price movement), the profit of a delta hedged option position is equal to a scaling factor
(gamma/2) multiplied by the square of the return. This means that the profit from a 2%
move in a stock price is four times (22=4) the profit from a 1% move in stock price.

This can also be seen from Figure 29 below, as the additional profit from the move from 1%
to 2% is three times the profit from 0% to 1% (for a total profit four times the profit for a
1% move).
48 CHAPTER 2: VOLATILITY TRADING

 Daily P&L ≈ constant × dt × ((σ2 + 2σx + x2) - σ2) / σ

 Daily P&L ≈ constant × dt × (2x + x2/σ)
 Daily P&L ≈ constant × dt × 2x as x is small
Daily P&L proportional to x, where x = realised volatility - σ

Hence, it is the difference between realised and implied volatility that is the key to daily P&L
(or carry).
2.2: Variance Is the Key, Not Volatility 49

2.2: VARIANCE IS THE KEY, NOT VOLATILITY

Partly due to its use in Black-Scholes, historically, volatility has been used as the
measure of deviation for financial assets. However, the correct measure of deviation
is variance (or volatility squared). Volatility should be considered to be a derivative of
variance. The realisation that variance should be used instead of volatility led
volatility indices, such as the VIX, to move away from ATM volatility (VXO index)
towards a variance-based calculation.

VAR, NOT VOL, IS CORRECT MEASURE FOR DEVIATION

There are three reasons why variance, not volatility, should be used as the correct measure
for volatility. However, despite these reasons, even variance swaps are normally quoted as
the square root of variance for an easier comparison with the implied volatility of options
(but we note that skew and convexity mean the fair price of variance should always trade
above ATM options).

 Variance takes into account implied volatility at all stock prices. Variance takes into
account the implied volatility of all strikes with the same expiry (while ATM implied will change
with spot, even if volatility surface does not change).
 Deviations need to be squared to avoid cancelling. Mathematically, if deviations were simply
summed then positive and negative deviations would cancel. This is why the sum of squared
deviations is taken (variance) to prevent the deviations from cancelling. Taking the square root of
this sum (volatility) should be considered a derivative of this pure measure of deviation
(variance).
 Profit from a delta-hedged option depends on the square of the return. Due to the
convexity of an option, if the volatility is doubled, the profits from delta hedging are multiplied
by a factor of four. For this reason, variance (which looks at squared returns) is a better measure
of deviation than volatility.
2.2: Variance Is the Key, Not Volatility 51

This can also be seen from Figure 29 Profile of a Delta-Hedged Option in the previous
section (page 46), as the additional profit from the move from 1% to 2% is three times the
profit from 0% to 1% (for a total profit four times the profit for a 1% move).

VOL SHOULD BE CONSIDERED A DERIVATIVE OF VAR

The three examples above show why variance is the natural measure for deviation. Volatility,
the square root of variance, should be considered a derivative of variance rather than a pure
measure of deviation. It is variance, not volatility, that is the second moment of a
distribution (the first moment is the forward or expected price). For more details on
moments, read the section 7.4 How to Measure Skew and Smile.

VIX CALCULATION MOVED FROM ATM VOL TO VAR

Due to the realisation that variance, not volatility, was the correct measure of deviation, on
Monday, September 22, 2003, the VIX index moved away from using ATM implied towards
a variance-based calculation. Variance-based calculations have also been used for by other
volatility index providers. The old VIX, renamed VXO, took the implied volatility for strikes
above and below spot for both calls and puts. As the first two-month expiries were used, the
old index was an average of eight implied volatility measures as 8 = 2 (strikes) × 2 (put/call)
× 2 (expiry). We note that the use of the first two expiries (excluding the front month if it
was less than eight calendar days) meant the maturity was on average 1.5 months, not one
month as for the new VIX.

VDAX calculation also moved from ATM vol to var

Similarly, the VDAX index, which was based on 45-day ATM-implied volatility, has been
superseded by the V1X index, which, like the new VIX, uses a variance swap calculation. All
recent volatility indices, such as the vStoxx (V2X), VSMI (V3X), VFTSE, VNKY and VHSI,
use a variance swap calculation, although we note the recent VIMEX index uses a similar
methodology to the old VIX (potentially due to illiquidity of OTM options on the Mexbol
index).

VARIANCE TERM STRUCTURE IS NOT ALWAYS FLAT

While average variance term structure should be flat in theory, in practice supply and
demand imbalances can impact variance term structure. The buying of protection at the long
end should mean that variance term structure is on average upward sloping, but in turbulent
markets it is usually inverted.
52 CHAPTER 2: VOLATILITY TRADING

2.3: VOLATILITY, VARIANCE AND GAMMA SWAPS

In theory, the profit and loss from delta hedging an option is fixed and is based
solely on the difference between the implied volatility of the option when it was
purchased and the realised volatility over the life of the option. In practice, with
discrete delta hedging and unknown future volatility, this is not the case, leading to
the creation of volatility, variance and gamma swaps. These products also remove
the need to continuously delta hedge, which can be very labour-intensive and
expensive. Until the credit crunch, variance swaps were the most liquid of the three,
but now volatility swaps are more popular for single stocks.

VOL, VAR & GAMMA SWAPS GIVE PURE VOL EXPOSURE

As spot moves away from the strike of an option the gamma decreases, and it becomes more
difficult to profit via delta hedging. Second-generation volatility products, such as volatility
swaps, variance swaps and gamma swaps, were created to give volatility exposure for all
levels of spot and also to avoid the overhead and cost of delta hedging. While volatility and
variance swaps have been traded since 1993, they became more popular post-1998, when
Russia defaulted on its debts and Long-Term Capital Management (LTCM) collapsed.

Variance and gamma swaps usually quoted in volatility terms

Variance and gamma swaps are normally quoted as the square root of variance. This allows
easier comparison with the options market.

VOLATILITY SWAP ≤ GAMMA SWAP ≤ VARIANCE SWAP

The square root of the variance strike is always above volatility swaps (and ATMf implied as
volatility swaps ≈ ATMf implied). This is due to the fact a variance swap payout is convex
(hence, will always be greater than or equal to volatility swap payout of identical vega, which
is explained later in the section). Only for the unrealistic case of no vol of vol (ie, future
volatility is constant and known) will the price of a volatility swap and variance swap (and
gamma swap) be the same4. The fair price of a gamma swap is between volatility swaps and
variance swaps.

4 A variance swap payout is based on cash return assuming zero mean, whereas a delta-hedged option
variance payout is based on a forward. Hence, a variance swap fair price will be slightly above a
constant and flat volatility surface if the drift is non-zero (as close-to-close cash returns will be lifted by
the drift).
2.3: Volatility, Variance and Gamma Swaps 53

 Volatility swaps. Volatility swaps were the first product to be traded significantly and
became increasingly popular in the late 1990s until interest migrated to variance swaps.
Following the collapse of the single-stock variance market in the credit crunch, they are
having a renaissance due to demand from dispersion traders. A theoretical drawback of
volatility swaps is the fact that they require a volatility of volatility (vol of vol) model for
pricing, as options need to be bought and sold during the life of the contract (which
leads to higher trading costs). However, in practice, the vol of vol risk is small and
volatility swaps trade roughly in line with ATM forward (ATMf) implied volatility.
 Variance swaps. The difficulty in hedging volatility swaps drove liquidity towards the
variance swap market, particularly during the 2002 equity collapse. As variance swaps
can be replicated by delta hedging a static portfolio of options, it is not necessary to buy
or sell options during the life of the contract. The problem with this replication is that it
assumes options of all strikes can be bought, but in reality very OTM options are either
not listed or not liquid. Selling a variance swap and only hedging with the available
roughly ATM options leaves the vendor short tail risk. As the payout is on variance,
which is volatility squared, the amount can be very significant. For this reason, liquidity
on single-stock variance disappeared in the credit crunch.
 Gamma swaps. Dispersion traders profit from overpriced index-implied volatility by
going long single-stock variance and short index variance. The portfolio of variance
swaps is not static; hence, rebalancing trading costs are incurred. Investment banks
attempted to create a liquid gamma swap market, as dispersion can be implemented via a
static portfolio of gamma swaps (and, hence, it could better hedge the exposure of their
books from selling structured products). However, liquidity never really took off due to
limited interest from other market participants.

(1) VOLATILITY SWAPS

The payout of a volatility swap is simply the notional, multiplied by the difference between
the realised volatility and the fixed swap volatility agreed at the time of trading. As can be
seen from the payoff formula below, the profit and loss is completely path independent as it
is solely based on the realised volatility. Volatility swaps were previously illiquid, but are now
more popular with dispersion traders, given the single stock variance market no longer exists
post credit crunch. Unless packaged as a dispersion, volatility swaps rarely trade. As
dispersion is short index volatility, long single stock volatility, single stock volatility swaps
tend to be bid only (and index volatility swaps offered only).
54 CHAPTER 2: VOLATILITY TRADING

Volatility swap payoff

(σ F – σ S ) × volatility notional

where:

σ F = future volatility (that occurs over the life of contract)

σ S = swap rate volatility (fixed at the start of contract)

Volatility notional = Vega = notional amount paid (or received) per volatility point

(2) VARIANCE SWAPS

Variance swaps are identical to volatility swaps except their payout is based on variance
(volatility squared) rather than volatility. Variance swaps are long skew (more exposure to
downside put options than upside calls) and convexity (more exposure to OTM options than
ATM). Typically variance swaps trade in line with the 30 delta put (if skew is downward
sloping as normal). One-year variance swaps are the most frequently traded.

Variance swap payoff

(σ F 2 - σ S 2) × variance notional

where:

Variance notional = notional amount paid (or received) per variance point

NB: Variance notional = Vega / (2 × σ S ) where σ S = current variance swap price

CAPPED VAR SWAPS ARE SHORT OPTION ON VAR

Variance swaps on single stocks and emerging market indices are normally capped at 2.5
times the strike, in order to prevent the payout from rising towards infinity in a crisis or
bankruptcy. A cap on a variance swap can be modelled as a vanilla variance swap less an
option on variance whose strike is equal to the cap. More details can be found in the
section 2.4 Options on Variance.

Capped var should be hedged with OTM calls, not OTM puts
The presence of a cap on a variance swap means that if it is to be hedged by only one option
it should be a slightly OTM call, not an OTM (approx delta 30) put. This is to ensure the
option is so far OTM when the cap is hit that the hedge disappears. If this is not done, then
if a trader is long a capped variance swap he would hedge by going short an OTM put. If
2.3: Volatility, Variance and Gamma Swaps 55

markets fall with high volatility hitting the cap, the trader would be naked short a (now close
to ATM) put. Correctly hedging the cap is more important than hedging the skew position.

S&P500 variance market is increasing in liquidity

The payout of volatility swaps and variance swaps of the same vega is similar for small
payouts, but for large payouts the difference becomes very significant due to the quadratic
(ie, squared) nature of variance. The losses suffered in the credit crunch from the sale of
variance swaps, particularly single stock variance (which, like single stock volatility swaps
now, was typically bid only), have weighed on their subsequent liquidity. Now variance
swaps only trade for indices (usually without cap, but sometimes with). The popularity of
VIX futures has raised awareness of variance swaps, which has helped S&P500 variance
swaps become more liquid than they were before the credit crunch. SX5E variance liquidity
used to be in line with the S&P500, but is now far less liquid.

CORRIDOR VARIANCE SWAPS ARE NOT LIQUID

As volatility and spot are correlated, volatility buyers would typically only want exposure to
volatility levels for low values of spot. Conversely, volatility sellers would only want exposure
for high values of spot. To satisfy this demand, corridor variance swaps were created. These
only have exposure when spot is between spot values A and B. If A is zero, then it is a down
variance swap. If B is infinity, it is an up variance swap. There is only a swap payment on
those days the spot is in the required range, so if spot is never in the range there is no
payment. Because of this, a down variance swap and up variance swap with the same spot
barrier is simply a vanilla variance swap. The liquidity of corridor variance swaps was always
far lower than for variance swaps and, post credit crunch, they are rarely traded.

Corridor variance swap payoff

(σ F when in corridor 2 - σ S 2) × variance notional × percentage of days spot is within corridor

where:

σ F when in corridor = future volatility (of returns P i /P i-1 which occur when B L < P i-1 ≤ B H )

B L and B H , are the lower and higher barriers, where B L could be 0 and B H could be infinity.

(3) GAMMA SWAPS

The payout of gamma swaps is identical to that of a variance swap, except the daily P&L is
weighted by spot (price n ) divided by the initial spot (price 0 ). If spot range trades after the
position is initiated, the payouts of a gamma swap are virtually identical to the payout of a
variance swap. Should spot decline, the payout of a gamma swap decreases. Conversely, if
spot increases, the payout of a gamma swap increases. This spot-weighting of a variance
swap payout has the following attractive features:
56 CHAPTER 2: VOLATILITY TRADING

 Spot weighting of gamma swap payout makes it unnecessary to have a cap, even for
single stocks (if a company goes bankrupt with spot dropping close to zero with very
high volatility, multiplying the payout by spot automatically prevents an excessive
payout).
 If a dispersion trade uses gamma swaps, the amount of gamma swaps needed does not
change over time (hence, the trade is ‘fire and forget’, as the constituents do not have to
be rebalanced as they would if variance swaps were used).
 A gamma swap can be replicated by a static portfolio of options (although a different
static portfolio to variance swaps), which reduces hedging costs. Hence, no volatility of
volatility model is needed (unlike volatility swaps).

Gamma swap market has never had significant liquidity

A number of investment banks attempted to kick start a liquid gamma swap market, partly
to satisfy potential demand from dispersion traders and partly to get rid of some of the
exposure from selling structured products (if the product has less volatility exposure if prices
fall, then a gamma swap better matches the change in the vega profile when spot moves).
While the replication of the product is as trivial as for variance swaps, it was difficult to
convince other market participants to switch to the new product and liquidity stayed with
variance swaps (although after the credit crunch, single-stock variance liquidity moved to the
volatility swap market). If the gamma swap market ever gains liquidity, long skew trades
could be put on with a long variance-short gamma swap position (as this would be long
downside volatility and short upside volatility, as a gamma swap payout decreases/increases
with spot).

Gamma swap payoff

(σ G 2 - σ S 2) × variance notional

where:
price n
σ G 2 = future spot weighted (ie, multiplied by ) variance
price0

σ S 2 = swap rate variance (fixed at the start of contract)

2.3: Volatility, Variance and Gamma Swaps 57

PAYOUT OF VOL, VAR AND GAMMA SWAPS

The payout of volatility swaps, variance swaps and gamma swaps is the difference between
the fixed and floating leg, multiplied by the notional. The calculation for volatility assumes
zero mean return (or zero drift) to make the calculation easier and to allow the variance
calculation to be additive.

 Fixed leg. The cost (or fixed leg) of going long a volatility, variance or gamma swap is
always based on the swap price, σ S (which is fixed at inception of the contract). The
fixed leg is σ S for volatility swaps, but is σ S 2 for variance and gamma swaps).

 Floating leg. The payout (or floating leg) for volatility and variance swaps is based on
the same variable σ F (see equation below). The only difference is that a volatility swap
payout is based on σ F , whereas for a variance swap it is σ F 2. The gamma swap payout is
based on a similar variable σ G 2, which is σ F 2 multiplied by price n /price 0 .

∑[Ln(return )] i
2

σ F = 100 × i =1
× number business days in year
Texp

T
pricei
∑ price [Ln(returni )]2
σ G = 100 × i =1 0
× number business days in year
Texp

pricei
return i = for indices
pricei −1

pricei + dividendi
return i = for single stocks (dividend i is dividend going ex on day n)
pricei −1

where:

number of business days in year = 252 (usual market practice)

T exp = Expected value of N (if no market disruption occurs). A market disruption is when
shares accounting for at least 20% of the index market cap have not traded in the last 20
minutes of the trading day.
58 CHAPTER 2: VOLATILITY TRADING

Basis risk between cash and futures can cause traders problems
We note that the payout of variance swaps is based on the cash close, but traders normally
delta hedge using futures. The difference between the cash and futures price is called the
basis, and the risk due to a change in basis is called basis risk. Traders have to take this basis
risk between the cash close and futures close, which can be significant as liquidity in the
futures market tends to be reduced after the cash market closes.

VOL, VAR AND GAMMA SWAPS ARE ACTUALLY FORWARDS

The naming of volatility swaps, variance swaps and gamma swaps is misleading, as they are
in fact forwards. This is because their payoff is at maturity, whereas swaps have intermediate
payments.

VARIANCE IS ADDITIVE WITH ZERO MEAN ASSUMPTION

Normally, standard deviation or variance looks at the deviation from the mean. The above
calculations assume a zero mean, which simplifies the calculation (typically, one would
expect the mean daily return to be relatively small). With a zero mean assumption, variance is
additive. A mathematical proof of the formula below is given in the section A2 Measuring
Historical Volatility in the Appendix.

Past variance + future variance = total variance

Lack of dividend adjustment for indices affects pricing

The return calculation for a variance swap on an index does not adjust for any dividend
payments that go ex. This means that the dividend modelling method can affect the pricing.
Near-dated and, hence, either known or relatively certain dividends should be modelled
discretely rather than as a flat yield. The changing exposure of the variance swap to the
volatility on the ex date can be as large as 0.5 volatility points for a three-year variance swap
(if all other inputs are kept constant, discrete (ie, fixed) dividends lift the value of both calls
and puts, as proportional dividends simply reduce the volatility of the underlying by the
dividend yield).

Calculation agents have discretion as to when a market disruption occurs

Normally, the investment bank is the calculation agent for any variance swaps traded. As the
calculation agent normally has some discretion over when a market disruption event occurs,
this can lead to cases where one calculation agent believes a market disruption occurs and
another does not. This led to a number of disputes in 2008, as it was not clear if a market or
exchange disruption had occurred. Similarly, if a stock is delisted, the estimate of future
volatility for settlement prices is unlikely to be identical between firms, which can lead to
issues if a client is long and short identical products at different investment banks. These
problems are less of an issue if the counterparties are joint calculation agents.
2.3: Volatility, Variance and Gamma Swaps 59

HEDGING OF VAR CAN IMPACT EQUITY & VOL MARKET

Hedging volatility, variance and gamma swaps always involve the trading of a strip of
options of all strikes and delta hedging at the close. The impact the hedging of all three
products has on equity and volatility markets is similar, but we shall use the term variance
swaps, as it has by far the most impact of the three (the same arguments will apply for
volatility swaps and gamma swaps).

Short end of volatility surfaces is now pinned to realised

If there is a divergence between short-dated variance swaps and realised volatility, hedge
funds will put on variance swap trades to profit from this divergence. This puts pressure on
the short-dated end of volatility surfaces to trade close to the current levels of realised
volatility. Due to the greater risk of unexpected events, it is riskier to attempt a similar trade
at the longer-dated end of volatility surfaces.

Skew levels affected by direction of volatility trading

As variance swaps became a popular way to express a view of the direction of implied
volatility, they impacted the levels of skew. This occurred as variance swaps are long skew
(explained below) and, if volatility is being sold through variance swaps, this weighs on skew.
This occurred between 2003 and 2005, which pushed skew to a multiple-year low. As
volatility bottomed, the pressure from variance swap selling abated and skew recovered.

Delta hedge can suppress or exaggerate market moves

As the payout of variance swaps is based on the close-to-close return, they all have an
intraday delta (which is equal to zero if spot is equal to the previous day’s close). As this
intraday delta resets to zero at the end of the day, the hedging of these products requires a
delta hedge at the cash close. A rule of thumb is that the direction of hedging flow is in the
direction that makes the trade the least profit (ensuring that if a trade is crowded, it makes
less money). This flow can be hundreds of millions of US dollars or euros per day, especially
when structured products based on selling short-dated variance are popular (as they were in
2006 and 2007, less so since the high volatility of the credit crunch).

 Variance buying suppresses equity market moves. If clients are net buyers of
variance swaps, they leave the counterparty trader short. The trader will hedge this short
position by buying a portfolio of options and delta hedging them on the close. If spot
has risen over the day the position (which was originally delta-neutral) has a positive
delta (in the same way as a delta-hedged straddle would have a positive delta if markets
rise). The end of day hedge of this position requires selling the underlying (to become
delta-flat), which suppresses the rise of spot. Similarly, if markets fall, the delta hedge
required is to buy the underlying, again suppressing the market movement.
2.3: Volatility, Variance and Gamma Swaps 67

GREEKS OF VOL, VAR AND GAMMA SWAPS

As a volatility swap needs a vol of vol model, the Greeks are dependent on the model used.
For variance swaps and gamma swaps, there is no debate as to the Greeks. However,
practical considerations can introduce ‘shadow Greeks’. In theory, a variance swap has zero
delta, but in practice it has a small ‘shadow delta’ due to the correlation between spot and
implied volatility (skew). Similarly, theta is not necessarily as constant as it should be in
theory, as movements of the volatility surface can cause it to change.

Variance swap vega decays linearly with time

As variance is additive, the vega decays linearly with time. For example, 100K vega of a one
year variance swap at inception will have 75K vega after three months, 50K after six months
and 25K after nine months.

Variance swaps offer constant cash gamma, gamma swaps have constant
share gamma
Share gamma is the number of shares that need to be bought (or sold) for a given change in
spot (typically 1%). It is proportional to the Black-Scholes gamma (second derivative of price
with respect to spot) multiplied by spot. Cash gamma (or dollar gamma) is the cash amount
that needs to be bought or sold for a given movement in spot; hence, it is proportional to
share gamma multiplied by spot (ie, proportional to Black-Scholes gamma multiplied by spot
squared). Variance swaps offer a constant cash gamma (constant convexity), whereas gamma
swaps offer constant share gamma (hence the name gamma swaps).

γ = gamma = number of shares bought (or sold) per €1 spot move

γ×S = number of shares bought (or sold) per 100% spot (S×€1) move

γ × S / 100 = share gamma = number of shares bought (or sold) per 1% move

γ × S2 / 100 = cash/dollar gamma = notional cash value traded per 1% move

68 CHAPTER 2: VOLATILITY TRADING

2.4: OPTIONS ON VARIANCE

As the liquidity of the variance swap market improved in the middle of the last
decade, market participants started to trade options on variance. As volatility is more
volatile at high levels, the skew is positive (the inverse of the negative skew seen in
the equity market). In addition, volatility term structure is inverted, as volatility mean
reverts and does not stay elevated for long periods of time.

OPTIONS ON VAR EXPIRY = EXPIRY OF VAR SWAP

An option on variance is a European option (like all exotics) on a variance swap whose
expiry is the same expiry as the option. As it is an option on variance, a volatility of volatility
model is needed in order to price the option. At inception, the underlying is 100% implied
variance, whereas at maturity the underlying is 100% realised variance (and in between it will
be a blend of the two). As the daily variance of the underlying is locked in every day, the
payoff could be considered to be similar to an Asian (averaging) option.

Options on variance are quoted in volatility points

Like a variance swap, the price of an option on variance is quoted in volatility points. The
typical 3-month to 18-month maturity of the option is in line with the length of time it takes
3-month realised volatility to mean revert after a crisis. The poor liquidity of options on
variance, and the fact the underlying tends towards a cash basket over time, means a trade is
usually held until expiry.

Option on variance swap payoff

Max(σ F 2 - σ K 2, 0) × Variance notional

where:

σ F = future volatility (that occurs over the life of contract)

σ K = strike volatility (fixed at the start of contract)

Variance notional = notional amount paid (or received) per variance point

NB: Variance notional = Vega / (2σ S ) where σ S = variance swap reference (current fair price
of variance swap, not the strike)
2.4: Options on Variance 71

Implied variance term structure is inverted, but not as inverted as realised

variance
As historical volatility tends to mean revert in an eight-month time horizon (on average), the
term structure of options on variance is inverted (while volatility can spike and be high for
short periods of time, over the long term it trades in a far narrower range). We note that, as
the highest volatility occurs due to unexpected events, the peak of implied volatility (which is
based on the market’s expected future volatility) is lower than the peak of realised volatility.
Hence, the volatility of implied variance is lower than the volatility of realised variance,
especially for short maturities.

Figure 42. Option on Variance Skew Option on Variance Term Structure

78% 80%
Options on variance have 70%
76%
positive skew 60%
74% 50%
Implied vol

Implied vol
72% 40%
30% Options on variance have
70% inverted term structure
20%
68% 10%
66% 0%
80% 85% 90% 95% 100% 105% 110% 115% 120% 0 0.5 1 1.5 Maturity 2
Strike …
Positive skew (6 months) Negative term structure (ATM)

CAPPED VAR HAVE EMBEDDED OPTION ON VAR

While options on variance swaps are not particularly liquid, their pricing is key for valuing
variance swaps with a cap. Capped variance swaps are standard for single stocks and
emerging market indices and can be traded on regular indices as well. When the variance
swap market initially became more liquid, some participants did not properly model the cap,
as it was seen to have little value. The advent of the credit crunch and resulting rise in
volatility made the caps more valuable, and now market participants fail to model them at
their peril.

Variance Swap with Cap C = Variance Swap - Option on Variance with Cap C

Option on Variance with Cap C = Variance Swap - Variance Swap with Cap C

While value of cap is small at inception, it can become more valuable as

market moves
A capped variance swap can be modelled as a vanilla variance swap less an option on
variance, whose strike is the cap. This is true as the value of an option on variance at the cap
will be equal to the difference between the capped and uncapped variance swaps. Typically,
the cap is at 2.5× the strike and, hence, is not particularly valuable at inception. However, as
the market moves, the cap can become closer to the money and more valuable.
72 CHAPTER 2: VOLATILITY TRADING

OPTIONS ON VAR STRATEGIES SIMILAR TO OPTIONS

Strategies that are useful for vanilla options have a read-across for options on variance. For
example, a long variance position can be protected or overwritten. The increased liquidity of
VIX options allows relative value trades to be put on.

Selling straddles on options on variance can also be a popular strategy, as volatility can be
seen to have a floor above zero. Hence, strikes can be chosen so that the lower breakeven is
in line with the perceived floor to volatility.

Options on variance can also be used to hedge a volatility swap position, as an option on
variance can offset the vol of vol risk embedded in a volatility swap.

OPTIONS ON VOL FUTURES ARE SLIGHTLY DIFFERENT

While options on volatility futures have many similarities to options on variance, they are
slightly different. For more details see the section 4.5: Options On Volatility Futures.
CHAPTER 3
WHY EVERYTHING YOU THINK YOU
KNOW ABOUT VOLATILITY IS WRONG

Assuming implied volatility is an unbiased estimate of future realised volatility is an

easy mistake to make, however the fair price of implied volatility is above the expected
future realised volatility. In addition, the impact of hedging both structured products
and variable annuity products can cause imbalances in the volatility market. These
distortions can create opportunities for investors willing to take the other side. We
examine the opportunities from these imbalances and dispel the myth of using
volatility as an equity hedge.
3.1: Implied Volatility Should Be Above Realized Volatility 75

Far-dated options are most overpriced, due to upward sloping term structure
Volatility selling strategies typically involve selling near-dated volatility (or variance).
Examples include call overwriting or selling near-dated variance (until the recent explosion
of volatility, this was a popular hedge fund strategy that many structured products copied).
As term structure is on average upward sloping, this implies that far-dated implieds are more
expensive than near dated implieds. The demand for long-dated protection (eg, from
variable annuity providers) offers a fundamental explanation for term structure being upward
sloping (see section 3.3 Variable Annuity Hedging Lifts Long-Term Vol). However, as 12× one
month options (or variance swaps) can be sold in the same period of time as 1× one-year
option (or variance swap), greater profits can be earned from selling the near-dated product
despite it being less overpriced. We note the risk is greater if several near-dated options (or
variance swap) are sold in any period.

VOLATILITY OVERPRICING IS UNLIKELY TO DISAPPEAR

There are several fundamental reasons why volatility, and variance, is overpriced. Since these
reasons are structural, we believe that implied volatility is likely to remain overpriced for the
foreseeable future. Given variance exposure to overpriced wings (and low strike puts) and
the risk aversion to variance post credit crunch, we view variance as more overpriced than
volatility.

 Demand for put protection. The demand for hedging products, either from investors,
structured products or providers of variable annuity products, needs to be offset by
market makers. As market makers are usually net sellers of volatility, they charge margin
for taking this risk and for the costs of gamma hedging.
 Demand for OTM options lifts wings. Investors typically like buying OTM options as
there is an attractive risk-reward profile (similar to buying a lottery ticket). Market
makers therefore raise their prices to compensate for the asymmetric risk they face. As
the price of variance swaps is based on options of all strikes, this lifts the price of
variance.
 Index implieds lifted from structured product demand. The demand from
structured products typically lifts index implied compared to single-stock implied. This is
why implied correlation is higher than it should be.
76 CHAPTER 3: WHY EVERYTHING YOU THINK YOU KNOW ABOUT VOL IS WRONG

SELLING VOL SHOULD BE LESS PROFITABLE THAN BEFORE

Hedge funds typically aim to identify mispricings in order to deliver superior returns.
However, as both hedge funds and the total hedge fund marketplace grow larger, their
opportunities are gradually being eroded. We believe that above-average returns are only
possible in the following circumstances:

 A fund has a unique edge (eg, through analytics, trading algorithms or proprietary
information/analysis).
 There are relatively few funds in competition, or it is not possible for a significant
number of competitors to participate in an opportunity (either due to funding or legal
restrictions, lack of liquid derivatives markets or excessive risk/time horizon of trade).
 There is a source of imbalance in the markets (eg, structured product flow or
regulatory demand for hedging), causing a mispricing of risk.

All of the above reasons have previously held for volatility selling strategies (eg, call
overwriting or selling of one/three-month variance swaps). However, given the abundance
of publications on the topic in the past few years and the launch of several structured
products that attempt to profit from this opportunity, we believe that volatility selling could
be less profitable than before. The fact there remains an imbalance in the market due to the
demand for hedging should mean volatility selling is, on average, a profitable strategy.
However, we would caution against using a back test based on historical data as a reliable
estimate of future profitability.
3.2: Long Volatility Is a Poor Hedge 77

3.2: LONG VOLATILITY IS A POOR HEDGE

An ideal hedging instrument for a security is an instrument with -100% correlation to
that security and zero cost. As the return on variance swaps have a c-70% correlation
with equity markets, adding long volatility positions (either through variance swaps
or futures on volatility indices such as VIX or vStoxx) to an equity position could be
thought of as a useful hedge. However, as volatility is on average overpriced, the cost
of this strategy far outweighs any diversification benefit.

VOLATILITY HAS UP TO 70% CORRELATION WITH EQUITY

Equity markets tend to become more volatile when they decline and less volatile when they
rise. A fundamental reason for this is the fact that gearing increases as equities decline 7. As
both gearing and volatility are measures of risk, they should be correlated; hence, they are
negatively correlated to equity returns. While short term measures of volatility (eg, vStoxx)
only have an R2 of 50%-60% against the equity market, longer dated variance swaps (purest
way to trade volatility) can have up to c70% R2.

1 YEAR VOLATILITY HAS HIGHEST CORRELATION TO EQUITY

There are two competing factors to the optimum maturity for a volatility hedge. The longer
the maturity, the more likely the prolonged period of volatility will be due to a decline in the
market. This should give longer maturities higher equity volatility correlation, as the impact
of short-term noise is reduced. However, for long maturities (years), there is a significant
chance that the equity market will recover from any downturn, reducing equity volatility
correlation. The optimum correlation between the SX5E and variance swaps, is for returns
between nine months and one year. This is roughly in line with the c8 months it takes
realised volatility to mean revert after a crisis.

SHORT-DATED VOLATILITY FUTURES ARE A POOR HEDGE

Recently, there have been several products based on rolling VIX or vStoxx futures whose
average maturity is kept constant. As these products have to continually buy far-dated
futures and sell near-dated futures (to keep average maturity constant as time passes), returns
suffer from upward sloping term structure. Since the launch of vStoxx futures, rolling one-
month vStoxx futures have had negative returns (see Figure 51 below). This is despite the
SX5E also having suffered a negative return, suggesting that rolling vStoxx futures are a
poor hedge. For more details on futures on volatility indices, see the section 4.1 Forward
Starting Products.

7 Assuming no rights issues, share buybacks, debt issuance or repurchase/redemption.

80 CHAPTER 3: WHY EVERYTHING YOU THINK YOU KNOW ABOUT VOL IS WRONG

Equities need to have positive return over hedging back-testing period

While we acknowledge that there are periods of time in which a long volatility position is a
profitable hedge, these tend to occur when equity returns are negative (and short futures are
usually a better hedge). We believe that the best back-testing periods for comparing hedging
strategies are those in which equities have a return above the risk-free rate (if returns below
the risk-free rate are expected, then investors should switch allocation away from equities
into risk-free debt). For these back-testing periods, long volatility strategies struggle to
demonstrate value as a useful hedging instrument. Hence, we see little reason for investors
to hedge with variance swaps rather than futures given the overpricing of volatility, and less
than 100% correlation between volatility and equity returns.

HEDGING WITH VARIANCE IS NOT COMPABLE TO PUTS

Due to the lack of convexity of a variance swap hedge, we believe it is best to compare long
variance hedges to hedging with futures rather than hedging with puts. Although variance
hedges might be cheaper than put hedges, the lack of convexity for long volatility makes this
an unfair comparison, in our view.
3.3: Variable Annuity Hedging Lifts Long-Term Vol 81

3.3: VARIABLE ANNUITY HEDGING LIFTS LONG-

TERM VOL
Since the 1980s, a significant amount of variable annuity products have been sold,
particularly in the USA. The size of this market is now over US$1trn. From the mid-
1990s, these products started to become more complicated and offered guarantees to
the purchaser (similar to being long a put). The hedging of these products increases
the demand for long-dated downside strikes, which lifts long-dated implied volatility
and skew.

VARIABLE ANNUITIES OFTEN EMBED A ‘PUT’ OPTION

With a fixed annuity, the insurance company that sold the product invests the proceeds and
guarantees the purchaser a guaranteed fixed return. Variable annuities, however, allow the
purchaser to pick which investments they want to put their funds into. The downside to this
flexibility is the unprotected exposure to a decline in the market. To make variable annuities
more attractive, from the 1990s many were sold with some forms of downside protection (or
put). The different types of protection are detailed below in order of the risk to the insurance
company.

 Return of premium. This product effectively buys an ATM put in addition to investing
proceeds. The investor is guaranteed returns will not be negative.
 Roll-up. Similar to return of premium; however, the minimum guaranteed return is
greater than 0%. The hedging of this product buys a put which is ITM with reference to
spot, but OTM compared with the forward.
 Ratchet (or maximum anniversary value). These products return the highest value the
underlying has ever traded at (on certain dates). The hedging of these products involves
payout look-back options, more details of which are in the section 5.4 Look-Back Options.
 Greater of ‘ratchet’ or ‘roll-up’. This product returns the greater of the ‘roll-up’ or
‘ratchet’ protection.

Hedging of variable annuity products lifts index term structure and skew
The hedging of variable annuity involves the purchase of downside protection for long
maturities. Often the products are 20+ years long, but as the maximum maturity with sufficient
liquidity available on indices can only be 3-5 years, the position has to be dynamically hedged
with the shorter-dated option. This constant bid for long-dated protection lifts index term
structure and skew, particularly for the S&P500 but also affects other major indices (potentially
due to relative value trading). The demand for protection (from viable annuity providers or
other investors), particularly on the downside and for longer maturities, could be considered to
be the reason why volatility (of all strikes and maturities), skew (for all maturities) and term
structure are usually overpriced.
3.4: Structured Products Vicious Circle 87

(4) SHORT VEGA POSITION INCREASES DUE TO VEGA CONVEXITY

Options have their peak vega when they are (approximately) ATM. As implied volatility
increases, the vega of OTM options increases and converges with the vega of the peak ATM
option. Therefore, as implied volatility increases, the vega of OTM options increases (see
Figure 50 above). The rate of change of vega given a change in volatility is called volga
(VOL-GAmma) or vomma, and is known as vega convexity.

Volga = dVega/dVol

Vega convexity causes short volatility position to increase

As the vega of options rises as volatility increases, this increases the size of the short
volatility position that needs to be hedged. As trading desks’ volatility short position has now
increased, they have to buy volatility to cover the increased short position, which leads to
further gains in implied volatility. This starts a vicious circle of increasing volatility, which we
call the ‘structured products vicious circle’.

VEGA CONVEXITY HIGHEST FOR LOW-TO-MEDIUM IMPLIEDS

As Figure 56 above shows, the slope of vega against volatility is steepest (ie, vega convexity
is highest) for low-to-medium implied volatilities. This effect of vega convexity is therefore
more important in volatility regimes of c20% or less; hence, the effect of structured products
can have a similar effect when markets rise and volatilities decline. In this case, trading desks
become long vega, due to skew, and have to sell volatility. Vega convexity causes this long
position to increase as volatility declines, causing further volatility sellings. This is typically
seen when a market recovers after a volatile decline (eg, in 2003 and 2009, following the end
of the tech bubble and credit crunch, respectively).

IMPACT GREATEST FOR FAR-DATED IMPLIEDS

While this position has the greatest impact at the far end of volatility surfaces, a rise in far-
dated term volatility and skew tends to be mirrored to a lesser extent for nearer-dated
expiries. If there is a disconnect between near- and far-dated implied volatilities, this can
cause a significant change in term structure.
88 CHAPTER 3: WHY EVERYTHING YOU THINK YOU KNOW ABOUT VOL IS WRONG

STRUCTURED PRODUCT GUARANTEE IS LONG AN OTM PUT

The capital guarantee of many structured products leaves the seller of the product effectively
short an OTM put. A short OTM put is short skew and short vega convexity (or volga). This
is a simplification, as structured products tend to buy visually cheap options (ie, OTM
options) and sell visually expensive options (ie, ATM options), leaving the seller with a long
ATM and short OTM volatility position. As OTM options have more volga (or vega
convexity) than ATM options (see the section A8 Greeks and Their Meaning in the Appendix)
the seller is short volga. The embedded option in structured products is floored, which
causes the seller to be short skew (as the position is similar to being short an OTM put).
3.5: Stretching Black-Scholes Assumptions 89

3.5: STRETCHING BLACK-SCHOLES ASSUMPTIONS

The Black-Scholes model assumes an investor knows the future volatility of a stock,
in addition to being able to continuously delta hedge. In order to discover what the
likely profit (or loss) will be in reality, we stretch these assumptions. If the future
volatility is unknown, the amount of profit (or loss) will vary depending on the path,
but buying cheap volatility will always show some profit. However, if the position is
delta-hedged discretely, the purchase of cheap volatility may reveal a loss. The
variance of discretely delta-hedged profits can be halved by hedging four times as
frequently. We also show why traders should hedge with a delta calculated from
expected – not implied – volatility, especially when long volatility.

MODEL ASSUMES KNOWN VOL AND CONTINUOUS HEDGING

While there are a number of assumptions behind Black-Scholes, the two which are the least
realistic are: (1) a continuous and known future realised volatility; and (2) an ability to delta
hedge continuously. There are, therefore, four different scenarios to investigate. We assume
that options are European.

 Continuous delta hedging with known volatility. In this scenario, the profit (or loss)
from volatility trading is fixed. If the known volatility is constant, then the assumptions
are identical to Black-Scholes. Interestingly, the results are the same if volatility is
allowed not to be constant (while still being known).

 Continuous delta hedging with unknown volatility. If volatility is unknown, then

typically traders hedge with the delta calculated using implied volatility. However, as
implied volatility is not a perfect predictor of future realised volatility, this causes some
variation in the profit (or loss) of the position. However, with these assumptions, if
realised volatility is above the implied volatility price paid, it is impossible to suffer a
loss.

 Discrete delta hedging with known volatility. As markets are not open 24/7,
continuous delta hedging is arguably an unreasonable assumption. The path dependency
of discrete delta hedging adds a certain amount of variation in profits (or losses), which
can cause the purchase of cheap volatility (implied less than realised) to suffer a loss.
The variance of the payout is inversely proportional to the frequency of the delta
hedging. For example, the payout from hedging four times a day has a variance that is a
quarter of the variance that results if the position is hedged only once a day.

 Discrete delta hedging with unknown volatility. The most realistic assumption is to
hedge discretely with unknown volatility. In this case, the payout of volatility trading is
equal to the sum of the variance due to hedging with unknown volatility plus the
variance due to discretely delta hedging.
98 CHAPTER 3: WHY EVERYTHING YOU THINK YOU KNOW ABOUT VOL IS WRONG

Hedging with delta using implied volatility is bad for long vol strategies
Typically, when volatility rises there is often a decline in the markets. The strikes of the
option are therefore likely to be above spot when actual volatility is above implied. This
reduces the profits of the delta-hedged position as the position is actually long delta when it
appears to be delta flat. Alternatively, the fact that the position hedged with the realised
volatility over the life of the option is profitable can be thought of as due to the fact it is
properly gamma hedged, as it has more time value than is being priced into the market.
Hence, if a trader buys an option when the implied looks 5pts too cheap, then the delta
should be calculated from a volatility 5pts above current implied volatility. Using the proper
volatility assumption to calculate the delta means the profit from delta hedging an option is
approximately the difference between the theoretical value of the option at inception (ie,
using actual future realised volatility in pricing) and the price of the option (ie, using implied
volatility in pricing).

While this analysis has concentrated on delta, a similar logic applies for the calculation of all
Greeks.
CHAPTER 4
FORWARD STARTING PRODUCTS AND
VOLATILITY INDICES

Forward starting options are a popular method of trading forward volatility and term
structure as there is no exposure to near-term volatility and, hence, zero theta (until
the start of the forward starting option). Recently, trading forward volatility via
volatility futures such as VIX and vStoxx futures has become increasingly popular.
However, as is the case with forward starting options, there are modelling issues.
100 CHAPTER 4: FORWARD STARTING PRODUCTS AND VOLATILITY INDICES

4.1: FORWARD STARTING PRODUCTS

Forward starting options are a popular method of trading forward volatility and term
structure as there is no exposure to near-term volatility and, hence, zero theta (until
the start of the forward starting option. As the exposure is to forward volatility rather
than volatility, more sophisticated models need to be used to price them than
ordinary options. Forward starting options will usually have wider bid-offer spreads
than vanilla options, as their pricing and hedging is more complex. Recently, trading
forward volatility via VIX and vStoxx futures has become increasingly popular,
however there are modelling issues. Forward starting variance swaps are easier to
price as the price is determined by two variance swaps.

ZERO THETA IS ADVANTAGE OF FWD STARTING PRODUCTS

The main attraction of forward starting products is that they provide investors with long-
term volatility (or vega) exposure, without having exposure to short-term volatility (or
gamma) 11. As there is zero gamma until the forward (fwd) starting product starts, the
product does not have to pay any theta. Forward starting products are most appropriate for
investors who believe that there is going to be volatility in the future (eg, during a key
economic announcement or a reporting date) but that realised volatility is likely to be low in
the near term (eg, over Christmas or the summer lull).

Forward starting products are low cost, but also lower payout
We note that while forward starting products have a lower theta cost than vanilla options, if
there is a rise in volatility surfaces before the forward starting period is over, they are likely
to benefit less than vanilla options (this is because the front end of volatility surfaces tends
to move the most, and this is the area to which forward start has no sensitivity). Forward
starting products can therefore be seen as a low-cost, lower-payout method of trading
volatility.

TERM STRUCTURE PENALISES FWD STARTING PRODUCTS

While forward starting products have zero mathematical theta, they do suffer from the fact
that volatility and variance term structure is usually expensive and upward sloping. The
average implied volatility of a forward starting product is likely to be higher than a vanilla
product, which will cause the long forward starting position to suffer carry as the volatility is
re-marked lower 12 during the forward starting period.

SKEW CAUSES NEGATIVE SHADOW DELTA

11We shall assume that the investor wishes to be long a forward starting product.
12 If a 3-month forward starting option is compared to a 3-month vanilla option, then during the
forward starting period the forward starting implied volatility should, on average, decline.
4.1: Forward Starting Products 101

The presence of skew causes a correlation between volatility and spot. This correlation
produces a negative shadow delta for all forward starting products (forward starting options
have a theoretical delta of zero). The rationale is similar to the argument that variance swaps
have negative shadow delta due to skew.

FIXED DIVIDENDS ALSO CAUSES SHADOW DELTA

If a dividend is fixed, then the dividend yield tends to zero as spot tends to infinity, which
causes a shadow delta (which is positive for calls and negative for puts).

Proportional dividends reduce volatility of underlying

Options, variance swaps and futures on volatility indices gain in value if dividends are fixed,
as proportional dividends simply reduce the volatility of an underlying.

THERE ARE 3 MAIN METHODS TO TRADE FORWARD VOL

Historically, forward volatility could only be traded via forward starting options, which had
to be dynamically hedged and, hence, had high costs and wide bid-offer spreads. When
variance swaps became liquid, this allowed the creation of forward starting variance swaps
(as a forward starting variance can be perfectly hedged by a long and short position in two
vanilla variance swaps of different maturity, which is explained later). The client base for
trading forward volatility has recently been expanded by the listing of forwards on volatility
indices (such as the VIX or vStoxx). The definition of the three main forward starting
products is given below:

(1) Forward starting options. A forward starting option is an option whose strike will be
determined at the end of the forward starting period. The strike will be quoted as a
percentage of spot. For example, a one-year ATM option three-month forward start,
bought in September 2012, will turn into a one-year ATM option in December 2012 (ie,
expiry will be December 2013 and the strike will be the value of spot in December
2012). Forward starting options are quoted OTC. For flow client requests, the maturity
of the forward starting period is typically three months and with an option maturity no
longer than a year. The sale of structured products creates significant demand for
forward starting products, but of much longer maturity (2-3 years, the length of the
structured product). Investment banks will estimate the size of the product they can sell
and buy a forward starting option for that size. While the structured product itself does
not incorporate a forward start, as the price for the product needs to be fixed for a
period of 1-2 months (the marketing period), the product needs to be hedged with a
forward start before marketing can begin.
(2) Forward starting variance swaps. The easiest forward starting product to trade is a
variance swap, as it can be hedged with two static variance swap positions (one long, one
short). Like plain variance swaps, these products are traded OTC and their maturities
102 CHAPTER 4: FORWARD STARTING PRODUCTS AND VOLATILITY INDICES

can be up to a similar length (although investors typically ask for quotes up to three
years).

(3) Futures on volatility index. A forward on a volatility index works in the same way as a
forward on an equity index: they both are listed and both settle against the value of the
index on the expiry date. While forwards on volatility indices such as the VIX and
vStoxx have been quoted for some time, their liquidity has only recently improved to
such an extent that they are now a viable method for trading. This improvement has
been driven by increasing structured issuance and by options on volatility indices (delta
hedging of these options has to be carried out in the forward market). Current listed
maturities for the VIX and vStoxx exist for expiries under a year.
HIGHER HEDGING RISKS INCREASE COST
While forward starting options do not need to be delta hedged before the forward starting
period ends, they have to be vega hedged with vanilla straddles (or very OTM strangles if
they are liquidity enough, as they also have zero delta and gamma). A long straddle has to be
purchased on the expiry date of the option, while a short straddle has to be sold on the strike
fixing date. As spot moves the strikes will need to be rolled, which increases costs (which are
likely to be passed on to clients) and risks (unknown future volatility and skew) to the trader.

Pricing of futures on vol indices tends to be slanted against long investors

Similarly, the hedging of futures on volatility indices is not trivial, as (like volatility swaps)
they require a volatility of volatility model. While the market for futures on volatility indices
has become more liquid, as the flow is predominantly on the buy side, forwards on volatility
indices have historically been overpriced. They are a viable instrument for investors who
want to short volatility, or who require a listed product.

Forward starting var swaps have fewer imbalances than other products
The price – and the hedging – of a forward starting variance swap is based on two vanilla
variance swaps (as it can be constructed from two vanilla variance swaps). The worst-case
scenario for pricing is therefore twice the spread of a vanilla variance swap. In practice, the
spread of a forward starting variance swap is usually slightly wider than the width of the
widest bid-offer of the variance swap legs (ie, slightly wider than the bid-offer of the furthest
maturity).
4.1: Forward Starting Products 103

(1) FORWARD STARTING OPTIONS

A forward starting option can be priced using Black-Scholes in a similar way to a vanilla
option. The only difference is that the forward volatility (rather than volatility) is needed as
an input 13. The three different methods of calculating the forward volatility, and examples of
how the volatility input changes, are detailed below:

 Sticky delta (or moneyness) and relative time. This method assumes volatility
surfaces never change in relative dimensions (sticky delta and relative time). This is not a
realistic assumption unless the ATM term structure is approximately flat.
 Additive variance rule. Using the additive variance rule takes into account the term
structure of a volatility surface. This method has the disadvantage that the forward skew
is assumed to be constant in absolute (fixed) time, which is not usually the case. As skew
is normally larger for shorter-dated maturities, it should increase approaching expiry.
 Constant smile rule. The constant smile rule combines the two methods above by
using the additive variance rule for ATM options (hence, it takes into account varying
volatility over time) and applying a sticky delta skew for a relative maturity. It can be
seen as ‘bumping’ the current volatility surface by the change in ATM forward volatility
calculated using the additive variance rule.

STICKY DELTA & RELATIVE TIME USES CURRENT VOL SURFACE

If the relative dimensions of a volatility surface are assumed to never change, then the
volatility input for a forward starting option can be priced with the current volatility surface.
For example, a three-month 110% strike option forward starting after a period of time T can
be priced using the implied volatility of a current three-month 110% strike option (the
forward starting time T is irrelevant to the volatility input 14). As term structure is normally
positive, this method tends to underprice forward starting options. An example of a current
relative volatility surface, which can be used for pricing forward starting options under this
method, is shown below:

13 Forwards of the other inputs, for example interest rates, are generally trivial to compute.
14 Hence, the price of the three-month 110% option forward start will only be significantly different
from the price of the vanilla three-month 110% option if there is a significant difference in interest
rates or dividends.
106 CHAPTER 4: FORWARD STARTING PRODUCTS AND VOLATILITY INDICES

impact of having a relative time skew on a fixed ATM volatility can be measured by volatility
slide theta (see the section A9.Advanced (Practical or Shadow) Greeks in the Appendix).

CONSTANT SMILE RULE IS THE BEST MODEL OF THE THREE

Pricing with static delta and relative time usually underprices forward volatility (as volatility
term structure is normally upward sloping, and long-dated forward volatility is sold at the
lower levels of near-dated implied volatility). While additive variance correctly prices forward
volatility, this rule does mean future skew will tend towards zero (as skew tends to decay as
maturity increases and the additive variance rule assumes absolute – fixed – time for skew).
While this rule has been used in the past, the mispricing of long-dated skew for products
such as cliquets has led traders to move away from this model. The constant smile rule
would appear to be the most appropriate.

(2) FORWARD STARTING VAR SWAPS

In the section A2.Measuring Historical Volatility in the Appendix we show that variance is
additive (variance to time T 2 = variance to time T 1 + forward variance T 1 to T 2 ). This
allows the payout of a forward starting variance swap between T 1 and T 2 to be replicated via
a long variance swap to T 2 , and short variance swap to T 1 . We define N 1 and N 2 to be the
notionals of the variance swaps to T 1 and T 2, respectively. It is important to note that N 1
and N 2 are the notionals of the variance swap, not the vega (N = vega ÷ 2 σ). As the
variance swap payout of the two variance swaps must cancel up to T 1 , the following
relationship is true (we are looking at the floating leg of the variance swaps, and ignore
constants that cancel such as the annualisation factor):

Payout long var to T 2 = Payout long var to T 1 + Payout long var T 1 to T 2

T2 T1 T2

∑ [Ln(returni )]2 ∑ [Ln(returni )]2 ∑ [Ln(return )]

i =T1 +1
i
2

 N2 i =1
= N2 i =1
+ N2
T2 T2 T2

T1

∑ [Ln(return )]
i
2

 Payout short variance = N 2 i =1

(= Payout long variance to T 1 )
T2

T1 T1

∑ [Ln(returni )]2 ∑ [Ln(return )] i

2

 N1 i =1
= N2 i =1
T1 T2
108 CHAPTER 4: FORWARD STARTING PRODUCTS AND VOLATILITY INDICES

2 σ 2 2T2 − σ 12T1
 σ 12 = = forward volatility from T 1 to T 2
T2 − T1
Forward starting var swaps have fewer imbalances than other fwd products
The price – and the hedging – of a forward starting variance swap is based on two vanilla
variance swaps (as it can be constructed from two vanilla variance swaps). The worst-case
scenario for pricing is therefore twice the spread of a vanilla variance swap. In practice, the
spread of a forward starting variance swap is usually slightly wider than the width of the
widest bid-offer of the variance swap legs (ie, slightly wider than the bid-offer of the furthest
maturity).

(3) FUTURE ON VOLATILITY INDEX

Futures on volatility indices have become one of the most popular forward starting
products. For more details on both volatility indices, and futures on those indices, please see
the following two sections 4.2: Volatility Indices and 4.3: Futures On Volatility Indices.
4.2: Volatility Indices 109

4.2: VOLATILITY INDICES

While volatility indices were historically based on ATM implied, most providers have
swapped to a variance swap-based calculation. The price of a volatility index will,
however, typically be 0.2-0.7pts below the price of a variance swap of the same
maturity as the calculation of the volatility index typically chops the tails to remove
illiquid prices. Each volatility index provider has to use a different method of
chopping the tails in order to avoid infringing the copyright of other providers.

THERE ARE 2 WAYS OF CALCULATING A VOLATILITY INDEX

Historically, volatility indices (old VIX and VDAX) were based on ATM implied volatility.
This level is virtually identical to the fair price of a volatility swap (as volatility swaps ≈
ATMf implied). This methodology has the advantage that it uses the most liquid strikes, and
it is still used by some providers in less liquid markets for this reason. Due to the realisation
that variance, not volatility, was the correct measure of deviation, on September 22, 2003,
the VIX index moved away from using ATM implied towards a variance-based calculation
(and also moved from using the S&P100 to the S&P500). While the calculation is variance-
based, the index is quoted as the square root of variance for an easier comparison with the
implied volatility of options (but we note that skew and convexity mean the fair price of
variance swaps and volatility indices should always trade above ATM options).

Volatility indices based on ATM implied usually average 8 different options

The old VIX, renamed VXO, took the implied volatility for the S&P100 strikes above and
below spot for both calls and puts. As the first two-month expiries were used, the old index
was calculated using eight implied volatility measures as 8 = 2 (strikes) × 2 (put/call) × 2
(expiry). Similarly, the VDAX index, which was based on DAX 45-day ATM-implied
volatility, has been superseded by the V1X index, which, like the new VIX, uses a variance-
based calculation.

MOST INDICES NOW USE VARIANCE-BASED CALCULATIONS

Variance-based calculations have also been used by other volatility index providers. All
recent volatility indices, such as the vStoxx (V2X), VSMI (V3X), VFTSE, VNKY and VHSI,
use a variance swap calculation, although we note the recent VIMEX index uses a similar
methodology to the old VIX (due to illiquidity of OTM options on the Mexican index).
While the formula for a variance is a mathematical formula and hence not subject to
copyright, if this formula is modified to exclude tails (eg, requiring a non-zero bid and/or
offer price, excluding strikes too far away from spot, etc), then this calculation becomes
proprietary and is subject to copyright. This is the reason why different volatility index
providers have chosen different calculation methods.
110 CHAPTER 4: FORWARD STARTING PRODUCTS AND VOLATILITY INDICES

DIFFERENCES BETWEEN VOL INDICES AND VAR SWAPS

While the calculation of a volatility index might be based on a variance swap calculation, the
price of a volatility index will typically be lower than that of a variance swap. The magnitude
of the difference depends on the calculation itself, the number of strikes with available prices
and the difference in width between strikes.

 Excluding very high and very low strikes. To increase the stability of the calculation,
volatility indices exclude the implied volatility of options with very high or very low
strikes. Given the importance of low strike implied volatility to variance swap pricing,
chopping the wings of low strike implieds has a greater impact than removing high strike
implieds, hence the level of a volatility index is below the variance swap price (typically
between 0.2 and 0.7 volatility points).

 Discrete sampling by using only listed strikes. When pricing a variance swap, the
value of a parameterised volatility surface is used. This surface is completely continuous
and, hence, is not subject to errors due to using discrete data. As a volatility index has to
rely on data from listed strikes, this introduces a small error which causes the level of the
implied volatility index to be slightly below the variance swap price.

 Noise due to rolling expiries. If a volatility index does not interpolate between
expiries then the implied volatility will ‘jump’ when the maturity rolls from one expiry to
another. This difference can be c2 volatility points. Some indices only interpolate over a
few days and take an exact maturity the rest of the time, which smoothes this effect (but
does not fully remove it). Indices calculated by the CBOE move from interpolation to
extrapolation which will cause similar noise, but has a much smaller effect than rolling.
The average value from a volatility index that uses rolling is below the value of a
variance swap as term structure is normally positive.

 Linear interpolation between expiries (should be square root of time). Linearly

interpolating between expiries assumes a flat volatility term structure. In reality, a
volatility surface follows a ‘square root of time’ rule, which means that the slope of term
structure is steeper for near-dated maturities than for far dated ones. As a volatility
surface is normally upward sloping this means a volatility index is on average below the
level of a variance swap (up to c0.8 volatility points).
114 CHAPTER 4: FORWARD STARTING PRODUCTS AND VOLATILITY INDICES

Interpolation between expiries usually lowers the value of a volatility index

The slope of near-dated implieds is typically steeper than far-dated implieds, as volatility
surfaces often move in a ‘square root of time’ manner (near-dated implieds fall more than
far-dated implieds when volatility declines). Given the steeper slope of near-dated implieds,
linearly interpolating underestimates the implied volatility for positive term structure (and
similarly overestimates it for negative term structure). Typically, the demand for long-dated
hedges and risk aversion causes far-dated implieds to be greater than near-dated implieds,
hence term structure is normally positive. The effect of linearly interpolating between
maturities for a fixed maturity volatility index therefore causes a volatility index to normally
underestimate a variance swap level. This effect will be greatest when the (typically 1-month)
maturity of the volatility index is exactly in between listed expiries, and an extreme example
is given in the graph above where the difference is c0.8 volatility points. We note that should
the maturity of a volatility index lie close to a listed maturity the error due to interpolating
between expiries will be close to zero.
4.3: Futures on Volatility Indices 115

4.3: FUTURES ON VOLATILITY INDICES

While futures on volatility indices were first launched on the VIX in March 2004, it
has only been since the more recent launch of structured products and options on
volatility futures that liquidity has improved enough to be a viable method of trading
volatility. As a volatility future payout is based on the square root of variance, the
payout is linear in volatility not variance. The fair price of a future on a volatility
index is between the forward volatility swap, and the square root of the forward
variance swap. Volatility futures are, therefore, short vol of vol, just like volatility
swaps. It is therefore possible to get the implied vol of vol from the listed price of
volatility futures.

PRICE IS BETWEEN FORWARD VAR AND FORWARD VOL

A future on a volatility index functions in exactly the same way as a future on an equity
index. However, as volatility future is a forward (hence linear) payout of the square root of
variance, the payoff is different from a variance swap (whose payout is on variance itself).
The price of a forward on a volatility index lies between the fair value of a forward volatility
swap and the square root of the fair value of a forward variance swap.

σ Forward volatility swap ≤ Future on volatility index ≤ σ Forward variance swap

FUTURES ON VOLATILITY INDICES ARE SHORT VOL OF VOL

A variance swap can be hedged by delta hedging a portfolio of options (the portfolio is
known as a log contract, where the weight of each option is 1/K2 where K is the strike). As
the portfolio of options does not change, the only hedging costs are the costs associated
with delta hedging. A volatility swap has to be hedged through buying and selling variance
swaps (or a log contract of options); hence, it needs to have a volatility of volatility model.
As a variance swap is more convex than a volatility swap (variance swap payout is on
volatility squared), a volatility swap is short convexity compared to a variance swap. A
volatility swap is, therefore, short volatility of volatility (vol of vol) as a variance swap has no
vol of vol risk. As the price of a future on a volatility index is linear in volatility, a future on a
volatility index is short vol of vol (like volatility swaps).
4.3: Futures on Volatility Indices 119

VIX SETTLEMENT PRICE COULD BE MANIPULATED

As the settlement price for the VIX is based on opening trades on S&P500 options, it is far
easier to manipulate than the settlement price for the vStoxx, which is based on a 30-minute
average ending midday (CET). As shown earlier, the payout of a variance swap is based on a
portfolio of options of all strikes weighted 1/K2, where K is the strike of the option. This
means the calculation of variance swaps is very sensitive to the price of downside puts.
Typically, there are offers for downside puts of all strikes at the tick value (ie, the smallest
possible non-zero price), as these puts have near zero theoretical value. As the VIX
calculation requires a non-zero bid, these offers are usually excluded for strikes below c50%-
60%. By entering the minimum bid of US$0.05 (= tick value), these prices will be included in
the calculation and could lift the settlement price by c1pt (as the implied volatility for these
low strike puts will be very large). There have been times when the VIX settlement (on the
open) has been significantly different to both the close of the day, and the close of the
previous day.

MEAN REVERSION MEANS VOL FUTURES HAVE DELTA <1

Unlike normal futures, volatility futures are not linear in the underlying index (as the mean
reversion of volatility has an effect) and, hence, have deltas significantly lower than 100%.
An equity future has near a 100% delta. While the front month VIX future has a high 90%
delta (delta vs the VIX), the 6-month VIX future has a lower 55% delta. The lower delta is
due to the mean reversion of volatility, as 6-month VIX futures will not trade at 80% even if
the VIX trades at 80% (as the VIX only briefly went above 80% post Lehman bankruptcy
and swiftly declined, it is highly unlikely to still be at 80% in 6 months’ time). The empirical
deltas of VIX futures by maturity are shown in Figure 70 below. These values decline in a
similar way but less rapidly than they would if volatility solely obeyed a square root of time
rule, which is to be expected as volatility surfaces sometimes move in parallel.
122 CHAPTER 4: FORWARD STARTING PRODUCTS AND VOLATILITY INDICES

4.4: VOLATILITY FUTURE ETN/ETF

Structured products based on constant maturity volatility futures have become
increasingly popular and in the US have at times had a greater size than the
underlying volatility futures market. As a constant maturity volatility product needs
to sell near-dated expiries and buy far-dated expiries, this flow supports term
structure for volatility futures and the underlying options on the index itself. The
success of VIX-based products has led to their size being approximately two-thirds
of the vega of the relevant VIX futures market (which is a similar size to the net listed
S&P500 market) and, hence, appears to be artificially lifting near-dated term
structure. The size of vStoxx products is not yet sufficient to significantly impact the
market, hence they are a more viable method of trading volatility in our view. We
recommend shorting VIX-based structured products to profit from this imbalance,
potentially against long vStoxx based products as a hedge. Investors who wish to be
long VIX futures should consider the front-month and fourth-month maturities, as
their values are likely to be depressed from structured flow.

STRUCTURED PRODUCTS IMPROVED FUTURES LIQUIDITY

As it is impossible to have a product (perpetual or otherwise) whose payout is the volatility
index itself, volatility futures were launched to give investors an easy method of trading
volatility. Initially, VIX and vStoxx futures had limited liquidity, potentially as they are not
perpetual; however, the creation of perpetual structured products has improved the liquidity
of volatility futures. Similarly, the introduction of options on these futures has increased the
need to delta hedge using these futures, also increasing liquidity. In the US, the size of
structured products on VIX futures is so large at times it was bigger than the underlying VIX
futures market and appears to have moved the underlying S&P500 market itself.

VIX PRODUCTS ARE 2/3 OF THE SIZE OF FUTURES MARKET

The size in vega of the US market for vanilla S&P500 options, VIX futures and VIX-based
ETN/ETF is shown in Figure 72 below. As can be seen, the size of VIX-based ETN/ETFs
is approximately two-thirds of the size of the relevant VIX future.
130 CHAPTER 4: FORWARD STARTING PRODUCTS AND VOLATILITY INDICES

VOLATILITY TERM STRUCTURE IS VERY NEGATIVE

The term structure of implied volatility of vanilla equity options is on average relatively
flat 17. In contrast, the term structure of implied volatility of option on vol futures is
sharply negative. The volatility of a vol future is significantly less than the volatility of
the vol index, but does converge as it approaches expiry (when it becomes as volatile as
the vol index itself).

COMPARING IMPLIED VOL AND REALISED VOL IS DIFFICULT

The realised volatility of a vol future increases as it approaches expiry (near-dated
volatility is more volatile than far-dated volatility). The implied volatility of an option on
a vol future should trade roughly in line with the average realised volatility of the vol
future over the life of the option. Hence the average realised volatility of a vol future
will be a blend of the initial low realised volatility, and the higher realised volatility close
to expiry. The implied volatility term structure of an option on vol future will therefore
be less negative than the realised volatility term structure of the vol future.

Realised of vol futures < implied of option on vol futures < realised of vol
index
The implied volatility level of options on vol futures is also higher than the realised
volatility of the vol future for that expiry (eg, implied of 6-month option on vol futures
is above current realised of 6-month vol future). The implied volatility of options on vol
futures will, however, be less than the realised volatility of the vol index, which makes
options on vol futures look cheap if an investor mistakenly compares its implied to the
realised of the vol index.

OPTION ON VOL FUTURE CHEAPER THAN OPTIONS ON VAR

Implied volatility is less volatile than realised volatility, as implied volatility will never
trade at the min or max level of realised (as it is an estimate of future volatility, and there
is never a time that the market can be 100% certain realised will reach its min or max).
As implied volatility is less volatile than realised volatility, an option on a vol future
should be at a lower implied than an option on realised variance (particularly for near-
dated expiries). They will, however, have a similar negative term structure.

17On average slightly upward sloping, but at a far shallower gradient to the negative term structure of
options on vol futures.
CHAPTER 5
LIGHT EXOTICS

Advanced investors can make use of more exotic equity derivatives. Some of the most
popular are light exotics, such as barriers, worst-of/best-of options, outperformance
options, look-back options, contingent premium options, composite options and quanto
options.
134 CHAPTER 5: LIGHT EXOTICS

5.1: BARRIER OPTIONS

Barrier options are the most popular type of light exotic product, as they are used
within structured products or to provide cheap protection. The payout of a barrier
option knocks in or out depending on whether a barrier is hit. There are eight types
of barrier option, but only four are commonly traded, as the remaining four have a
similar price to vanilla options. Barrier puts are more popular than calls (due to
structured product and protection flow), and investors like to sell visually expensive
knock-in options and buy visually cheap knock-out options. Barrier options (like all
light exotics) are always European (if they were American, the price would be
virtually the same as a vanilla option, as the options could be exercised just before
the barrier was hit).

BARRIER OPTIONS CAN HAVE DELTA OF MORE THAN ±100%

The hedging of a barrier option is more involved than for vanilla options, as the delta near
the barrier can be significantly more than ±100% near expiry. The extra hedging risk of
barriers widens the bid-offer spread in comparison with vanilla options. Barrier options are
always European and are traded OTC.

THERE ARE THREE KEY VARIABLES FOR BARRIER OPTIONS

There are three key variables to a barrier option, each of which has two possibilities. These
combinations give eight types of barrier option (8=2×2×2).

 Down/up. The direction of the barrier in relation to spot. Almost all put barriers are
down barriers and, similarly, almost all call barriers are up barriers.
 Knock in/out. Knock-out (KO) options have a low premium and give the impression
of being cheap; hence, they are usually bought by investors. Conversely, knock-in (KI)
options are visually expensive (as knock-in options are a similar price to a vanilla) and
are usually sold by investors (through structured products). For puts, a knock-in is the
most popular barrier (structured product selling of down and knock-in puts). However,
for calls this is reversed and knock-outs are the most popular. Recent volatility has made
knock-out products less popular than they once were, as many hit their barrier and
became worthless.
 Put/call. The type of payout of the option. Put barriers are three to four times more
popular than call barriers, due to the combination of selling from structured products
(down and knock-in puts) and cheap protection buying (down and knock-out puts).
138 CHAPTER 5: LIGHT EXOTICS

KNOCK-OUTS COST C15%-25% OF PUT SPREAD COST

The payout of a knock-out put is equal to a ‘shark fin’ (see top left chart in Figure 83 above)
until the barrier is reached. A ‘shark fin’ is equal to a short digital position (at the barrier)
plus a put spread (long put at strike of knock-out put, short put at barrier of knock-out put).
The price of a knock-out put can therefore be considered to be the cost of a put spread, less
a digital and less the value of the knock-out. As pricing digitals and barriers is not trivial,
comparing the price of a knock-out put as a percentage of the appropriate put spread can be
a quick way to evaluate value (the knock-out will have a lower value as it offers less payout
to the downside). For reasonable barriers between 10% and 30% below the strike, the price
of the knock-out option should be between c15% and c25% of the cost of the put spread.

CONTINUOUS BARRIERS ARE CHEAPER THAN DISCRETE

There are two types of barriers, continuous and discrete. A continuous barrier is triggered if
the price hits the barrier intraday, whereas a discrete barrier is only triggered if the closing
price passes through the barrier. Discrete knock-out barriers are more expensive than
continuous barriers, while the reverse holds for knock-in barriers (especially during periods
of high volatility). There are also additional hedging costs to discrete barriers, as it is possible
for spot to move through the barrier intraday without the discrete barrier being triggered (ie,
if the close is the correct side of the discrete barrier). As these costs are passed on to the
investor, discrete barriers are far less popular than continuous barriers for single stocks
(c10%-20% of the market), although they do make up almost half the market for indices.

Jumps in stock prices between close and open is a problem for all barriers
While the hedge for a continuous barrier should, in theory, be able to be executed at a level
close to the barrier, this is not the case should the underlying jump between close and open.
In this case, the hedging of a continuous barrier suffers a similar problem to the hedging of a
discrete barrier (delta hedge executed at a significantly different level to the barrier).

DOUBLE BARRIERS ARE POSSIBLE, BUT RARE

Double barrier options have both an up barrier and a down barrier. As only one of the
barriers is significant for pricing, they are not common (as their pricing is similar to an
ordinary single-barrier option). They make up less than 5% of the light exotic market.
5.1: Barrier Options 139

REBATES CAN COMPENSATE FOR TRIGGER OF BARRIER

The main disadvantage of knock-out barrier options is that the investor receives nothing for
purchasing the option if they are correct about the direction of the underlying (option is
ITM) but incorrect about the magnitude (underlying passes through barrier). In order to
provide compensation, some barrier options give the long investors a rebate if the barrier is
triggered: for example, an ATM call with 120% knock-out that gives a 5% rebate if the
barrier is touched. Rebates comprise approximately 20% of the index barrier market but are
very rare for single-stock barrier options.
140 CHAPTER 5: LIGHT EXOTICS

5.2: WORST-OF/BEST-OF OPTIONS

Worst-of (or best-of) options give payouts based on the worst (or best) performing
asset. They are the second most popular light exotic due to structured product flow.
Correlation is a key factor in pricing these options, and investor flow typically buys
correlation (making uncorrelated assets with low correlation the most popular
underlyings). The underlyings can be chosen from different asset classes (due to low
correlation), and the number of underlyings is typically between three and 20. They
are always European, and normally ATM options.

NORMALLY 1 YEAR MATURITY AND CAN BE CALLS OR PUTS

Worst-of/best-of options can be any maturity. Although the most popular is one-year
maturity, up to three years can trade. As an option can be a call or a put, and either ‘worst-of’
or ‘best-of’; there are four types of option to choose from. However, the most commonly
traded are worst-of options (call or put). The payouts of the four types are given below:

Worst-of call = Max (Min (r 1 , r 2 ,... , r N ) , 0) where r i is the return of N assets

Worst-of put = Max (-Min (r 1 , r 2 ,... , r N ) , 0) where r i is the return of N assets

Best-of call = Max (Max (r 1 , r 2 ,... , r N ) , 0) where r i is the return of N assets

Best-of put = Max (-Max (r 1 , r 2 ,... , r N ) , 0) where r i is the return of N assets

WORST-OF CALLS ARE POPULAR TO BUY (AS CHEAP)

The payout of a worst-of call option will be equal to the lowest payout of individual call
options on each of the underlyings. As it is therefore very cheap, they are popular to buy. If
all the assets are 100% correlated, then the value of the worst-of call is equal to the value of
calls on all the underlyings (hence, in the normal case of correlation less than 100%, a worst-
of call will be cheaper than any call on the underlying). If we lower the correlation, the price
of the worst-of call also decreases (eg, the price of a worst-of call on two assets with -100%
correlation is zero, as one asset moves in the opposite direction to the other). A worst-of call
option is therefore long correlation. As worst-of calls are cheap, investors like to buy them
and, therefore, provide buying pressure to implied correlation.
142 CHAPTER 5: LIGHT EXOTICS

LIGHT EXOTIC OPTIONS FLOW LIFTS IMPLIED CORRELATION

As the flow from worst-of/best-of products tends to support the levels of implied
correlation, implied correlation typically trades above fair value. While other light exotic flow
might not support correlation (eg, outperformance options, which are described below),
worst-of/best-of options are the most popular light exotic, whose pricing depends on
correlation and are therefore the primary driver for this market. We would point out that the
most popular light exotics – barrier options – have no impact on correlation markets. In
addition, worst-of/best-of flow is concentrated in uncorrelated assets, whereas
outperformance options are usually on correlated assets.
5.3: Outperformance Options 143

5.3: OUTPERFORMANCE OPTIONS

Outperformance options are an option on the difference between returns on two
different underlyings. They are a popular method of implementing relative value
trades, as their cost is usually cheaper than an option on either underlying. The key
unknown parameter for pricing outperformance options is implied correlation, as
outperformance options are short correlation. The primary investor base for
outperformance options is hedge funds, which are usually buyers of outperformance
options on two correlated assets (to cheapen the price). Outperformance options are
European and can always be priced as a call. Unless they are struck with a hurdle,
they are an ATM option.

OUTPERFORMANCE OPTIONS ARE SHORT-DATED CALLS

Outperformance options give a payout based on the difference between the returns of two
underlyings. While any maturity can be used, they tend to be for maturities up to a year
(maturities less than three months are rare). The payout formula for an outperformance
option is below – by convention always quoted as a call of ‘r A over r B’ ’ (as a put of ‘r A over
r B ’ can be structured as a call on ‘r B over r A ’). Outperformance options are always European
(like all light exotics) and are traded OTC.

Payout = Max (r A – r B , 0) where r A and r B are the returns of assets A and B,

respectively

OPTIONS CAN HAVE HURDLE AND ALLOWABLE LOSS

While outperformance options are normally structured ATM, they can be cheapened by
making it OTM through a hurdle or by allowing an allowable loss at maturity (which simply
defers the initial premium to maturity). While outperformance options can be structured
ITM by having a negative hurdle, as this makes the option more expensive, this is rare. The
formula for outperformance option payout with these features is:

Payout = Max (r A – r B – hurdle, – allowable loss)

OUTPERFORMANCE OPTIONS ARE SHORT CORRELATION

The pricing of outperformance options depends on both the volatility of the two
underlyings and the correlation between them. As there tends to be a more liquid and visible
market for implied volatility than correlation, it is the implied correlation that is the key
factor in determining pricing. Outperformance options are short correlation, which can be
intuitively seen as: the price of an outperformance option must decline to zero if one
assumes correlation rises towards 100% (two identical returns give a zero payout for the
outperformance option).
144 CHAPTER 5: LIGHT EXOTICS

As flow is to the buy side, some hedge funds outperformance call overwrite
Outperformance options are ideal for implementing relative value trades, as they benefit
from the upside, but the downside is floored to the initial premium paid. The primary
investor base for outperformance options are hedge funds. While flow is normally to the buy
side, the overpricing of outperformance options due to this imbalance has led some hedge
funds to call overwrite their relative value position with an outperformance option.

MARGRABE’S FORMULA CAN BE USED FOR PRICING

An outperformance option volatility σ A-B can be priced using Margrabe’s formula given the
inputs of the volatilities σ A and σ B of assets A and B, respectively, and their correlation ρ.
This formula is shown below.

σ A − B = σ A 2 + σ B 2 − 2 ρσ Aσ B

TEND TO BE USED FOR CORRELATED ASSETS

The formula above confirms mathematically that outperformance options are short
correlation (due to the negative sign of the final term with correlation ρ). From an investor
perspective, it therefore makes sense to sell correlation at high levels; hence, outperformance
options tend to be used for correlated assets (so cross-asset outperformance options are very
rare). This is why outperformance options tend to be traded on indices with a 60%-90%
correlation and on single stocks that are 30%-80% correlated. The pricing of an
outperformance option offer tends to have an implied correlation 5% below realised for
correlations of c80%, and 10% below realised for correlations of c50% (outperformance
option offer is a bid for implied correlation).

Best pricing is with assets of similar volatility

The price of an outperformance is minimised if volatilities σ A and σ B of assets A and B are
equal (assuming the average of the two volatilities is kept constant). Having two assets of
equal volatility increases the value of the final term 2ρσ A σ B (reducing the outperformance
volatility σ A-B ).

LOWER FORWARD FLATTERS OUTPERFORMANCE PRICING

Assuming that the two assets have a similar interest rate and dividends, the forwards of the
two assets approximately cancel each other out, and an ATM outperformance option is also
ATMf (ATM forward or At The Money Forward). When comparing relative costs of
outperformance options with call options on the individual underlyings, ATMf strikes must
be used. If ATM strikes are used for the individual underlyings, the strikes will usually be
lower than ATMf strikes and the call option will appear to be relatively more expensive
compared to the ATMf (= ATM) outperformance option.
5.3: Outperformance Options 145

Pricing of ATM outperformance options is usually less than ATMf on either

underlying
If two assets have the same volatility (σ A = σ B ) and are 50% correlated (ρ = 50%), then the
input for outperformance option pricing σ A-B is equal to the volatilities of the two
underlyings (σ A-B = σ A = σ B ). Hence, ATMf (ATM forward) options on either underlying
will be the same as an ATMf (≈ATM) outperformance option. As outperformance options
tend to be used on assets with higher than 50% correlation and whose volatilities are similar,
outperformance options are usually cheaper than similar options on either underlying.
146 CHAPTER 5: LIGHT EXOTICS

5.4: LOOK-BACK OPTIONS

There are two types of look-back options, strike look-back and payout look-back,
and both are usually multi-year options. Strike reset (or look-back) options have their
strike set to the highest, or lowest, value within an initial look-back period (of up to
three months). These options are normally structured so the strike moves against the
investor in order to cheapen the cost. Payout look-back options conversely tend to be
more attractive and expensive than vanilla options, as the value for the underlying
used is the best historical value. As with all light exotics, these options are European
and OTC.

STRIKE OF RESET OPTIONS MOVES AGAINST INVESTOR

There are two main strike reset options, and both have an initial look-back period of
typically one to three months, where the strike is set to be the highest (for a call) or lowest
(for a put) traded value. While the look-back optionality moves against the investor, as the
expiry of these options is multi-year (typically three), there is sufficient time for spot to move
back in the investor’s favour, and the strike reset cheapens the option premium. While
having a strike reset that moves the strike to be the most optimal for the investor is possible,
the high price means they are unpopular and rarely trade. While the cheaper form of strike
reset options does attract some flow due to structured products, they are not particularly
popular.

Strike resets are cheaper alternative to ATM option at end of reset period
There are three possible outcomes to purchasing a strike reset option. Strike reset options
can be considered a cheaper alternative to buying an ATM option at the end of the strike
reset period, as the strike is roughly identical for two of the three possible outcomes (but at a
lower price).

 Spot moves in direction of option payout. If spot moves in a direction that would
make the option ITM, the strike is reset to be equal to spot as it moves in a favourable
direction, and the investor is left with a roughly ATM option.
 Range-trading markets. Should markets range trade, the investor will similarly receive
a virtually ATM option at the end of the strike reset period.
 Spot moves in opposite direction to option payout. If spot initially moves in the
opposite direction to the option payout (down for calls, up for puts), then the option
strike is identical to an option that was initially ATM (as the key value of the underlying
for the strike reset is the initial value) and, hence, OTM at the end of the strike rest
period. The downside of this outcome is why strike reset options can be purchased for a
lower cost than an ATM option.
5.4: Look-Back Options 147

Strike resets perform best when there is an initial period of range trading
Strike reset options are therefore most suitable for investors who believe there will be an
initial period of range trading, before the underlying moves in a favourable direction.

PAYOUT LOOK-BACK OPTIONS

Having a look-back option that selects the best value of the underlying (highest for calls,
lowest for puts) increases the payout of an option – and cost. These options typically have a
five-year maturity and typically use end-of-month or end-of-year values for the selection of
the optimal payout.
148 CHAPTER 5: LIGHT EXOTICS

5.5: CONTINGENT PREMIUM OPTIONS

Contingent premium options are initially zero-premium and only require a premium
to be paid if the option becomes ATM on the close. The contingent premium to be
paid is, however, larger than the initial premium would be, compensating for the fact
that it might never have to be paid. Puts are the most popular, giving protection with
zero initial premium. These typically one-year put options are OTM (or the
contingent premium would almost certainly have to be paid immediately) and
European.

CONTINGENT PREMIUM OPTION HAS ZERO UPFRONT COST

While contingent premium calls are possible, the most popular form is for a contingent
premium put to allow protection to be bought with no initial cost. The cost of the premium
to be paid is roughly equal to the initial premium of the vanilla option, divided by the
probability of spot trading through the strike at some point during the life of the option (eg,
an 80% put whose contingent premium has to be paid if the underlying goes below 80%).
Using contingent premium options for protection has the benefit that no cost is suffered if
the protection is not needed, but if spot dips below the strike/barrier, then the large
premium has to be paid (which is likely to be more than the put payout unless there was a
large decline). These can be thought of as a form of ‘crash put’.

Conditional premium on a level other than strike is possible, but rare

The usual structure for contingent premium options is to have the level at which the
premium is paid equal to the strike. The logic is that although investors have to pay a large
premium, they do have the benefit of holding an option that is slightly ITM. Having the
conditional premium at a level other than strike is possible, but rare (eg, an 80% put whose
contingent premium has to be paid if the underlying reaches 110%).
5.6: Composite and Quanto Options 149

5.6: COMPOSITE AND QUANTO OPTIONS

There are two types of option involving different currencies. The simplest is a
composite option, where the strike (or payoff) currency is in a different currency to
the underlying. A slightly more complicated option is a quanto option, which is
similar to a composite option, but the exchange rate of the conversion is fixed.

COMPOSITE OPTIONS USE DIFFERENT VOLATILITY INPUT

A composite option is a cash or physical option on a security whose currency is different
from the strike or payoff currency (eg, Euro strike option on Apple). If an underlying is in a
foreign currency, then its price in the payout (or strike) currency will usually be more volatile
(and hence more expensive) due to the additional volatility associated with currency
fluctuations. Only for significantly negative correlations will a composite option be less
expensive than the vanilla option (if there is zero correlation the effect of FX still lifts
valuations). The value of a composite option can be calculated using Black-Scholes as usual,
by substituting the volatility of the asset with the volatility of the asset in payout currency
terms. The payout (or strike) currency risk-free rate should be used rather than the (foreign)
security currency risk-free rate. The dividend yield assumption is unchanged (as it has no
currency) between a composite option and a vanilla option.

σ Payout = σ Security 2 + σ FX 2 + 2 ρσ Securityσ FX

where

σ Payout = volatility of asset in payout (strike) currency

σ Security = volatility of asset in (foreign) security currency

σ FX = volatility of FX rate (between payout currency and security currency)

ρ = correlation of FX rate (security currency in payoff currency terms) and security price
150 CHAPTER 5: LIGHT EXOTICS

Composite options are long correlation (if FX is foreign currency in domestic

terms)
The formula to calculate the volatility of the underlying is given above. As the payoff
increases with a positive correlation between FX and the underlying, a composite option is
long correlation (the positive payout will be higher due to FX, while FX moving against the
investor is irrelevant when the payout is zero). Note that care has to be taken when
considering the definition of the FX rate; it should be the (foreign) security currency given in
(domestic) payoff currency terms.

For example, if we are pricing a euro option on a dollar-based security and assume an
extreme case of ρ = 100%, the volatility of the USD underlying in EUR will be the sum of
the volatility of the underlying and the volatility of USD.

QUANTO OPTIONS USE DIFFERENT DIVIDEND INPUT

Quanto options are similar to a composite option, except the payout is always cash settled
and a fixed FX rate is used to determine the payout. Quanto options can be modelled using
Black-Scholes. As the FX rate for the payout is fixed, quanto options are modelled using the
normal volatility of the underlying (as FX volatility has no effect). The payout is simply the
fixed FX rate multiplied by the price of a vanilla option with the same volatility, but a
different carry. The carry (risk-free rate - dividend) to be used is shown below (the risk-free
rate for quanto options is assumed to be the risk-free rate of the security currency, ie, it is
not the same as for composite options).

cQuanto = rfrSecurity − d − ρσ Securityσ FX

 dQuanto = d + ρσ Securityσ FX as d quanto = rfr Security - c Quanto

where

c Quanto = carry for quanto pricing

d Quanto = dividend for quanto pricing

d = dividend yield

rfr Security = risk free rate of security currency

rfr Payout = risk free rate of payout currency

5.6: Composite and Quanto Options 151

Quanto options are either long or short correlation depending on the sign of
the delta
The correlation between the FX and the security has an effect on quanto pricing, the
direction (and magnitude) of which depends on the delta of the option. This is because the
dividend risk of an option is equal to its delta, and the dividend used in quanto pricing
increases as correlation increases.

Quanto option calls are short correlation (if FX is foreign currency in

domestic terms)
As a call option is short dividends (call is an option on the price of underlying, not the total
return of the underlying), a quanto call option is short correlation. A quanto put option is
therefore slightly long correlation. In both cases, we assume the FX rate is the foreign
security currency measured in domestic payout terms.

Intuitively, we can see a quanto call option is short correlation by assuming the dividend
yield and both currency risk-free rates are all zero and comparing its value to a vanilla call
option priced in the (foreign) security currency. If correlation is high, the vanilla call option
is worth more than the quanto call option (as FX moves in favour of the investor if the price
of the security rises). The reverse is also true (negative correlation causes a vanilla call option
to be worth less than a quanto call option). As the price of a vanilla (single currency) call
does not change due to the correlation of the underlying with the FX rate, this shows a
quanto call option is short correlation.
CHAPTER 6
RELATIVE VALUE AND CORRELATION
TRADING

Advanced investors often use equity derivatives to gain different exposures; for
example, relative value or the jumps on earnings dates. We demonstrate how this can
be done and also reveal how profits from equity derivatives are both path dependent
and dependent on the frequency of delta hedging.
154 CHAPTER 6: RELATIVE VALUE AND CORRELATION TRADING

6.1: RELATIVE VALUE TRADING

Relative value is the name given to a variety of trades that attempt to profit from the
mean reversion of two related assets that have diverged. The relationship between
the two securities chosen can be fundamental (different share types of same
company or significant cross-holding) or statistical (two stocks in same sector).
Relative value can be carried out via cash (or delta-1), options or outperformance
options.

TRADES ARE USUALLY CHOSEN ON CORRELATED ASSETS

The payout of a relative value trade on two uncorrelated securities is completely random,
and the investor on average gains no benefit. However, if two securities have a strong
fundamental or statistical reason to be correlated, they can be thought of as trading in a
similar direction with a random noise component. Assuming the correlation between the
securities is sufficiently strong, the noise component should mean revert. Relative value
trades attempt to profit from this mean reversion. There are five main types of relative value
trades.

 Dual listing. If a share trades on different exchanges (eg, an ADR), the two prices
should be equal. This is not always the case due to execution risk (different trading
times) and perhaps due to indexation flow. Non-fungible shares or those with shorting
restrictions are most likely to show the largest divergence in price. Of all relative value
trades, dual-listing ones are likely to show the strongest correlation.

 Share class. If there is more than one type of share, perhaps with voting or ownership
restrictions, then the price of these shares can diverge from one another. For example,
preference shares typically have a higher dividend to compensate for lack of voting
rights, but suffer from less liquidity and (normally) exclusion from equity indices. During
special situations, for example, during the Porsche/VW saga, the difference in price
between the two shares can diverge dramatically.

 Cross-holding. If one company (potentially a holding company) owns a significant

amount of another company, the prices of the two companies will be linked. Sometimes
putting on a cross-holding trade is difficult in practice due to the high borrow cost of
the smaller company. This trade is also known as a stub trade when the investor wants
pure exposure to the larger company, and hedges out the unwanted exposure to the
equity holdings of the larger company. Potentially, these trades can occur when a larger
company spins off a subsidiary but keeps a substantial stake post spin-off.
6.1: Relative Value Trading 155

 Event-driven. In the event of a takeover that is estimated to have a significant chance

of succeeding, the share prices of the acquiring and target company should be
correlated. The target will usually trade at a discount to the bid price, to account for the
probability the deals falls through (although if the offer is expected to be improved, or
beaten by another bidder, the target could trade above the offer price).

 Long-short. If a long and short position is initiated in two securities that do not have
one of the above four reasons to be correlated, it is a long-short trade. The correlation
between the two securities of a long-short trade is likely to be lower than for other
relative values trades. Because of this, often two stocks within a sector are chosen, as
they should have a very high correlation and the noise component is likely to be
bounded (assuming market share and profitability is unlikely to change substantially over
the period of the relative value trade).

Long-short can focus returns on stock picking ability (which is c10% of equity
return)
General market performance is typically responsible for c70% of equity returns, while c10%
is due to sector selection and the remaining c20% due to stock picking. If an investor wishes
to focus returns on the proportion due to sector or stock picking, they can enter into a long
position in that security and a short position in the appropriate market index (or vice versa).
This will focus returns on the c30% due to sector and stock selection. Typically, relatively
large stocks are selected, as their systematic risk (which should cancel out in a relative value
trade) is usually large compared to specific risk. Alternatively, if a single stock in the same
sector (or sector index) is used instead of the market index, then returns should be focused
on the c20% due to stock picking within a sector.

SIZE OF POSITIONS SHOULD BE WEIGHTED BY BETA

If the size of the long-short legs are chosen to have equal notional (share price × number of
shares × FX), then the trade will break even if both stock prices go to zero. However, the
legs of the trade are normally weighted by beta to ensure the position is market neutral for
more modest moves in the equity market. The volatility (historical or implied) of the stock
divided by the average volatility of the market can be used as an estimate of the beta.

DELTA-1, OPTIONS AND OUTPERFORMANCE OPTIONS

Relative value trades can be implemented via cash/delta-1, vanilla options or
outperformance options. They have very different trade-offs between liquidity and risk. No
one method is superior to others, and the choice of which instrument to use depends on the
types of liquidity and risk the investor is comfortable with.
6.1: Relative Value Trading 157

Weighting options by volatility is similar to weighting by beta and roughly

zero cost
The most appropriate weighting for two relative value legs is beta weighting the size of the
delta hedge of the option (ie, same beta × number of options × delta × FX), rather than
having identical notional (share price × number of options × FX). Beta weighting ensures
the position is market neutral. Volatility weighting can be used as a substitute for beta
weighting, as volatility divided by average volatility of the market is a reasonable estimate for
beta. Volatility weighting ATM (or ATMf) options is roughly zero cost, as the premium of
ATM options is approximately linear in volatility.

Choosing strike and maturity of option is not trivial

One disadvantage of using options in place of equity is the need to choose a maturity,
although some investors see this as an advantage as it forces a view to be taken on the
duration or exit point of the trade at inception. If the position has to be closed or rolled
before expiry, there are potentially mark-to-market risks. Similarly, the strike of the option
needs to be chosen, which can be ATM (at the money), ATMf (ATM forward), same
percentage of spot/forward or same delta. Choosing the same delta of an OTM option
means trading a strike further away from spot/forward for the more volatile asset (as delta
increases as volatility increases). We note that trading the same delta option is not the same
as volatility weighting the options traded as delta is not linear in volatility. Delta also does
not take into account the beta of the underlyings.

(3) OUTPERFORMANCE OPTIONS: LIMITED DOWNSIDE BUT LOW

LIQUIDITY

Outperformance options are ideally suited to relative value trades, as the maximum loss is
the premium paid and the upside is potentially unlimited. However, outperformance options
suffer from being relatively illiquid. While pricing is normally cheaper than vanilla options
(for normal levels of correlation), it might not be particularly appealing depending on the
correlation between the two assets. As there are usually more buyers than sellers of
outperformance options, some hedge funds use outperformance options to overwrite their
relative value trades.
158 CHAPTER 6: RELATIVE VALUE AND CORRELATION TRADING

6.2: RELATIVE VALUE VOLATILITY TRADING

Volatility investors can trade volatility pairs in the same way as trading equity pairs.
For indices, this can be done via options, variance swaps or futures on a volatility
index (such as the VIX or vStoxx). For indices that are popular volatility trading
pairs, if they have significantly different skews this can impact the volatility market.
Single-stock relative value volatility trading is possible, but less attractive due to the
wider bid-offer spreads.

TWO WAYS TO PROFIT FROM VOLATILITY PAIR TRADING

When a pair trade between two equities is attempted, the main driver of profits is from a
mean reversion of the equity prices. With volatility relative value trading, there are two ways
of profiting:

 Mean reversion. In the same way an equity pair trade profits from a mean reversion of
stock prices, a volatility pair trade can profit from a mean reversion of implied volatility.
For short-term trades, mean reversion is the primary driver for profits (or losses). For
relative value trades using forward starting products (eg, futures on volatility indices),
this is the only driver of returns as forward starting products have no carry. The method
for finding suitable volatility pair trades that rely on a short-term mean reversion are
similar to that for a vanilla pair trade on equities.
 Carry. For an equity pair trade, the carry of the position is not as significant as, typically,
the dividend yields of equities do not differ much from one another and are relatively
small compared to the movement in spot. However, the carry of a volatility trade
(difference between realised volatility and implied volatility) can be significant. As the
duration of a trade increases, the carry increases in importance. Hence, for longer term
volatility pair trades it is important to look at the difference between realised and implied
volatility.

IMPLIED VOL SPREAD BETWEEN PAIRS IS KEPT STABLE

While the skew of different indices is dependent on correlation, traders tend to keep the
absolute difference in implied volatility stable due to mean reversion. This is why if equity
markets move down, the implied volatility of the S&P500 or FTSE (as they are large
diversified indices that hence have high skew) tends to come under pressure, while the
implied volatility of country indices with fewer members, such as the DAX, are likely to be
supported. The SX5E tends to lie somewhere in between, as it has fewer members than the
S&P500 or FTSE but is more diverse than other European country indices. Should markets
rise, the reverse tends to occur (high skew indices implieds are lifted, low skew implieds are
weighed on).
160 CHAPTER 6: RELATIVE VALUE AND CORRELATION TRADING

PAIR TRADES CAN BE THETA OR VEGA WEIGHTED

Theta and vega weighted are the most common methods of weighting pair trades. Dollar
gamma weighted is rarely used and is included for completeness purposes only. Theta-
weighted trades assume proportional volatility changes (eg, if stock A has 20% implied and
stock B has 25% implied, if stock A rises from 20% to 30% implied that is a 50% increase
and stock B rises 50% to 37.5% implied). Vega-weighted trades assume absolute volatility
changes (eg, if stock A has 20% implied and stock B has 25% implied, if stock A rises from
20% to 30% that is a 10 volatility point increase and stock B rises 10 volatility points to 35%
implied).

Pair trade between two securities of same type should be theta weighted
If a pair trade between two securities of the same type (ie, two indices, or two single stocks)
is attempted, theta weighting is the most appropriate. This is because the difference between
a low volatility security and a high volatility security (of the same type) usually increases as
volatility increases (ie, a proportional move). If a pair trade between an index and a single
stock is attempted, vega weighting is the best as the implied volatility of an index is
dependent not only on single-stock implied volatility but also on implied correlation. As
volatility and correlation tend to move in parallel, this means the payout of a vega-weighted
pair trade is less dependent on the overall level of volatility (hence the volatility mispricing
becomes a more significant driver of the P&L of the trade) 18.

18 There is evidence to suggest that vega-weighted index vs single-stock pair trades on average

associate 2%-5% too much weight to the single-stock leg compared to the index leg. However, as this
is so small compared to stock specific factors, it should be ignored.
6.3: Correlation Trading 163

where
σI = index volatility

σi = single stock volatility (of ith member of index)

wi = single stock weight in index (of ith member of index)

n = number of members of index

INDEX CORRELATION CAN BE ESTIMATED FROM VARIANCE

If the correlation of all the different members of an index is assumed to be identical (a
heroic assumption, but a necessary one if we want to have a single measure of correlation),
the correlation implied by index and single-stock implied volatility can be estimated as the
variance of the index divided by the weighted average single-stock variance. This measure is
a point or two higher than the actual implied correlation but is still a reasonable
approximation.

σI2
ρimp =
∑
n 2
i =1
wiσ i

where

ρ imp = implied correlation (assumed to be identical between all index members)

Proof implied correlation can be estimated by index variance divided by

single stock variance
The formula for calculating the index volatility from the members of the index is given
below.

σ I 2 = ∑i =1 wi 2σ i 2 + ∑i =1, j ≠ i wi w jσ iσ j ρij
n n

where

ρ ij = correlation between single stock i and single stock j

If we assume the correlations between each stock are identical, then this correlation can be
implied from the index and single stock volatilities.
164 CHAPTER 6: RELATIVE VALUE AND CORRELATION TRADING

σ I 2 − ∑i =1 wi 2σ i 2
n

ρimp =
∑
n
i =1, j ≠ i
wi w jσ iσ j

Assuming reasonable conditions (correlation above 15%, c20 members or more, reasonable
weights and implied volatilities), this can be rewritten as the formula below.

σI2
ρimp =
(∑i =1 wiσ i ) 2
n

This can be approximated by the index variance divided by the weighted average single-stock
variance.

σI2
ρimp ≈ eg, if index variance=20% and members average variance=25%,
∑
n 2
wσ
i =1 i i
ρ≈64%.

This approximation is slightly too high (c2pts) due to Jensen’s inequality (shown below).

(∑i =1 wiσ i ) 2 ≤ ∑i =1 wiσ i

n n 2

STRUCTURED PRODUCTS LIFT IMPLIED CORRELATION

Using correlation to visually cheapen payouts through worst-of/best-of options is common
practice for structured products. Similarly, the sale of structured products, such as Altiplano
(which receives a coupon provided none of the assets in the basket has fallen), Everest (payoff on
the worst performing) and Himalayas (performance of best share of index), leave their
vendors short implied correlation. This buying pressure tends to lift implied correlation
above fair value. We estimate that the correlation exposure of investment banks totals
c€200mn per percentage point of correlation. The above formulae can show that two
correlation points is equivalent to 0.3 to 0.5 (single-stock) volatility points. Similarly, the fact
that institutional investors tend to call overwrite on single stocks but buy protection on an
index also leads to buying pressure on implied correlation. The different methods of trading
correlation are shown below.

 Dispersion trading. Going short index implied volatility and going long single-stock
implied volatility is known as a dispersion trade. As a dispersion trade is short Volga, or
vol of vol, the implied correlation sold should be c10pts higher value than for a
6.3: Correlation Trading 165

correlation swap. A dispersion trade was historically put on using variance swaps, but
the large losses from being short single stock variance led to the single stock market
becoming extinct. Now dispersion is either put on using straddles, or volatility swaps.
Straddles benefit from the tighter bid-offer spreads of ATM options (variance swaps
need to trade a strip of options of every strike). Using straddles does imply greater
maintenance of positions, but some firms offer delta hedging for 5-10bp. A
disadvantage of using straddles is that returns are path dependent. For example, if half
the stocks move up and half move down, then the long single stocks are away from their
strike and the short index straddle is ATM.

 Correlation swaps. A correlation swap is simply a swap between the (normally equal
weighted) average pairwise correlation of all members of an index and a fixed amount
determined at inception. Market value-weighted correlation swaps are c5 correlation
points above equal weighted correlation, as larger companies are typically more
correlated than smaller companies. While using correlation swaps to trade dispersion is
very simple, the relative lack of liquidity of the product is a disadvantage. We note the
levels of correlation sold are typically c5pts above realised correlation.

 Covariance swaps. While correlation swaps are relatively intuitive and are very similar
to trading correlation via dispersion, the risk is not identical to the covariance risk of
structured product sellers (from selling options on a basket). Covariance swaps were
invented to better hedge the risk on structure books, and they pay out the correlation
multiplied by the volatility of the two assets.

 Basket options. Basket options (or options on a basket) are similar to an option on an index,
except the membership and weighting of the members does not change over time. The most
popular basket option is a basket of two equal weighted members, usually indices.

 Worst-of/best-of option. The pricing of worst-of and best-of options has a correlation
component. These products are discussed in the section 5.2 Worst-of/Best-of Options in the
Forward Starting Products and Light Exotics chapter.

 Outperformance options. Outperformance options pricing has as an input the

correlation between the two assets. These products are also discussed in the section 5.3
Outperformance Options in the Forward Starting Products and Light Exotics chapter.
6.3: Correlation Trading 169

where

Notional= notional paid (or received) per correlation point

ρ K = strike of correlation swap (agreed at inception of trade)

2
ρ = ρij (equal weight correlation swap)
n(n − 1)

∑
n
i =1, j > i
wi w j ρij
ρ = (market value weight correlation swap)
∑
n
ww
i =1, j > i i j

n = number of stocks in basket

Correlation swaps tend to trade c5 correlation points above realised

A useful rule of thumb for the level of a correlation swap is that it trades c5 correlation
points above realised correlation (either equal-weighted or market value-weighted, depending
on the type of correlation swap). However, for very high or very low values of correlation,
this formula makes less sense. Empirically, smaller correlations are typically more volatile
than higher correlations. Therefore, it makes sense to bump the current realised correlation
by a larger amount for small correlations than for higher correlations (correlation swaps
should trade above realised due to demand from structured products). The bump should
also tend to zero as correlation tends to zero, as having a correlation swap above 100%
would result in arbitrage (can sell correlation swap above 100% as max correlation is 100%).
Hence, a more accurate rule of thumb (for very high and low correlations) is given by the
formula below.

Correlation swap level = ρ + α (1 - ρ)

where

ρ = realised correlation

α = bump factor (typical α = 0.1)

Maturity of correlation swap is typically between one and three years

Structured products typically have a maturity of 5+ years; however, many investors close
their positions before expiry. The fact that a product can also delete a member within the
lifetime of the product has led dealers to concentrate on the three-year maturity rather than
170 CHAPTER 6: RELATIVE VALUE AND CORRELATION TRADING

five-year. As the time horizon of hedge funds is short dated, correlation swaps typically trade
between one and three years. The size is usually between €250k and €1,000k.

Correlation swaps suffer from lack of liquidity

The market for correlation swaps has always been smaller than for dispersion. Whereas the
variance swap or option market has other market participants who ensure liquidity and
market visibility, the investor base for correlation swaps is far smaller. This can be an issue
should a position wish to be closed before expiry. It can also cause mark-to-market
problems. The correlation swap market grew from 2002 onwards until the credit crunch,
when investor appetite for exotic products disappeared. At its peak, it is estimated that some
structured derivative houses shed up to c10% of their short correlation risk to hedge funds
using correlation swaps.

DISPERSION IS THE MOST POPULAR METHOD OF TRADING

CORRELATION
As the levels of implied correlation are usually overpriced (a side effect of the short
correlation position of structured product sellers), index implied volatility is expensive when
compared with the implied volatility of single stocks. A long dispersion trade attempts to
profit from this by selling index implied and going long single-stock implied 19. Such a long
dispersion trade is short implied correlation. While dispersion is the most common method
of trading implied correlation, the payoff is also dependent on the level of volatility. The
payout of (theta-weighted) dispersion is shown below. Because of this, and because
correlation is correlated to volatility, dispersion trading is short vol of vol (volga).

P & L dispersion = ∑i =1 wiσ i ( ρimp − ρ )

n 2

There are four instruments that can be used to trade dispersion:

 Straddle (or call) dispersion. Using ATM straddles to trade dispersion is the most
liquid and transparent way of trading. Because it uses options, the simplest and most
liquid volatility instrument, the pricing is usually the most competitive. Trading 90%
strike rather than ATM allows higher levels of implied correlation to be sold. Using
options is very labour intensive, however, as the position has to be delta-hedged (some
firms offer delta hedging for 5-10bp). In addition, the changing vega of the positions
needs to be monitored, as the risks are high given the large number of options that have
to be traded. In a worst-case scenario, an investor could be right about the correlation
position but suffer a loss from lack of vega monitoring. We believe that using OTM
strangles rather than straddles is a better method of using vanilla options to trade
dispersion as OTM strangles have a flatter vega profile. This means that spot moving

19 Less liquid members of an index are often excluded, eg, CRH for the Euro STOXX 50.
6.3: Correlation Trading 171

away from strike is less of an issue, but we acknowledge that this is a less practical way
of trading.
 Variance swap dispersion. Because of the overhead of developing risk management
and trading infrastructure for straddle dispersion, many hedge funds preferred to use
variance swaps to trade dispersion. With variance dispersion it is easier to see the profits
(or losses) from trading correlation than it is for straddles. Variance dispersion suffers
from the disadvantage that not all the members of an index will have a liquid variance
swap market. Since 2008, the single-stock variance market has disappeared due to the
large losses suffered from single-stock variance sellers (as dispersion traders want to go
long single-stock variance, trading desks were predominantly short single-stock
variance). It is now rare to be able to trade dispersion through variance swaps.
 Volatility swap dispersion. Since liquidity disappeared from the single-stock variance
market, investment banks have started to offer volatility swap dispersion as an
alternative. Excluding dispersion trades, volatility swaps rarely trade.
 Gamma swap dispersion. Trading dispersion via gamma swaps is the only ‘fire and
forget’ way of trading dispersion. As a member of an index declines, the impact on the
index volatility declines. As a gamma swap weights the variance payout on each day by
the closing price on that day, the payout of a gamma swap similarly declines with spot.
For all other dispersion trades, the volatility exposure has to be reduced for stocks that
decline and increased for stocks that rise. Despite the efforts of some investment banks,
gamma swaps never gained significant popularity.

Need to decide on weighting scheme for dispersion trades

While a dispersion trade always involves a short index volatility position and a long single-
stock volatility position, there are different strategies for calculating the ratio of the two trade
legs. If we assume index implied is initially 20%, if it increases to 30% the market could be
considered to have risen by ten volatility points or risen by 50%. If the market is considered
to rise by ten volatility points and average single-stock implied is 25%, it would be expected
to rise to 50% (vega-weighted). If the market is considered to rise by 50% and average
single-stock implied is 30%, it would be expected to rise to 45% (theta- or correlation-
weighted). The third weighting, gamma-weighted, is not often used in practice.

 Vega-weighted. In a vega-weighted dispersion, the index vega is equal to the sum of

the single-stock vega. If both index and single-stock vega rise one volatility point, the
two legs cancel and the trade neither suffers a loss or reveals a profit.
 Theta- (or correlation-) weighted. Theta weighting means the vega multiplied by
√variance (or volatility for volatility swaps) is equal on both legs. This means there is a
smaller single-stock vega leg than for vega weighting (as single-stock volatility is larger
than index volatility, so it must have a smaller vega for vega × volatility to be equal).
Under theta-weighted dispersion, if all securities have zero volatility, the theta of both
174 CHAPTER 6: RELATIVE VALUE AND CORRELATION TRADING

Theta-weighted dispersion is best weighting for almost pure correlation

exposure
The sole factor that determines if theta-weighted dispersion makes a profit or loss is the
difference between realised and implied correlation. For timing entry points for theta-
weighted dispersion, we believe investors should look at the implied correlation of an index
(as theta-weighted dispersion returns are driven by correlation). Note that theta-weighted
dispersion breaks even if single stock and index implied moves by the same percentage
amount (eg, index vol of 20%, single-stock vol of 25% and both rise 50% to 30% and
37.5%, respectively).

Vega-weighted dispersion gives hedged exposure to mispricing of correlation

When a dispersion trade is vega-weighted, it can be thought of as being the sum of a theta-
weighted dispersion (which gives correlation exposure), plus a long single-stock volatility
position. This volatility exposure can be thought of as a hedge against the short correlation
position (as volatility and correlation are correlated); hence, a vega-weighted dispersion gives
greater exposure to the mispricing of correlation. When looking at the optimal entry point
for vega-weighted dispersion, it is better to look at the difference between average single-
stock volatility and index volatility (as this applies an equal weight to both legs, like in a vega-
weighted dispersion). Note that vega-weighted dispersion breaks even if single stock and
index implied moves by the same absolute amount (eg, index vol of 20%, single-stock vol of
25% and both rise ten volatility points to 30% and 35%, respectively). Empirically, the
difference between single-stock and index volatility (ie, vega-weighted dispersion) is not
correlated to volatility 21, which supports our view of vega-weighted dispersion being the
best.

Gamma-weighted dispersion is rare, and not recommended

While gamma weighting might appear mathematically to be a suitable weighting for
dispersion, in practice it is rarely used. It seems difficult to justify a weighting scheme where
more single-stock vega is bought than index (as single stocks have a higher implied than
index and, hence, should move more). We include the details of this weighting scheme for
completeness, but do not recommend it.

21Single-stock leg is arguably 2%-5% too large; however, slightly over-hedging the implicit short
volatility position of dispersion could be seen as an advantage.
6.3: Correlation Trading 175

DISPERSION TRADES ARE SHORT VOL OF VOL (VOLGA)

The P&L of a theta-weighted dispersion trade is proportional to the spread between implied
and realized market value-weighted correlation (ρ), multiplied by a factor that corresponds to
a weighted average variance of the components of the index 22.

P & L theta weighted dispersion = ∑i =1 wiσ i ( ρimp − ρ )

n 2

where:

ρ = market value weighted correlation

The payout of a theta-weighted dispersion is therefore equal to the difference in implied and
realised correlation (market value-weighted pairwise realised correlation) multiplied by the
weighted average variance. If vol of vol was zero and volatility did not change, then the
payout would be identical to a correlation swap and both should have the same correlation
price. If volatility is assumed to be correlated to correlation (as it is, as both volatility and
correlation increase in a downturn) and the correlation component is profitable, the profits
are reduced (as it is multiplied by a lower volatility). Similarly, if the correlation suffers a loss,
the losses are magnified (as it is multiplied by a higher volatility). Dispersion is therefore
short volga (vol of vol) as the greater the change in volatility, the worse the payout. To
compensate for this short volga position, the implied correlation level of dispersion is c10
correlation points above the level of correlation swaps.

BASKET OPTIONS ARE MOST LIQUID CORRELATION PRODUCT

The most common product for trading correlation is a basket option (otherwise known as an
option on a basket). If the members of a basket are identical to the members of an index and
have identical weights, then the basket option is virtually identical to an option on the index.
The two are not completely identical, as the membership and weight of a basket option does
not change 23, but it can for an index (due to membership changes, rights issues, etc). The
formula for basket options is below.

∑
n
Basket = i =1
wi Si where S i is the ith security in the basket

∑
n 2
Basket call payoff at expiry = Max(0, i =1
wi Si − K ) where K is the strike

22Proof of this result is outside the scope of this publication.

23Weighting for Rainbow options is specified at maturity based on the relative performance of the
basket members, but discussion of these options is outside of the scope of this publication.
176 CHAPTER 6: RELATIVE VALUE AND CORRELATION TRADING

MOST POPULAR BASKET OPTIONS ARE ON TWO INDICES

While the above formula can be used for all types of basket, the most popular is a basket on
two equal weighted indices. In this case the correlation traded is not between multiple
members of a basket (or index) but the correlation between only two indices. As the options
usually wants the two indices to have identical value, it is easier to define the basket as the
equal weighted sum of the two security returns (see the below formula setting n = 2). The
previous formula could be used, but the weight w would not be 0.5 (would be 0.5 / S i at
inception ).

Si at expiry
∑
n
Basket = w where S i is the ith security (and w normally = 1/n)
i =1
Si at inception

PAYOFF IS BASED ON COVARIANCE, NOT CORRELATION

The payout of basket options is based on the correlation multiplied by the volatility of the
two securities, which is known as covariance. The formula for covariance is shown below.
As basket options are typically the payout of structured products, it is better to hedge the
exposure using products whose payout is also based on covariance. It is therefore better to
use covariance swaps rather than correlation swaps or dispersion to offset structured
product risk.

Covariance(A,B) = ρσ Aσ B where ρ is the correlation between A and B

COVARIANCE SWAPS BETTER REPRESENT STRUCTURED

PRODUCT RISK
The payout of structured products is often based on a basket option. The pricing of an
option on a basket involves covariance, not correlation. If an investment bank sells an
option on a basket to a customer and hedges through buying correlation (via correlation
swaps or dispersion) there is a mismatch 24. Because of this, attempts were made to create a
covariance swap market, but liquidity never took off.

Correlation swap payoff

[Covariance(A,B) - K covariance ] × Notional

where

Notional= notional paid (or received) per covariance point

24Results in being short cross-gamma. Cross-gamma is the effect a change in the value of one
underlying has on the delta of another.
6.3: Correlation Trading 177

ρ = correlation between A and B

σ i = volatility of i

Covariance(A,B) = ρσ Aσ B (note if A = B then covariance = variance as ρ = 1)

K covariance = strike of covariance swap (agreed at inception of trade)

178 CHAPTER 6: RELATIVE VALUE AND CORRELATION TRADING

6.4: TRADING EARNINGS

ANNOUNCEMENTS/JUMPS
From the implied volatilities of near dated options, it is possible to calculate the
implied jump on key dates. Trading these options in order to take a view on the
likelihood of unanticipated (low or high) volatility on reporting dates is a very
common strategy. We examine the different methods of calculating the implied
jump, and show how the jump calculation should normalise for index term structure.

TOTAL VOL = DIFFUSIVE VOL + JUMP VOL

While stock prices under Black-Scholes are modelled as having a GBM (Geometric
Brownian Motion) with constant volatility, in reality there are certain dates where there is
likely to be more volatility than average. These key dates are usually reporting dates, but
could also coincide with conference dates or investor days (in fact, any day where material
non-public information is released to the public). The implied volatility of an option whose
expiry is after a key date can be considered to be the sum of the normal diffusive volatility
(normal volatility for the stock in the absence of any significantly material information being
released) and the volatility due to the anticipated jump on the key date. While options of any
expiry after the key date could be used, we shall assume the expiry chosen is the expiry just
after the key date (to ensure the greatest percentage of the options’ time value is associated
with the jump).

Implied jumps can be traded as part of a relative value trade

Trading an option whose expiry is just after the key date gives exposure not only to the
implied jump, but also the normal diffusive volatility. If the expiry is far away then there
should be another expiry just before the key date. In this case the long position in the expiry
just after the key date can be hedged by shorting the expiry just before the key date. This
relative value trade assumes the investor wishes to be long the implied volatility jump. If an
investor wishes to short an expensive implied volatility jump then the reverse position (short
expiry just after key date, long expiry just before key date) should be put on.
180 CHAPTER 6: RELATIVE VALUE AND CORRELATION TRADING

Implied jumps normally calculated for near-dated events

Implied jumps are normally only calculated for near-dated events, as the effect of the jump
tends to be too diluted for far dated expiries (and hence would be untradeable taking bid-
offer spreads into account). Forward starting options could be used to trade jumps after the
first expiry, but the wider bid-offer spread could be greater than potential profits.

Forward volatility can be calculated with implied of two options

The calculation for forward volatility is derived from the fact variance (time weighted) is
additive. The formula is given below (σ x is the implied volatility for options of maturity T x ).

σ 2 2T2 − σ 12T1
σ 12 = = forward volatility T 1 to T 2
T2 − T1

JUMP VOLATILITY CAN BE CALCULATED FROM DIFFUSIVE

VOLATILITY
As variance is additive, the volatility due to the jump can be calculated from the total
volatility and the diffusive volatility. We note this assumes the normal diffusive volatility is
constant, whereas volatility just after a reporting date is, in fact, typically ¾ of the volatility
just before a reporting date (as the previously uncertain earnings are now known).

σ Expiry after jump 2T = σ Jump 2 + σ Diffusive 2 (T − 1)

 σ Jump = (σ Expiry after jump 2T − σ Diffusive 2 (T − 1)

where

σ Expiry after jump = implied volatility of option whose expiry is after the jump

T = time to the expiry after jump (= T 1 )

σ Diffusive = diffusive volatility (σ Before jump if there is an expiry before the jump, if not it is σ 12 )

σ Jump = implied volatility due to the jump

CHAPTER 7
SKEW AND TERM STRUCTURE TRADING

We examine how skew and term structure are linked and the effect on volatility
surfaces of the square root of time rule. The correct way to measure skew and smile is
examined, and we show how skew trades only breakeven when there is a static local
volatility surface.
184 CHAPTER 7: SKEW AND TERM STRUCTURE TRADING

7.1: SKEW AND TERM STRUCTURE ARE LINKED

When there is an equity market decline, there is normally a larger increase in ATM
implied volatility at the near end of volatility surfaces than the far end. Assuming
sticky strike, this causes near-dated skew to be larger than far-dated skew. The
greater the term structure change for a given change in spot, the higher skew is.
Skew is also positively correlated to term structure (this relationship can break down
in panicked markets). For an index, skew (and potentially term structure) is also
lifted by the implied correlation surface. Diverse indices tend to have higher skew for
this reason, as the ATM correlation is lower (and low strike correlation tends to 100%
for all indices).

SKEW & TERM STR CUT SURFACE IN DIFFERENT DIMENSIONS

A volatility surface has three dimensions (strike, expiry and implied volatility), which is
difficult to show on a two dimensional page. For simplicity, a volatility surface is often
plotted as two separate two dimensional graphs. The first plots implied volatility vs expiry
(similar to the way in which a yield curve plots credit spread against expiry) in order to show
term structure (the difference in implied volatility for options with different maturities and
the same strike). The second plots implied volatility vs strike to show skew (the difference in
implied volatility for options with different strikes and the same maturity). We examine a
volatility surface in both these ways (ie, term structure and skew) and show how they are
related.

TERM STRUCTURE IS NORMALLY UPWARD SLOPING

When there is a spike in realised volatility, near-dated implied volatility tends to spike in a
similar way (unless the spike is due to a specific event such as earnings). This is because the
high realised volatility is expected to continue in the short term. Realised volatility can be
expected to mean revert over a c8-month period, on average. Hence far-dated implied
volatilities tend to rise by a smaller amount than near-dated implied volatilities (as the
increased volatility of the underlying will only last a fraction of the life of a far-dated option).
Near-dated implieds are therefore more volatile than far-dated implieds. The theoretical term
structure for different strikes is shown in Figure 98 below, which demonstrates that near-
dated implieds are more volatile. We have shown ATM (100%) term structure as upward
sloping as this is how it trades on average (for the same reasons credit spread term structure
is normally upward sloping, ie, risk aversion and supply-demand imbalances for long
maturities).
7.1: Skew and Term Structure Are Linked 189

There are three reasons why skew and term structure are correlated

 Credit events, such as bankruptcy, lift both skew and term structure
 Implied volatility is ‘sticky’ for low strikes and long maturities
 Implied correlation is ‘sticky’ for low strikes and long maturities (only applies to index)

(1) BANKRUPTCY LIFTS BOTH SKEW AND TERM STRUCTURE

There are various models that show the effect of bankruptcy (or credit risk) lifting both skew
and term structure. As implieds with lower strikes have a greater sensitivity to credit risk (as
most of the value of low strike puts is due to the risk of bankruptcy), their implieds rise
more, which causes higher skew. Similarly, options with longer maturity are more sensitive to
credit risk (causing higher term structure, as far-dated implieds rise more). Longer-dated
options have a higher sensitivity to credit risk as the probability of entering default increases
with time (hence a greater proportion of an option’s value will be associated with credit
events as maturity increases). More detail on the link between volatility and credit can be
seen in section A12 capital Structure Arbitrage in the Appendix.

(2) IMPLIED VOL IS ‘STICKY’ FOR LOW STRIKES AND LONG MATURITIES

Term structure should rise if near-dated ATM implieds fall, as far-dated ATM implieds are
relatively constant (as they tend to include complete economic cycles). This is shown
in Figure 102 below.
192 CHAPTER 7: SKEW AND TERM STRUCTURE TRADING

In the same way implied volatility is ‘sticky’ for low strikes and long maturities, so is implied
correlation (in a crisis correlation tends to 100% hence low strike correlation is ‘sticky’, and
in the long run macro trends are the primary driver for all stock prices hence long dated
correlation is also ‘sticky’). This can be an additional reason why index skew and index term
structure are correlated.

CORRELATION LIFTS INDEX SKEW ABOVE SINGLE-STOCK SKEW

An approximation for implied correlation is the index volatility squared divided by the
average single-stock volatility squared [ρ = σ Index ² ÷ average(σ Single stock )²]. Implied correlation
is assumed to tend towards 100% for low strikes, as all stocks can be expected to decline in a
crisis. This causes index skew to be greater than single stock skew. Index skew can be
thought of as being caused by both the skew of the single stock implied volatility surface,
and the skew of the implied correlation surface.

Example of how index skew can be positive with flat single-stock skew
We shall assume all single stocks in an index have the same (flat) implied volatility and
single-stock skew is flat. Low strike index volatility will be roughly equal to the constant
single-stock volatility (as implied correlation is close to 100%), but ATM index volatility will
be less than this value due to diversification (as implied correlation ρ for ATM strikes is less
than 100% and σ Index ² = ρ × average(σ Single stock )². Despite single stocks having no skew, the
index has a skew (as low strike index implieds > ATM index implieds) due to the change in
correlation. For this reason, index skew is always greater than the average single-stock skew.

Implied correlation is likely to be sticky for low strikes and long maturities
A correlation surface can be constructed for options of all strikes and expiries, and this
surface is likely to be close to 100% for very low strikes. The surface is likely to be relatively
constant for far maturities; hence, implied correlation term structure and skew will be
correlated (as both rise when near-dated ATM implied correlation falls, similar to volatility
surfaces). This also causes skew and term structure to be correlated for indices.

DIVERSE INDICES HAVE HIGHER SKEW

As index skew is caused by both single-stock skew and implied correlation skew, a more
diverse index should have a higher skew than a less diverse index (assuming there is no
significant difference in the skew of the single-stock members). This is due to the fact that
diverse indices have a lower ATM implied, but low strike implieds are in line with (higher)
average single-stock implieds for both diverse and non-diverse indices.
7.2: √Root of Time Rule Can Compare Different Term Structures and Skews 193

7.2: SQUARE ROOT OF TIME RULE CAN COMPARE

DIFFERENT TERM STRUCTURES AND SKEWS
When implied volatility changes, the change in ATM volatility multiplied by the
square root of time is generally constant. This means that different (T 2 -T 1 ) term
structures can be compared when multiplied by √(T 2 T 1 )/(√T 2 -√T 1 ), as this
normalises against 1Y-3M term structure. Skew weighted by the square root of time
should also be constant. Looking at the different term structures and skews, when
normalised by the appropriate weighting, can allow us to identify calendar and skew
trades in addition to highlighting which strike and expiry is the most attractive to
buy (or sell).

REALISED VOLATILITY MEAN REVERTS AFTER 8 MONTHS

When there is a spike in realised volatility, it takes on average eight months for three-month
realised volatility to settle back down to levels seen before the spike. The time taken for
volatility to normalise is generally longer if the volatility is caused by a negative return, than if
it is caused by a positive return (as a negative return is more likely to be associated with an
event that increases uncertainty than a positive return). This mean reversion is often
modelled via the square root of time rule.

VOL MOVE MULTIPLIED BY √TIME IS USUALLY CONSTANT

The near-dated end of volatility surfaces is highly correlated to realised volatility, as hedge
funds and prop desks typically initiate long/short gamma positions should there be a
significant divergence. As volatility mean reverts, the far-dated end of volatility surfaces is
more stable (as investors know that any spike in volatility will be short-lived and not last for
the full length of a far-dated option). A common way to model the movement of volatility
surfaces, is to define the movement of one-year implied and then adjust the rest of the curve
by that move divided by time (in years) to the power of p. Only two parameters (the one-
year move and p) are needed to adjust the whole surface. Fixing the power (or p) at 0.5 is the
most common and is known as the square root of time rule (which only has one parameter,
the one-year change).

One year implied volatility move

Implied vol move for maturity T years =
Tp
196 CHAPTER 7: SKEW AND TERM STRUCTURE TRADING

√TIME RULE CAN COMPARE DIFFERENT TERM STRUCTURES

ATM term structure can be modelled as flat volatility, with a square root of time adjustment
on top. With this model, flat volatility is equal to the volatility for an option of infinite
maturity. There are, therefore, two parameters to this model, the volatility at infinity (V ∞ )
and the scale of the square root of time adjustment ,which we define to be z (for one-year
implied).

Volatility = V ∞ – z /√T

where z = scale of the square root of time adjustment (which we define as normalised term
structure)

We have a negative sign in front of z, so that a positive z implies an upward sloping term
structure and a negative z is a downward sloping term structure.

Different term structures are normalised by multiplying by √(T2T1)/(√T2-√T1)

Using the above definition, we can calculate the normalised term structure z from two
volatility points V 1 and V 2 (whose maturity is T 1 and T 2 ).

V 1 = V ∞ – z /√T 1

V 2 = V ∞ – z /√T 2

V 1 + z /√T 1 = V 2 + z /√T 2 = V ∞

z (1/√T 1 – 1/√T 2 ) = V 2 – V 1

T2T1
z = (V 2 – V 1 ) ×
T2 − T1

V 2 – V 1 is the normal definition for term structure. Hence, term structure can be normalised
by multiplying by √(T 2 T 1 )/(√T 2 -√T 1 ). We note that the normalisation factor for 1Y-3M
term structure is 1. Therefore, normalising allows all term structure to be compared to 1Y-
3M term structure.
7.3: Term Structure Trading 199

IDENTIFYING WHEN TO GO LONG, OR SHORT, CALENDARS

When examining term structure trades, the power of the movement in volatility surfaces can
be compared to the expected 0.5 power of the square root of time rule. If the movement has
a power significantly different from 0.5, then a long (or short) calendar position could be
initiated to profit from the anticipated correction. This method assumes calendars were
previously fairly priced (otherwise the move could simply be a mean reversion to the norm).

If vol rises with power less than 0.5, investors should short calendars
If surfaces rise with a power less than 0.5 (ie, a more parallel move) then near-dated implieds
have not risen as much as expected and a short calendar (long near-dated, short far-dated)
position should be initiated. This position will profit from the anticipated correction. Should
surfaces fall with a power less than 0.5, a long calendar (short near-dated, long far-dated)
would profit from the anticipated further decline of near-dated implieds.

If vol rises with power more than 0.5, investors should go long calendars
Conversely, if surfaces rise with a power greater than 0.5, near-dated implieds have risen too
far and a long calendar position should be initiated. On the other hand, if surfaces fall with a
power greater than 0.5, a short calendar position should be initiated (as near-dated implieds
have fallen too far).

POWER VEGA IS VEGA DIVIDED BY THE √TIME

As volatility surfaces tend to move in a square root of time manner, many systems report
power vega (vega divided by square root of time). Power vega takes into account the fact
that the implied volatility of near-dated options is more volatile than far-dated options.

VARIANCE TERM STRUCTURE CAN INDENTIFY TRADES

To determine if a term structure trade is needed, we could look at variance term structure
rather than implied volatility term structure. Using variance term structure eliminates the
need to choose a strike (an ATM term structure will not be ATM as soon as the spot moves,
so it is effectively strike dependent, but simply delegates the choice of strike to the equity
market). Variance term structure is similar to ATM term structure, despite variance being
long skew and skew being greater for near-dated implieds. This is because the time value of
an OTM option increases with maturity. Hence, the increased weight associated with OTM
options cancels the effect of smaller skew for longer maturities.
200 CHAPTER 7: SKEW AND TERM STRUCTURE TRADING

Forward starting var swaps (or options) can be used to trade term structure
Trading term structure via a long and short variance swap is identical to a position in a
forward starting variance swap (assuming the weights of the long and short variance swap
are correct; if not, there will be a residual variance swap position left over). The correct
weighting for long and short variance swaps to be identical to a forward starting variance
swap is detailed in the section 4.1 Forward Starting Products. If an investor wants to trade term
structure, but does not want to have exposure to current volatility (ie, wants to have zero
theta and gamma), then forward starting products (variance swaps or options) can be used.
Note that while forward starting products have no exposure to current realised volatility,
they do have exposure to future expectations of volatility (ie, implied volatility hence has
positive vega).
7.4: How to Measure Skew and Smile 201

7.4: HOW TO MEASURE SKEW AND SMILE

The implied volatilities for options of the same maturity, but of different strike, are
different from each other for two reasons. Firstly, there is skew, which causes low
strike implieds to be greater than high strike implieds due to the increased leverage
and risk of bankruptcy. Secondly, there is smile (or convexity/kurtosis), when OTM
options have a higher implied than ATM options. Together, skew and smile create
the ‘smirk’ of volatility surfaces. We look at how skew and smile change by maturity
in order to explain the shape of volatility surfaces both intuitively and
mathematically. We also examine which measures of skew are best, and why.

MOMENTS DESCRIBE THE PROBABILITY DISTRIBUTION

In order to explain skew and smile, we shall break down the probability distribution of log
returns into moments. Moments can describe the probability distribution 29. From the
formula below we can see that the zero-th moment is 1 (as the sum of a probability
distribution is 1, as the probability of all outcomes is 100%). The first moment is the
expected value (ie, mean or forward) of the variable. The second, third and fourth moments
are variance, skew and kurtosis, respectively (see table on the left below). For moments of
two or greater it is usual to look at central moments, or moments about the mean (we
cannot for the first moment as the first central moment would be 0). We shall normalise the
central moment by dividing it by σn in order to get a dimensionless measure. The higher the
moment, the greater the number of data points that are needed in order to get a reasonable
estimate.

∞
Raw moment = Ε( X k ) = ∫x
k
f ( x)
−∞

∞
Normalised central moment = Ε(( X − µ ) k ) / σ k = ∫ (x − µ) f ( x) / σ k
k

−∞

where
f (x) is the probability distribution function

29 The combination of all moments can perfectly explain any distribution as long as the distribution

has a positive radius of convergence or is bounded (eg, a sine wave is not bounded; hence, it cannot
be explained by moments alone).
7.4: How to Measure Skew and Smile 207

THERE ARE 3 WAYS TO MEASURE SKEW

There are 3 main ways skew can be measured. While the first is the most mathematical, in
practice the other 2 are more popular with market participants.

 Third moment
 Strike skew (eg, 90%-110%)
 Delta skew (eg, [25 delta put – 25 delta call] / 50 delta)

(1) THIRD MOMENT IS DEFINITION OF CBOE SKEW INDEX

CBOE have created a skew index on the S&P500. This index is based on the normalised
third central moment; hence, it is strike independent. The formula for the index is given
below. For normal negative skew, if the size of skew increases, so does the index (as negative
skew is multiplied by -10).

SKEW = 100 – 10 × 3rd moment

(2) STRIKE SKEW SHOULD NOT BE DIVIDED BY VOLATILITY

The most common method of measuring skew is to look at the difference in implied
volatility between two strikes, for example 90%-110% skew or 90%-ATM skew. It is a
common mistake to believe that strike skew should be divided by ATM volatility in order to
take into account the fact that a 5pt difference is more significant for a stock with 20%
volatility than 40% volatility. This ignores the fact that the strikes chosen (say 90%-110% for
20% volatility stocks) should also be wider for high volatility stocks (say 80%-120%, or two
times wider, for 40% volatility stocks as the volatility is 2×20%). The difference in implied
volatility should be taken between two strikes whose width between the strikes is
proportional to the volatility (similar to taking the implied volatility of a fixed delta, eg, 25%
delta). An approximation to this is to take the fixed strike skew, and multiply by volatility, as
shown below. As the two effects cancel each other out, we can simply take a fixed strike
skew without dividing by volatility.

Difference in vol between 2 strikes = 90-110%

Difference in vol between 2 strikes whose width increases with vol = 90-110% × ATM

Skew = Difference in vol between 2 strikes whose width increases with vol
ATM
 Skew = 90-110% × ATM
ATM
 Skew = 90-110%
210 CHAPTER 7: SKEW AND TERM STRUCTURE TRADING

7.5: SKEW TRADING

The profitability of skew trades is determined by the dynamics of a volatility surface.
We examine sticky delta (or ‘moneyness’), sticky strike, sticky local volatility and
jumpy volatility regimes. Long skew suffers a loss in both a sticky delta and sticky
strike regimes due to the carry cost of skew. Long skew is only profitable with jumpy
volatility. We also show how the best strikes for skew trading can be chosen.

4 IDEALISED REGIMES DESCRIBE MOVEMENT OF VOL SURFACE

There are four idealised regimes for a volatility surface. While sticky delta, sticky strike and
(sticky) local volatility are well known and widely accepted names, we have added ‘jumpy
volatility’ to define volatility with a high negative correlation with spot. These regimes are
summarised below, and more details are given on pages 216-222 of this section.

(1) Sticky delta (or sticky moneyness). Sticky delta assumes a constant volatility for
options of the same strike as a percentage of spot. For example, ATM or 100% strike
volatility has constant volatility. As this model implies there is a positive correlation
between volatility and spot, the opposite of what is usually seen in the market, it is not a
particularly realistic model (except over a very long time horizon).
(2) Sticky strike. A sticky strike volatility surface has a constant volatility for options with
the same fixed currency strike. Sticky strike is usually thought of as a stable (or
unmoving) volatility surface as real-life options (with a fixed currency strike) do not
change their implied volatility.
(3) Sticky local volatility. Local volatility is the instantaneous volatility of stock at a certain
stock price. When local volatility is static, implied volatility rises when markets fall (ie,
there is a negative correlation between stock prices and volatility). Of all the four
volatility regimes, it is arguably the most realistic and fairly prices skew.
(4) Jumpy volatility. We define a jumpy volatility regime as one in which there is an
excessive jump in implied volatility for a given movement in spot. There is a very high
negative correlation between spot and volatility. This regime usually occurs over a very
short time horizon in panicked markets (or a crash).
7.5: Skew Trading 215

Typically, traders use two main ways to examine implied volatility surfaces. Absolute
dimensions tend to be used when examining individual options, a snapshot of volatilities, or
plotting implied volatilities over a relatively short period of time. Relative dimensions tend
to be used when examining implied volatilities over relatively long periods of time 37.

 Absolute dimensions. In absolute dimensions, implied volatility surfaces are examined

in terms of fixed maturity (eg, Dec14 expiry) and fixed strike (eg, €4,000). This surface is
a useful way of examining how the implied volatility of actual traded options changes.
 Relative dimensions. An implied volatility surface is examined in terms of relative
dimensions when it is given in terms of relative maturity (eg, three months or one year)
and relative strike (eg, ATM, 90% or 110%). Volatility surfaces tend to move in relative
dimensions over a very long period of time, whereas absolute dimensions are more
suitable for shorter periods of time.

Care must be taken when examining implieds in relative dimensions

As the options (and variance swaps) investors buy or sell are in fixed dimensions with fixed
expiries and strikes, the change in implied volatility in absolute dimensions is the key driver
of volatility profits (or losses). However, investors often use ATM volatility to determine
when to enter (or exit) volatility positions, which can be misleading. For example, if there is
a skew (downside implieds higher than ATM) and equity markets decline, ATM implieds will
rise even though volatility surfaces remain stable. A plot of ATM implieds will imply buying
volatility was profitable over the decline in equity markets; however, in practice this is not
the case.

Absolute implied volatility is the key driver for equity derivative profits
As options that are traded have a fixed strike and expiry, it is absolute implied volatility that
is the driver for equity derivative profits and skew trades. However, we accept that relative
implied volatility is useful when looking at long-term trends. For the volatility regimes (1)
sticky delta and (2) sticky strike, we shall plot implieds using both absolute and relative
dimensions in order to explain the difference. For the remaining two volatility regimes
(sticky local volatility and jumpy volatility), we shall only plot implied volatility using absolute
dimensions (as that is the driver of profits for traded options and variance swaps).

37This is usually for liquidity reasons, as options tend to be less liquid for maturities greater than two
years (making implied volatility plots of more than two years problematic in absolute dimensions).
218 CHAPTER 7: SKEW AND TERM STRUCTURE TRADING

LONG SKEW IS UNPROFITABLE WITH STICKY STRIKE

While there is no profit or loss from re-marking a surface in a sticky strike model, a long
skew position still has to pay skew theta. Overall, a long skew position is still unprofitable
with sticky strike, but it is less unprofitable than with sticky delta.

(3) STICKY LOCAL VOLATILITY PRICES SKEW FAIRLY

As a sticky local volatility causes a negative correlation between spot and Black-Scholes
volatility (shown below), this re-mark is profitable for long skew positions. As the value of
this re-mark is exactly equal to the cost of skew theta, skew trades break even in a sticky local
volatility regime. If volatility surfaces move as predicted by sticky local volatility, then skew is
priced fairly (as skew trades do not make a loss or profit).

BLACK-SCHOLES VOL IS AVERAGE OF LOCAL VOL

Local volatility is the name given for the instantaneous volatility of an underlying (ie, the
exact volatility it has at a certain point). The Black-Scholes volatility of an option with strike
K is equal to the average local (or instantaneous) volatility of all possible paths of the
underlying from spot to strike K. This can be approximated by the average of the local
volatility at spot and the local volatility at strike K. This approximation gives two results 39:

 The ATM Black-Scholes volatility is equal to the ATM local volatility.

 Black-Scholes skew is half the local volatility skew (due to averaging).

Example of local volatility skew = 2x Black-Scholes skew

The second point can be seen if we assume the local volatility for the 90% strike is 22% and
the ATM local volatility is 20%. The 90%-100% local volatility skew is therefore 2%. As the
Black-Scholes 90% strike option will have an implied volatility of 21% (the average of 22%
and 20%), it has a 90%-100% skew of 1% (as the ATM Black-Scholes volatility is equal to
the 20% ATM local volatility).

39 As this is an approximation, there is a slight difference which we shall ignore.

7.5: Skew Trading 221

VOL RE-MARK WITH STICKY LOCAL VOL = SKEW THETA

While a sticky local volatility regime causes long skew positions to profit from (Black-
Scholes) implied volatility changes, the position still suffers from skew theta. The
combination of these two cancel exactly, causing a long (or short) skew trade to break even.
As skew trades break even under a static local volatility model, and as there is a negative spot
vol correlation, it is arguably the most realistic volatility model.

Figure 129. Breakdown of P&L for Skew Trades

P&L breakdown for long skew

(e.g. long put, short call)
Volatility
Remark Skew theta Total
regime

Sticky delta + =

Sticky strike + =

Sticky local
volatility + =

Jumpy
volatility + =

(4) JUMPY VOLATILITY IS ONLY REGIME WHERE LONG

SKEW IS PROFITABLE
During very panicked markets, or immediately after a crash, there is typically a very high correlation
between spot and volatility. During this volatility regime (which we define as jumpy volatility) volatility
surfaces move in excess of that implied by sticky local volatility. As the implied volatility surface re-mark
for a long skew position is in excess of skew theta, long skew positions are profitable. A jumpy volatility
regime tends to last for a relatively short period of time.
APPENDIX

This includes technical detail and areas related to volatility trading that do not fit into
earlier sections.
A.1: Local Volatility 231

BLACK-SCHOLES VOL IS AVERAGE OF LOCAL VOLATILITIES

It is possible to calculate the local (or instantaneous) volatility surface from the Black-
Scholes implied volatility surface. This is possible as the Black-Scholes implied volatility of
an option is the average of all the paths between spot (ie, zero maturity ATM strike) and the
maturity and strike of the option. A reasonable approximation is the average of all local
volatilities on a direct straight-line path between spot and strike. For a normal relatively flat
skew, this is simply the average of two values, the ATM local volatility and the strike local
volatility.

Black-Scholes skew is half local volatility skew as it is the average

If the local volatility surface has a 22% implied at the 90% strike, and 20% implied at the
ATM strike, then the Black-Scholes implied volatility for the 90% strike is 21% (average of
22% and 20%). As ATM implieds are identical for both local and Black-Scholes implied
volatility, this means that 90%-100% skew is 2% for local volatility but 1% for Black-
Scholes. Local volatility skew is therefore twice the Black-Scholes skew.

ATM volatility is the same for both Black-Scholes and local volatility
For ATM implieds, the local volatility at the strike is equal to ATM, hence the average of the
two identical numbers is simply equal to the ATM implied. For this reason, Black-Scholes
ATM implied is equal to local volatility ATM implied.

LOCAL VOL IS THE ONLY COMPLETE CONSISTENT VOL MODEL

A local volatility model is complete (it allows hedging based only on the underlying asset)
and consistent (does not contain a contradiction). It is often used to calculate exotic option
implied volatilities to ensure the prices for these exotics are consistent with the values of
observed vanilla options and hence prevent arbitrage. A local volatility model is the only
complete consistent volatility model; a constant Black-Scholes volatility model (constant
implied volatility for all strikes and expiries) can be considered to be a special case of a static
local volatility model (where the local volatilities are fixed and constant for all strikes and
expiries).
232 APPENDIX

A.2: MEASURING HISTORICAL VOLATILITY

The implied volatility for a certain strike and expiry has a fixed value. There is,
however, no single calculation for historical volatility. The number of historical days
for the historical volatility calculation changes the calculation, in addition to the
estimate of the drift (or average amount stocks are assumed to rise). There should,
however, be no difference between the average daily or weekly historical volatility.
We also examine different methods of historical volatility calculation, including
close-to-close volatility and exponentially weighted volatility, in addition to advanced
volatility measures such as Parkinson, Garman-Klass (including Yang-Zhang
extension), Rogers and Satchell and Yang-Zhang. We also show that it is best to
assume a zero drift assumption for close-to-close volatility, and that under this
condition variance is additive.

DEFINITION OF VOLATILITY
Assuming that the probability distribution of the log returns of a particular security is
normally distributed (or follows a normal ‘bell-shape distribution’), volatility σ of that
security can be defined as the standard deviation of the normal distribution of the log
returns. As the mean absolute deviation is √(2/π) (≈0.8) × volatility, the volatility can be
thought of as c1.25× the expected percentage change (positive or negative) of the security.

σ = standard deviation of log returns × 1 / ∆t

CLOSE-TO-CLOSE HISTORICAL VOLATILITY IS THE MOST

COMMON
Volatility is defined as the annualised standard deviation of log returns. For historical
volatility the usual measure is close-to-close volatility, which is shown below.

 c + di 
Log return = x i = Ln  i
 c  where d i = ordinary dividend and c i is close price
 i −1 

1 N
Volatility 43 (not annualised) = σ x =
N
∑ (x
i =1
i − x )2

where x = drift = Average (x i )

Historical volatility calculation is an estimate from a sample

43We take the definition of volatility of John Hull in Options, Futures and Other Derivatives in which n
day volatility uses n returns and n+1 prices. We note Bloomberg uses n prices and n-1 returns.
A.2: Measuring Historical Volatility 233

Historical volatility is calculated as the standard deviation of the log returns of a particular
securities’ time series. If the log returns calculation is based on daily data, we have to
multiply this number by the square root of 252 (the number of trading days in a calendar
year) in order to annualise the volatility calculation (as ∆t = 1/252 hence 1 / ∆t = 252 ). As
a general rule, to annualise the volatility calculation, regardless of the periodicity of the data,
the standard deviation has to be multiplied by the square root of the number of
days/weeks/months within a year (ie, 252, 52 , 12 ).

σ Annualised = σ x × values in year

VARIANCE IS ADDITIVE IF ZERO MEAN IS ASSUMED

Frequency of returns in a year = F (eg, 252 for daily returns)

1 N
σ Annualised = √F× σ x = √F
N
∑ (x − x)
i =1
i
2

As x ≈ 0 if we assume zero average returns

1 N

∑x
2
σ Annualised = √F i
N i =1

F N

∑x
2
σ Annualised 2 = i
N i =1

Now if we assume that the total sample N can be divided up into period 1 and period 2
where period 1 is the first M returns then:

F N
× ∑ xi
2
σ Total 2 =
N Total i =1

F M
× ∑ xi (where N Period 1 = M)
2
σ Period 1 2 =
N Period 1 i =1

F N

∑x
2
σ Period 2 2 = × i (where N Period 2 = N – M)
N Period 2 i = M +1
234 APPENDIX

then

F N
F M 2 N
2
× ∑ xi =  ∑ xi + ∑ xi 
2
σ Total 2 =
N Total i =1 N Total  i =1 i = M +1 

F M
F N
× ∑ xi + ∑x
2 2
σ Total 2 = × i
N Total i =1 N Total i = M +1

N Period 1  F M
2 N  F N
2
σ Total 2 =  × ∑ xi  + Period 2  × ∑ xi 
N Total  N Period 1 i =1  N Total  N Period 2 i = M +1 

N Period 1 N
σ Total 2 = σ Period 12 + Period 2 σ Period 2 2
N Total N Total

Hence variance is additive (when weighted by the time in each period / total time)

BEST TO ASSUME ZERO DRIFT FOR VOL CALCULATION

The calculation for standard deviation calculates the deviation from the average log return
(or drift). This average log return has to be estimated from the sample, which can cause
problems if the return over the period sampled is very high or negative. As over the long
term very high or negative returns are not realistic, the calculation of volatility can be
corrupted by using the sample log return as the expected future return. For example, if an
underlying rises 10% a day for ten days, the volatility of the stock is zero (as there is zero
deviation from the 10% average return). This is why volatility calculations are normally more
reliable if a zero return is assumed. In theory, the expected average value of an underlying at
a future date should be the value of the forward at that date. As for all normal interest rates
(and dividends, borrow cost) the forward return should be close to 100% (for any reasonable
sampling frequency, ie, daily/weekly/monthly). Hence, for simplicity reasons it is easier to
assume a zero log return as Ln(100%) = 0.

WHICH HISTORICAL VOLATILITY SHOULD I USE?

When examining how attractive the implied volatility of an option is, investors will often
compare it to historical volatility. However, historical volatility needs two parameters.

 Length of time (eg, number of days/weeks/months)

 Frequency of measurement (eg, daily/weekly)
A.2: Measuring Historical Volatility 235

LENGTH OF TIME FOR HISTORICAL VOLATILITY

Choosing the historical volatility number of days is not a trivial choice. Some investors
believe the best number of days of historical volatility to look at is the same as the implied
volatility of interest. For example, one-month implied should be compared to 21 trading day
historical volatility (and three-month implied should be compared to 63-day historical
volatility, etc). While an identical duration historical volatility is useful to arrive at a realistic
minimum and maximum value over a long period of time, it is not always the best period of
time to determine the fair level of long-dated implieds. This is because volatility mean reverts
over a period of c8 months. Using historical volatility for periods longer than c8 months is
not likely to be the best estimate of future volatility (as it could include volatility caused by
earlier events, whose effect on the market has passed). Arguably a multiple of three months
should be used to ensure that there is always the same number of quarterly reporting dates in
the historical volatility measure. Additionally, if there has been a recent jump in the share
price that is not expected to reoccur, the period of time chosen should exclude that jump.

The best historical volatility period does not have to be the most recent
If there has been a rare event which caused a volatility spike, the best estimate of future
volatility is not necessary the current historical volatility. A better estimate could be the past
historical volatility when an event that caused a similar volatility spike occurred. For
example, the volatility post credit crunch could be compared to the volatility spike after the
Great Depression or during the bursting of the tech bubble.

FREQUENCY OF HISTORICAL VOLATILITY

While historical volatility can be measured monthly, quarterly or yearly, it is usually measured
daily or weekly. Normally, daily volatility is preferable to weekly volatility as five times as
many data points are available. However, if volatility over a long period of time is being
examined between two different markets, weekly volatility could be the best measure to
reduce the influence of different public holidays (and trading hours 44). If stock price returns
are independent, then the daily and weekly historical volatility should on average be the
same. If stock price returns are not independent, there could be a difference.
Autocorrelation is the correlation between two different returns so independent returns have
an autocorrelation of 0%.

Trending markets imply weekly volatility is greater than daily volatility

With 100% autocorrelation, returns are perfectly correlated (ie, trending markets). Should
autocorrelation be -100% correlated, then a positive return is followed by a negative return
(mean reverting or range trading markets). If we assume markets are 100% daily correlated
with a 1% daily return, this means the weekly return is 5%. The daily volatility is therefore

44 Advanced volatility measures could be used to remove part of the effect of different trading hours.
238 APPENDIX

historical volatility and then gradually decline to lower levels (arguably in a similar way to
how implied volatility spikes, then mean reverts).

ADVANCED VOLATILITY MEASURES

Close-to-close volatility is usually used as it has the benefit of using the closing auction prices
only. Should other prices be used, then they could be vulnerable to manipulation or a ‘fat
fingered’ trade. However, a large number of samples need to be used to get a good estimate
of historical volatility, and using a large number of closing values can obscure short-term
changes in volatility. There are, however, different methods of calculating volatility using
some or all of the open (O), high (H), low (L) and close (C). The methods are listed in order
of their maximum efficiency (close-to-close var divided by alternative measure var).

 Close to close (C). The most common type of calculation that benefits from only using
reliable prices from closing auctions. By definition its efficiency is one at all times.

 Parkinson (HL). As this estimate only uses the high and low price for an underlying, it
is less sensitive to differences in trading hours. For example, as the time of the EU and
US closes are approximately half a trading day apart, they can give very different returns.
Using the high and low means the trading over the whole day is examined, and the days
overlap. As it does not handle jumps, on average it underestimates the volatility, as it
does not take into account highs and lows when trading does not occur (weekends,
between close and open). Although it does not handle drift, this is usually small. While
other measures are more efficient based on simulated data, some studies have shown it
to be the best measure for actual empirical data.

 Garman-Klass (OHLC). This estimate is the most powerful for stocks with Brownian
motion, zero drift and no opening jumps (ie, opening price is equal to closing price of
previous period). Like Parkinson, it also underestimates the volatility (as it assumes no
jumps).

 Rogers-Satchell (OHLC). The efficiency of the Rogers-Satchell estimate is similar to

that for Garman-Klass; however, it benefits from being able to handle non-zero drift.
Opening jumps are not handled well though, which means it underestimates the
volatility.

 Garman-Klass Yang-Zhang extension (OHLC). Yang-Zhang extended the Garman-

Klass method that allows for opening jumps hence it is a fair estimate, but does assume
zero drift. It has an efficiency of eight times the close-to-close estimate.

 Yang-Zhang (OHLC). The most powerful volatility estimator which has minimum
estimation error. It is a weighted average of Rogers-Satchell, the close-open volatility and
A.2: Measuring Historical Volatility 241

Bias depends on the type of distribution of the underlying

While efficiency (how volatile the measure is) is important, so too is bias (whether the
measure is, on average, too high or low). Bias depends on the sample size, and the type of
distribution the underlying security has. Generally, the close-to-close volatility estimator is
too big 46 (as it does not model overnight jumps), while alternative estimators are too small
(as they assume continuous trading, and discrete trading will have a smaller difference
between the maximum and minimum). The key variables that determine the bias are:

 Sample size. As the standard close-to-close volatility measure suffers with small sample
sizes, this is where alternative measures perform best (the highest efficiency is reached
for only two days of data).

 Volatility of volatility. While the close-to-close volatility estimate is relatively

insensitive to a changing volatility (vol of vol), the alternative estimates are far more
sensitive. This bias increases the more vol of vol increases (ie, more vol of vol means a
greater underestimate of volatility).

 Overnight jumps between close and open. Approximately one-sixth of equity

volatility occurs outside the trading day (and approximately twice that amount for
ADRs). Overnight jumps cause the standard close-to-close estimate to overestimate the
volatility, as jumps are not modelled. Alternative estimates that do not model jumps
(Parkinson, Garman Klass and Rogers-Satchell) underestimate the volatility. Yang-
Zhang estimates (both Yang-Zhang extension of Garman Klass and the Yang-Zhang
measure itself) will converge with standard close-to-close volatility if the jumps are large
compared to the overnight volatility.

 Drift of underlying. If the drift of the underlying is ignored as it is for Parkinson and
Garman Klass (and the Yang Zhang extension of Garman Glass), then the measure will
overestimate the volatility. This effect is small for any reasonable drifts (ie, if we are
looking at daily, weekly or monthly data).

 Correlation daily volatility and overnight volatility. While Yang-Zhang measures

deal with overnight volatility, there is the assumption that overnight volatility and daily
volatility are uncorrelated. Yang-Zhang measures will underestimate volatility when
there is a correlation between daily return and overnight return (and vice versa), but this
effect is small.

46 Compared to integrated volatility.

242 APPENDIX

CLOSE-TO-CLOSE
The simplest volatility measure is the standard close-to-close volatility. We note that the
volatility should be the standard deviation multiplied by √N/(N-1) to take into account the
fact we are sampling the population (or take standard deviation of the sample) 47. We ignored
this in the earlier definition as for reasonably large n it √N/(N-1) is roughly equal to one.

F N
Standard dev of x = s x =
N
∑ (x
i =1
i − x )2

As σ = σ 2 = E (s2 ) < E ( s 2 ) = E (s ) by Jensens´s inequality

N
Volatility = σ x = s x ×
N −1

2
F N
 c 
F
∑
N
Volatility close to close = σ cc = ∑ (x 2
− x) =  Ln( i )  with zero
N −1 i =1
i
N −1 i =1  ci −1 
drift

PARKINSON
The first advanced volatility estimator was created by Parkinson in 1980, and instead of
using closing prices it uses the high and low price. One drawback of this estimator is that it
assumes continuous trading, hence it underestimates the volatility as potential movements
when the market is shut are ignored.

2
F 1 N
 h 
Volatility Parkinson = σ P = ∑  Ln( i ) 
N 4 Ln(2) i =1  li 

47As the formula for standard deviation has N-1 degrees of freedom (as we subtract the sample
average from each value of x)
A.2: Measuring Historical Volatility 243

GARMAN-KLASS
Later in 1980 the Garman-Klass volatility estimator was created. It is an extension of
Parkinson which includes opening and closing prices (if opening prices are not available the
close from the previous day can be used instead). As overnight jumps are ignored the
measure underestimates the volatility.

2 2
F N
1 h   c 
Volatility Garman-Klass = σ GK =
N
∑  Ln( i )  − (2 Ln(2) − 1) Ln( i ) 
i =1 2  li   oi 

ROGERS-SATCHELL
All of the previous advanced volatility measures assume the average return (or drift) is zero.
Securities that have a drift, or non-zero mean, require a more sophisticated measure of
volatility. The Rogers-Satchell volatility created in the early 1990s is able to properly measure
the volatility for securities with non-zero mean. It does not, however, handle jumps; hence, it
underestimates the volatility.

F N
hi hi li li
Volatility Rogers-Satchell = σ RS =
N
∑ Ln( c )Ln( o ) + Ln( c ) Ln( o )
i =1 i i i i

GARMAN-KLASS YANG-ZHANG EXTENSION

Yang-Zhang modified the Garman-Klass volatility measure in order to let it handle jumps.
The measure does assume a zero drift; hence, it will overestimate the volatility if a security
has a non-zero mean return. As the effect of drift is small, the fact continuous prices are not
available usually means it underestimates the volatility (but by a smaller amount than the
previous alternative measures).

Volatility GKYZ =

2 2 2
F N
 o  1 h   c 
σ GKYZ =
N
∑  Ln( i )  +  Ln( i )  − (2 Ln(2) − 1) Ln( i ) 
i =1  ci −1  2  li  oi 

244 APPENDIX

YANG-ZHANG
In 2000 Yang-Zhang created a volatility measure that handles both opening jumps and drift.
It is the sum of the overnight volatility (close-to-open volatility) and a weighted average of
the Rogers-Satchell volatility and the open-to-close volatility. The assumption of continuous
prices does mean the measure tends to slightly underestimate the volatility.

Volatility Yang-Zhang =

σ YZ = σ overnight volatility 2 + k σ open to close volatility 2 + (1 − k )σ RS 2

0.34
where k =
N +1
1.34 +
N −1
2
F N  oi oi 
σ overnight volatility 2
= ∑  Ln( ) − Ln( )
N − 1 i =1  ci −1 ci −1 

2
F N  ci ci 
∑
2
σ open to close volatility =  Ln( ) − Ln( )
N − 1 i =1  oi oi 
A.3: Proof Implied Jump Formula 245

A.3: PROOF IMPLIED JUMP FORMULA

In the section 6.4 Trading earnings announcements/jumps we showed that the implied jump
from an earnings announcement (or any one off event) can be backed out from the
implied volatility of two options. We prove the formula for the expected daily return
from the implied volatility jump.

ADDITIVE VAR ALLOWS JUMP VOL TO BE CALCULATED

The formula for calculating the jump volatility (which relies on variance being additive) is
shown below.

σ Jump = (σ Expiry after jump 2T − σ Diffusive 2 (T − 1)

where
σ Expiry after jump = implied volatility of option whose expiry is after the jump

T = time to the expiry after jump (= T 1 )

σ Diffusive = diffusive volatility (σ Before jump if there is an expiry before the jump, if not it is σ 12 )

σ Jump = implied volatility due to the jump

PROOF OF IMPLIED JUMP (EXPECTED DAILY RETURN)

The derivation of how to calculate the implied daily return on the day of the jump (which is
a combination of the normal daily move and the effect of the jump) from the implied
volatility due to jump (σ Jump ) is below.

S  S 
( )
∆S = S1 − S0 = S0  1 − 1 = S0 e r − 1 where r = Ln 1 
 S0   S0 
 ∆S 
 Ε
 S  = Ε e − 1
r
( )
 0 

 Expected daily return = Ε e − 1 ( r

)
as r is normally distributed
246 APPENDIX

−r 2
∞
1
∫ (e )
2σ 2
r
 Expected daily return = −1 e dr
2πσ 2 −∞

−r 2 −r 2
∞ 0
1 1
∫ (e ) ∫ (e )
2σ 2 2σ 2
r r
 Expected daily return = −1 e dr + −1 e dr
2πσ 2
0 2πσ 2
−∞

−r 2
∞
1
∫ (e )
2σ 2
r
 Expected daily return = − e− r e dr
2πσ 2 0

if we define x such that r = xσ

−x2
∞
1
∫ (e )
2
xσ
 Expected daily return = − e − xσ e dx
2π 0

x2
∞ xσ − x2
1 2 − xσ −

2π ∫0
 Expected daily return = e −e 2
dx

x2 x2
∞ xσ − ∞ − xσ − 2
1 2 1
2π ∫0 2π ∫0
 Expected daily return = e dx − e dx

− ( x −σ ) 2 σ 2 − ( x +σ ) 2 σ 2
∞ + ∞ +
1 2 2 1 2 2

2π ∫0 2π ∫0
 Expected daily return = e dx − e dx

− ( x −σ ) 2 − ( x +σ ) 2
σ2 ∞ σ2 ∞
1 2 1 2

2π ∫0 2π ∫0
 Expected daily return = e 2
e dx − e 2
e dx

− ( x −σ ) 2 − ( x +σ ) 2
σ2 0 σ2 ∞
1 2 1 2
 Expected daily return = e 2
2π ∫e
−∞
dx − e 2
2π ∫0
e dx
A.3: Proof Implied Jump Formula 247

−x2 −x2
σ2 σ σ2 ∞
1 2 1 2
 Expected daily return = e 2
2π ∫e
−∞
dx − e 2
2π σ∫
e dx

2
σ σ2
 Expected daily return = e 2
N (σ ) − e 2
[1 − N (σ )]
σ2
 Expected daily return = e 2
[2 × N (σ ) − 1]
248 APPENDIX

A.4: PROOF VARIANCE SWAPS CAN BE HEDGED BY

LOG CONTRACT (=1/K2)
A log contract is a portfolio of options of all strikes (K) weighted by 1/K2. When this
portfolio of options is delta hedged on the close, the payoff is identical to the payoff
of a variance swap. We prove this relationship and hence show that the volatility of a
variance swap can be hedged with a static position in a log contract.

PORTFOLIO OF OPTIONS WITH CONST VEGA WEIGHTED 1/K2

In order to prove that a portfolio of options with flat vega has to be weighted 1/K2, we will
define the variable x to be K/S (strike K divided by spot S). With this definition and
assuming zero interest rates, the standard Black-Scholes formula for vega of an option
simplifies to:

Vega of option = τ × S × f(x, v)

where

x = K / S (strike a ratio of spot)

τ = time to maturity

v = σ2 τ (total variance)

− d12
1
f(x, v) = ×e 2

2π

1 v
Ln( ) +
d1 = x 2
v

If we have a portfolio of options where the weight of each option is w(K), then the vega of
the portfolio of options V(S) is:

∞
V (S ) = τ ∫ w( K ) × S × f ( x, v)dK
K =0

As K = xS this means dK / dx = S, hence dK = S × dx and we can change variable K for x.

A.4: Proof Variance Swaps Can Be Hedged By Log Contract (=1/K2) 249

∞
V (S ) = τ ∫ w( xS ) × S
2
× f ( x, v)dx
x =0

In order for the portfolio of options to have a constant vega – no matter what the level of
spot – dV(S)/dS has to be equal to zero.

∞
dV
dS
=τ ∫
d 2
dS
[ ]
S w( xS ) × f ( x, v)dx = 0
x =0

And by the chain rule:

∞
 d 
 τ ∫ 2Sw( xS ) + S
2
w( xS ) × f ( x, v)dx = 0
x =0
dS 
∞
 d 
 τ ∫ S 2w( xS ) + S dS w( xS ) × f ( x, v)dx = 0
x =0

As d/dS = (d/dK) × (dK/dS), and dK/dS = x

∞
 d 
 τ ∫ S 2w( xS ) + xS dK w( xS ) × f ( x, v)dx = 0
x =0

As K = xS

∞
 d 
 τ ∫ S 2w( xS ) + K dK w( K ) × f ( x, v)dx = 0
x =0

d
 2w + K w( K ) = 0 for all values of S
dK

constant
 w( K ) =
K2
250 APPENDIX

A.5: PROOF VARIANCE SWAP NOTIONAL = VEGA/2σ

For small differences between the future volatility and current (implied) swap
volatility, the payout of a volatility swap can be approximated by a variance swap. We
show how the difference in their notionals should be weighted by 2σ.

Proof that Variance Swap Notional = vega/2σ

We intend to calculate the relative size (Z) of the variance swap notional compared to
volatility swap notional (volatility swap notional = vega by definition) so they have a similar
payout (for small differences between realised and implied volatility).

Notional variance swap ≈ Z × Notional volatility swap

 (σ F – σ S ) ≈ Z (σ F 2 – σ S 2)
where:

σ F = future volatility (that occurs over the life of contract)

σ S = swap rate volatility (fixed at the start of contract)

As there is a small difference between future (realised) volatility and swap rate (implied)
volatility, then we can define σ F = σ S + x where x is small.

 ((σ S + x) – σ S ) ≈ Z ( (σ S + x)2 - σ S 2) for simplification we shall replace σ S with σ

 x ≈ Z ( (σ2 + 2σx + x2) - σ2)
 x ≈ Z (2σx + x2)
 1 ≈ Z (2σ + x) and as x is small
 1/2σ ≈ Z
Hence Notional variance swap = vega / 2σ (as vega = Notional volatility swap )
A.6: Modelling Volatility Surfaces 251

A.6: MODELLING VOLATILITY SURFACES

There are a variety of constraints on the edges of a volatility surface, and this section
details some of the most important constraints from both a practical and theoretical
point of view. We examine the considerations for very short-dated options (a few
days or weeks), options at the wings of a volatility surface and very long-dated
options.

IMPLIED VOL IS LESS USEFUL FOR NEAR-DATED OPTIONS

Options that only have a few days or a few weeks to expiry have a very small premium. For
these low-value options, a relatively small change in price will equate to a relatively large
change in implied volatility. This means the implied volatility bid-offer arbitrage channel is
wider, and hence less useful. The bid-offer spread is more stable in cash terms for options of
different maturity, so shorter-dated options should be priced more by premium rather than
implied volatility.

Need to price short-dated options with a premium after a large collapse in the
market
If there has been a recent dip in the market, there is a higher than average probability that
the markets could bounce back to their earlier levels. The offer of short-dated ATM options
should not be priced at a lower level than the size of the decline. For example, if markets
have dropped 5%, then a one-week ATM call option should not be offered for less than
c5% due to the risk of a bounce-back.

SKEW SHOULD DECAY BY SQUARE ROOT OF TIME

The payout of a put spread (and call spread) is always positive; hence, it should always have a
positive cost. If it was possible to enter into a long put (or call) spread position for no cost
(or potentially earning a small premium), any rational investor would go long as large a
position as possible and earn risk-free profits (as the position cannot suffer a loss). A put
spread will have a negative cost if the premium earned by selling the lower strike put is more
than the premium of the higher strike put bought. This condition puts a cap on how
negative skew can be: for high (negative) skew, the implied of the low strike put could be so
large the premium is too high (ie, more than the premium of higher strike puts). The same
logic applies for call spreads, except this puts a cap on positive skew (ie, floor on negative
skew). As skew is normally negative, the condition on put spreads (see

Figure 144 below on the left) is usually the most important. As time increases, it can be
shown that the cap and floor for skew (defined as the gradient of first derivative of volatility
with respect to strike, which is proportional to 90%-100% skew) decays by roughly the
square root of time. This gives a mathematical basis for the ‘square root of time rule’ used by
traders.
252 APPENDIX

Figure 144. Put Spread Ratio put spread

2% 200%

Put spread must have Positive cost for ratio put

positive cost as payout is spread enforces skew
always positive decaying by time
Premium

Premium
1% 100%
(this enforces skew decaying (for maturity > c5 years)
by square root of time)

0% 0%
0% 50% 100% 150% 200% 0% 50% 100% 150% 200%
99-101 put spread 99-101 101x99 ratio put spread

Far-dated skew should decay by time for long maturities (c5 years)
It is possible to arrive at a stronger limit to the decay of skew by considering leveraged ratio
put spreads (see chart above on the right). For any two strikes A and B (assume A<B), then
the payout of going long A× puts with strike B, and going short B× puts with strike A
creates a ratio put spread whose value cannot be less than zero. This is because the
maximum payouts of both the long and short legs (puts have maximum payout with spot at
zero) is A×B. This can be seen in the

Figure 144 above on the right (showing a 99-101 101x99 ratio put spread). Looking at such
leveraged ratio put spreads enforces skew decaying by time, not by the square root of time.
However, for reasonable values of skew this condition only applies for long maturities (c5
years).

PROOF SKEW IS CAPPED AND FLOORED BY √TIME

Enforcing positive values for put and call spreads is the same as the below two conditions:

 Change in price of a call when strike increases has to be negative (intuitively

makes sense, as you have to pay more to exercise the higher strike call).

 Change in price of a put when strike increases has to be positive (intuitively makes
sense, as you receive more value if the put is exercised against you).

These conditions are the same as saying the gradient of x (=Strike/Forward) is bound by:

d1 2 d1 2
Lower bound = − 2π .e 2
[1 − N (d 2 )]≤ x ≤ 2π .e 2
.N (d 2 ) = upper bound

It can be shown that these bounds decay by (roughly) the square root of time. This is plotted
below.
254 APPENDIX

In practice, skew is likely to be bounded well before mathematical limits

While a 90%-100% one-year skew of 13.9% is very high for skew, we note buying cheap put
spreads will appear to be attractive long before the price is negative. Hence, in practice,
traders are likely to sell skew long before it hits the mathematical bounds for arbitrage (as a
put spread’s price tends to zero as skew approaches the mathematical bound). However, as
the mathematical bound decays by the square root of time, so too should the ‘market
bound’.

OTM IMPLIEDS AT WINGS HAVE TO BE FLAT IN LOG SPACE

While it is popular to plot implieds vs delta, it can be shown for many models 48 that implied
volatility must be linear in log strike (ie, Ln[K/F]) as log strike goes to infinity. Hence a
parameterisation of a volatility surface should, in theory, be parameterised in terms of log
strike, not delta. In practice, however, as the time value of options for a very high strike is
very small, modelling implieds against delta can be used as the bid-offer should eliminate any
potential arbitrage.

48 Eg, stochastic volatility plus jump models.

A.7: Black-Scholes Formula 255

A.7: BLACK-SCHOLES FORMULA

The most popular method of valuing options is the Black-Scholes-Merton model. We
show the key formulas involved in this calculation. The assumptions behind the
model are also discussed.

BLACK-SCHOLES MAKES A NUMBER OF ASSUMPTIONS

It is often joked that Black-Scholes is the wrong model with the wrong assumptions that gets
the right price. The simplicity of the model has ensured that it is still used despite the
competition from other, more complicated models. The assumptions are below:
 Constant (known) volatility
 Constant interest rates
 No dividends (a constant dividend yield can, however, be incorporated into the interest
rate)
 Zero borrow cost, zero trading cost and zero taxes
 Constant trading
 Stock price return is log normally distributed
 Can trade infinitely divisible amounts of securities
 No arbitrage

BLACK-SCHOLES PRICE OF EUROPEAN OPTIONS

Call option price = S × N (d1 ) − Ke − rT N (d 2 )

Put option price = − S × N (− d1 ) + Ke − rT N (d 2 )

where

S σ2
Ln  + (r + )T
 K 2
d1 =
σ T
256 APPENDIX

S σ 2
Ln  + (r − )T
K 2
d 2 = d1 − σ T =  
σ T
S = Spot

K = Strike

r = risk free rate (– dividend yield)

σ = volatility

T = time (years)
262 APPENDIX

A.9: ADVANCED (PRACTICAL OR SHADOW) GREEKS

How a volatility surface changes over time can impact the profitability of a position.
While the most important aspects have already been covered (and are relatively well
understood by the market) there are ‘second order’ Greeks that are less well
understood. Two of the most important are the impact of the passage of time on
skew (volatility slide theta), and the impact of a movement in spot on OTM options
(anchor delta). These Greeks are not mathematical Greeks, but are practical or
‘shadow’ Greeks.

SKEW INCREASE AS TIME PASSES CAUSES ‘VOL SLIDE THETA’

As an option approaches expiry, its maturity decreases. As near-dated skew is larger than far-
dated skew, the skew of a fixed maturity option will increase as time passes. This can be seen
by assuming that skew by maturity (eg, three-month or one-year) is constant (ie, relative
time, the maturity equivalent of sticky moneyness or sticky delta). We also assume that three-
month skew is larger than the value of one year skew. If we buy a low strike one year option
(ie, we are long skew) then, assuming spot and ATM volatility stay constant, when the option
becomes a three-month option its implied will have risen (as three-month skew is larger than
one-year skew and ATM volatility has not changed). We define ‘volatility slide theta’ as the
change in price of an option due to skew increasing with the passage of time49.

VOL SLIDE THETA IS MOST IMPORTANT FOR NEAR EXPIRIES

Given that skew increases as maturity decreases, this change in skew will increase the value
of long skew positions (as in the example) and decrease the value of short skew positions.
The effect of ‘volatility slide theta’ is negligible for medium- to far-dated maturities, but
increases in importance as options approach expiry. If a volatility surface model does not
take into account ‘volatility slide theta’, then its impact will be seen when a trader re-marks
the volatility surface.

VOL SLIDE THETA MEASURES IMPACT OF CONST SMILE RULE

The constant smile rule (CSR) details how forward starting options should be priced. The
impact of this rule on valuations is given by the ‘volatility slide theta’ as they both assume a
fixed maturity smile is constant. The impact of this assumption is more important for
forward starting options than for vanilla options.

49While we concentrate on Black-Scholes implied volatilities, volatility slide theta also affects local
volatility surfaces.
A.9: Advanced (Practical Or Shadow) Greeks 265

OTM options have a second order ‘anchor delta’

To simplify the example we shall assume the call wing parameter increases the implied
volatility for strikes 110% and more, and the put wing parameter decreases the implied
volatility for strikes 90% or lower. If spot rises 10%, the 120% call implied volatility will
suffer when the anchor is re-marked 10% higher, because the call implied volatility is initially
lifted by the call wing parameter (which no longer has an effect). OTM calls therefore have a
negative ‘anchor delta’ as they lose value as anchor rises. Similarly, as anchor rises the effect
of the put wing will increase, lowering the implied volatility of puts of strike 90% or less as
anchor rises. So, under this scenario all options that are OTM have a negative ‘anchor delta’
that needs to be hedged.
A.11: Sortino Ratio 269

A.11: SORTINO RATIO

If an underlying is distributed normally, standard deviation is the perfect measure of
risk. For returns with a skewed distribution, such as with option trading or call
overwriting, there is no one perfect measure of risk; hence, several measures of risk
should be used. The Sortino is one of the most popular measures of skewed risk, as it
only takes into account downside risk.

SORTINO RATIO IS MODIFICATION OF SHARPE RATIO

The Sharpe ratio measures the excess return, or amount of return (R) that is greater than the
target rate of return (T). Often zero or risk-free rate is used as the target return. To take
volatility of returns into account, this excess return is divided by the standard deviation.
However, this takes into account both upside and downside risks. As investors are typically
more focused on downside risks, the Sortino ratio modifies the Sharpe ratio calculation to
only divide by the downside risk (DR). The downside risk is the square root of the target
semivariance, which can be thought of as the amount of standard deviation due to returns
less than the target return. The Sortino ratio therefore only penalises large downside moves,
and is often thought of as a better measure of risk for skewed returns.

R −T
Sortino =
DR
where

05
T 
DR =  ∫ (T − x) f ( x)dx 
∞ 

R = Return

T = Target return
270 APPENDIX

A.12: CAPITAL STRUCTURE ARBITRAGE

When Credit Default Swaps were created in the late 1990’s, they traded independently
of the equity derivative market. The high levels of volatility and credit spreads during
the bursting of the TMT bubble demonstrated the link between credit spreads,
equity, and implied volatility. We examine four models that demonstrate this link
(Merton model, jump diffusion, put vs CDS, and implied no-default volatility).

NORMALLY TRADE CREDIT VS EQUITY, NOT VOLATILITY

Capital structure arbitrage models can link the price of equity, credit and implied volatility.
However, the relatively wide bid-offer spreads of equity derivatives mean trades are normally
carried out between credit and equity (or between different subordinations of credit and
preferred shares vs ordinary shares). The typical trade is for an investor to go long the
security that is highest in the capital structure, for example, a corporate bond (or potentially
a convertible bond), and short a security that is lower in the capital structure, for example,
equity. Reverse trades are possible, for example, owning a subordinated higher yielding bond
and shorting a senior lower yielding bond (and earning the positive carry as long as
bankruptcy does not occur). Only for very wide credit spreads and high implied volatility is
there a sufficiently attractive opportunity to carry about an arbitrage between credit and
implied volatility. We shall concentrate on trading credit vs equity, as this is the most
common type of trade.

Credit spread is only partly due to default risk

The OAS (Option Adjusted Spread) of a bond over the risk-free rate can be divided into
three categories. There is the expected loss from default; however, there is also a portion due
to general market risk premium and additionally a liquidity cost. Tax effects can also have an
effect on the corporate bond market. Unless a capital structure arbitrage model takes into
account the fact that not all of a bond’s credit spread is due to the risk of default, the model
is likely to fail. The fact that credit spreads are higher than they should be if bankruptcy risk
was the sole risk of a bond was often a reason why long credit short equity trades have
historically been more popular than the reverse (in addition to the preference to being long
the security that is highest in the capital structure in order to reduce losses in bankruptcy).
A.12: Capital Structure Arbitrage 271

CDS usually better than bonds for credit leg, as they are unfunded and easier
to short
Using CDS rather than corporate bonds can reduce many of the discrepancies in spread that
a corporate bond suffers and narrow the difference between the estimated credit spread and
the actual credit spread. We note that CDS are an unfunded trade (ie, leveraged), whereas
corporate bonds are a funded trade (have to fund the purchase of the bond) that has many
advantages when there is a funding squeeze (as occurred during the credit crunch). CDS also
allow a short position to be easily taken, as borrow for corporate bonds is not always
available, is usually expensive and can be recalled at any time. While borrow for bonds was
c50bp before the credit crunch it soared to c5% following the crisis.

Credit derivatives do not have established rules for equity events

While credit derivatives have significant language against credit events, they have no
language for equity events, such as special dividends or rights issues. Even for events such as
takeovers and mergers, where there might be relevant documentation, credit derivatives are
likely to behave differently than equity (and equity derivatives).

CREDIT CAN LEAD EQUITY MARKET AND VICE VERSA

We note that there are occasions when corporate bond prices lag a movement in equity
prices, simply as traders have not always updated levels (but this price would be updated
should an investor request a firm price). CDS prices suffer less from this effect, and we note
for many large companies the corporate bond market is driven by the CDS market and not
vice versa (the tail wags the dog). Although intuitively the equity market should be more
likely to lead the CDS market than the reverse (due to high frequency traders and the greater
number of market participants), when the CDS market is compared to the equity market on
average neither consistently leads the other. Even if the CDS and equity on average react
equally as quickly to new news, there are still occasions when credit leads equity and vice
versa. Capital structure arbitrage could therefore be used on those occasions when one
market has a delayed reaction to new news compared to the other.

GREATEST OPPORTUNITY ON BBB OR BB RATED BONDS

In order for capital structure arbitrage to work, there needs to be a strong correlation
between credit and equity. This is normally found in companies that are rated BBB or BB.
The credit spread for companies with ratings of A or above is normally more correlated to
the general credit supply and interest rates than the equity price. For very speculative
companies (rated B or below), the performance of their debt and equity is normally very
name-specific, and often determined by the probability of takeover or default.
272 APPENDIX

Capital structure arbitrage works best when companies don’t default

Capital structure arbitrage is a bet on the convergence of equity and credit markets. It has
the best result when a company in financial distress recovers, and the different securities it
has issued converge. Should the company enter bankruptcy, the returns are less impressive.
The risk to the trade is that the company becomes more distressed, and as the likelihood of
bankruptcy increases the equity and credit markets cease to function properly. This could
result in a further divergence or perhaps closure of one of the markets, potentially forcing a
liquidation of the convergence strategy.

FUNDAMENTALS CAN DWARF STATISTICAL RELATIONSHIPS

Capital structure arbitrage assumes equity and credit markets move in parallel. However,
there are many events that are bullish for one class of investors and bearish for another. This
normally happens when the leverage of a company changes suddenly. Takeovers and rights
issues are the two main events that can quickly change leverage. Special dividends, share
buybacks and a general reduction of leverage normally have a smaller, more gradual effect.

Rights issue. A rights issue will always reduce leverage, and is effectively a transfer of value
from equity holders to debt holders (as the company is less risky, and earnings are now
divided amongst a larger number of shares).

Takeover bid (which increases leverage). When a company is taken over, unless the
acquisition is solely for equity, a portion of the acquisition will have to be financed with cash
or debt (particularly during a leveraged buyout). In this case, the leverage of the acquiring
company will increase, causing an increase in credit spreads and a reduction in the value of
debt. Conversely, the equity price of the acquiring company is more stable. For the acquired
company, the equity price should jump close to the level of the bid and, depending on the
structure of the offer, the debt could fall (we note that if the acquired company is already in
distress the value of debt can rise; for example, when Household was acquired by HSBC).

GM EQUITY SOARED A DAY BEFORE CREDIT SANK

On May 4, 2005, Kirk Kerkorian announced the intention to increase his (previously
unknown) stake in GM, causing the troubled company’s share price to soar 18% intraday
(7.3% close to close). The following day, S&P downgraded GM and Ford to ‘junk’, causing a
collapse in the credit market and a 122bp CDS rise in two days. As many capital structure
arbitrage investors had a long credit short equity position, both legs were loss making and
large losses were suffered.
A.12: Capital Structure Arbitrage 275

the credit spread of a company rises, this increases the likelihood of a jump to
bankruptcy and increases the skew. A jump diffusion model therefore shows a link
between credit spread and implied volatility.

 Put vs CDS. As the share price of a company in default tends to trade close to zero, a
put can be assumed to pay out its strike in the even of default. This payout can be
compared to the values of a company’s CDS, or its debt market (as the probability of a
default can be estimated from both). As a put can also have a positive value even if a
company does not default, the value of a CDS gives a floor to the value of puts. As 1xN
put spreads can be constructed to never have a negative payout, various caps to the
value of puts can be calculated by ensuring they have a cost. The combination of the
CDS price floor, and put price cap, gives a channel for implieds to trade without any
arbitrage between CDS and put options.

 No-default implied volatility. Using the above put vs CDS methodology, the value of
a put price due to the payout in default can be estimated. If this value is taken away from
the put price, the remaining price can be used to calculate a no-default implied volatility
(or implied diffusive volatility). The skew and term structure of implied no-default
implied volatilities are flatter than Black-Scholes implied volatility, which allows an easier
comparison and potential for identifying opportunities.

(1) MERTON MODEL

The Merton model assumes that a company has an enterprise value (V) whose debt (D)
consists of only one zero coupon bond whose value at maturity is K. These assumptions are
made in order to avoid the possibility of a default before maturity (which would be possible
if there was more than one maturity of debt, or a coupon had to be paid. The company has
one class of equity (E) that does not pay a dividend. The value of equity (E) and debt (D) at
maturity is given below.

Enterprise value = V = E + D

Equity = Max(V – K, 0) = call on V with strike K

Debt = Min(V, K) = K – Max(K – V, 0) = Face value of debt K – put on V with strike K

A.12: Capital Structure Arbitrage 277

DEBT HAS A DELTA THAT CAN BE ARBITRAGED VS EQUITY

As the value of the short put has a delta, debt has a delta. It is therefore possible to go long
debt and short equity (at the calculated delta using the Merton model) as part of a capital
structure arbitrage trade.

If enterprise value is unchanged, then if value of equity rises, value of credit

falls
As enterprise value is equal to the sum of equity and debt, if enterprise value is kept constant
then for equity to rise the value of debt must fall. An example would be if a company
attempts to move into higher-risk activity, lifting its volatility. As equity holders are long a
call on the value of the company they benefit from the additional time value. However, as
debt holders are short a put they suffer should a firm move into higher-risk activities.

Merton model assumes too high a recovery rate

Using the vanilla Merton model gives unrealistic results with credit spreads that are too tight.
This is because the recovery rate (of V/K) is too high. However, using more advanced
models (eg, stochastic barrier to take into account the default point is unknown), the model
can be calibrated to market data.

MERTON MODEL EXPLAINS EQUITY SKEW

The volatility of an enterprise should be based on the markets in which it operates, interest
rates and other macro risks. It should, however, be independent of how it is funded. The
proportion of debt to equity therefore should not change the volatility of the enterprise V;
however, it does change the volatility of the equity E. It can be shown that the volatility of
equity is approximately equal to the volatility of the enterprise multiplied by the leverage
(V/E). Should the value of equity fall, the leverage will rise, lifting the implied volatility. This
explains skew: the fact that options of lower strike have an implied volatility greater than
options of high strike.

σ E ≈ σ V × V / E (= σ V × leverage)
278 APPENDIX

Figure 160. Value of Enterprise and Equity with Low Debt

Equity has low volatility (≈ volatility of enterprise)

160
if debt is low % of enterprise
140
V (value of enterprise) = D + E

120

100

80

60

40

20

0
0 365 730 Days

D (debt) E (equity)

Firms with a small amount of debt have equity volatility roughly equal to firm
volatility
If a firm has a very small (or zero) amount of debt, then the value of equity and the
enterprise are very similar. In this case, the volatility of the equity and enterprise should be
very similar (see Figure 160 above).

Firms with high value of debt to equity have very high equity volatility
For enterprises with very high levels of debt, a relatively small percentage change in the value
of the enterprise V represents a relatively large percentage change in the value of equity. In
these cases equity volatility will be substantially higher than the enterprise volatility (see
Figure 161).
A.12: Capital Structure Arbitrage 281

JUMP DIFFUSION CAUSES CREDIT-INDUCED SKEW

To show how credit spread (or bankruptcy) causes credit-induced skew, we shall price
options of different strike with jump diffusion, keeping the diffusive volatility and credit
spread constant. Using the price of the option, we shall then calculate the Black-Scholes
implied volatility. The Black-Scholes implied volatility is higher for lower strikes than higher
strikes, causing skew.

Credit-induced skew is caused by ‘option on bankruptcy’

The time value of an option will be divided between the time value due to diffusive volatility
and the time value due to the jump to zero in bankruptcy. High strike options will be
relatively unaffected by the jump to bankruptcy, and the Black-Scholes implied volatility will
roughly be equal to the diffusive volatility. However, the value of a jump to a zero stock
price will be relatively large for low strike put options (which, due to put call parity, is the
implied for all options). The difference between the Black-Scholes implied and diffusive
volatility could be considered to be the value due to the ‘option on bankruptcy’.

(3) PUT VS CDS

The probability distribution of a stock price can be decomposed into the probability of a
jump close to zero due to credit events or bankruptcy, and the log-normal probability
distribution that occurs when a company is not in default. While the value of a put option
will be based on the whole probability distribution, the value of a CDS will be driven solely
by the probability distribution due to default. The (bi-modal) probability distribution of a
stock price due to default, and when not in default, is shown below.
A.12: Capital Structure Arbitrage 283

CDS PRICES PROVIDE FLOOR FOR PUTS

As a put can have a positive value even if a stock is not in default, a CDS must be cheaper
than the equivalent number of puts (equivalent number of puts chosen to have same payout
in event of default, ie, using the formula above). If a put is cheaper than a CDS, an investor
can initiate a long put-short CDS position and profit from the difference. This was a popular
capital structure arbitrage trade in the 2000-03 bear market, as not all volatility traders were
as focused on the CDS market as they are now, and arbitrage was possible.

CDS in default must have greater return than put in default (without
arbitrage)
As a CDS has a lower price for an identical payout in default, a CDS must have a higher
return in default than a put. Given this relationship, it is possible to find the floor for the
value of a put. This assumes the price of a CDS is ‘up front’ ie, full cost paid at inception of
the contract rather than quarterly.

Puts return in default = Strike / Put Price

CDS return in default = (100% – Recovery Rate) / CDS Price

As CDS return in default must be greater than or equal to put return in default.

 (100% – Recovery Rate) / CDS Price ≥ Strike / Put Price

 Put Price ≥ Strike × CDS Price / (100% – Recovery Rate)
PUT VS CDS IS A POPULAR TRADE
As the prices of the put and CDS are known, the implied recovery rate can be backed out
using the below formula. If an investor’s estimate of recovery value differs significantly from
this level, a put vs CDS trade can be initiated. For a low (or zero) recovery rate, the CDS
price is too high and a short CDS long put position should be initiated. Conversely, if the
recovery rate is too high, a CDS price is too cheap and the reverse (long CDS, short put)
trade should be initiated.

Put Price = Strike × CDS Price / (100% – Implied Recovery Rate)

RATIO PUT SPREADS CAP VALUE OF PUTS

CDSs provide a floor to the price of a put. It is also possible to cap the price of a put by
considering ratio put spreads. For example, if we have the price for the ATM put, this means
we know that the value of a 50% strike put cannot be greater than half the ATM put price. If
not, we could purchase an ATM-50% 1×2 put spread (whose payout is always positive) and
earn a premium for free. This argument can be used for all strikes K and all 1xN put
spreads, and is shown below:
284 APPENDIX

K
N × put of strike ≤ put of strike K
N

ARBITRAGE MOST LIKELY WITH LOW STRIKE AND LONG MATURITY

The combination of CDS prices providing a floor, and put prices of higher strikes providing
a cap, gives a corridor for the values of puts. The width of this corridor is narrowest for low
strike long maturity options, as these options have the greatest percentage of their value
associated with default risk. As for all capital structure arbitrage strategies, companies with
high credit spreads are more likely to have attractive opportunities and arbitrage is potentially
possible for near-dated options.

(4) NO-DEFAULT IMPLIED VOLATILITY

The volatility of a stock price can be decomposed into the volatility due to credit events or
bankruptcy and the volatility that occurs when a company is not in default. This is similar to
the volatility due to jumps and the diffusive volatility of a jump diffusion model. As the
value of a put option due to the probability of default can be calculated from the CDS or
credit market, if this value was taken away from put prices this would be the ‘no-default put
price’ (ie, the value the put would have if a company had no credit risk). The implied
volatility calculated using this ‘no-default put price’ would be the ‘no-default implied
volatility’. No-default implied volatilities are less than the vanilla implied volatility, as vanilla
implied volatilities include credit risk).

No-default implied volatilities have lower skew and term structure

While we derive the no-default implied volatility from put options, due to put call parity the
implied volatility of calls and puts is identical for European options. As the value of a put
associated with a jump to default is highest for low-strike and/or long-dated options, no-
default implied volatilities should have a lower skew and term structure than vanilla Black-
Scholes implied volatilities. A no-default implied volatility surface should therefore be flatter
than the standard implied volatility surface and, hence, could be used to identify potential
trading opportunities.
 INDEX

A B
"abnormal" market moves, 7 back-testing, 76, 80
absolute dimensions, 215 bankruptcy, 189, 201
absolute time, 228 barrier options:
additive variance rule, 103, 104, 105, 106, call, 134, 135;
107 and correlation, 142;
allowable loss, 143 delta of, 134, 136;
alpha, 23, 24 down, 134, 135;
Altiplano, 164 as European options, 134;
American options, 2, 11, 12, 13, 37, 230 hedging of, 134;
anchor delta, 217n.38, 262, 263–65 key variables for, 134;
annualized volatility, 42 naming conventions for, 135;
annuities, 81 payoff, 135;
arbitrage, 169 payout of, 134, 135, 136, 138;
ATM forward, 144, 145 pricing of, 138;
ATM options: put, 134, 135, 136, 137, 138;
and "abnormal" market moves, 7; rebates with, 139;
definition, 3; types of, 134;
delta of, 7; up, 134, 135
leverage of, 7; basis, 58
liquidity of, 110; basis risk, 58
calculating premium, 41, 44; basket options, 14, 68, 159, 164, 165, 167,
pricing with additive variance rule, 105; 175–76
return on, 7; bear market, protection strategies for, 27,
time value of, 3, 6, 110; 28
vega, 203 bearish strategies, 3, 4, 5, 28, 31
ATM straddles, 38 beta, 155, 157
ATM volatility, 222 beta weighting, 157
at-the-money options. See: ATM options bias, 239, 241
autocallables, 135 bid-offer arbitrage, 38
autocorrelation, 235, 236 bid-offer spreads:
automatic exercise of options, 11, 13 and barrier options, 134;
AXA, pinning of, 43 and call overwriting, 18, 22;
and capital structure arbitrage, 270;
and dispersion trading, 165;
Index 287

bid-offer spreads (continued): C

and forward starting options, 100, 101,
calculation agent, 58
102, 108;
calendar collars, 29
and implied jumps, 180;
calendar put spreads, 30
and relative value trading, 156, 158;
calendar trades, 34, 193, 198, 199
single stocks vs. index, 25, 158;
call options:
and volatility trading, 41
American, 2, 12;
Black, Fischer, 2
and anchor delta, 265;
Black-Scholes. See: Black-Scholes-Merton
ATM, 6, 12, 41, 44, 69, 139, 251;
Black-Scholes-Merton formula:
and barrier options, 134, 135, 137;
assumptions behind, 2, 89, 92, 255;
as bearish or bullish strategy, 4, 31;
and calculating ATM premium, 44;
and borrow cost, 38;
and call option pricing, 255;
breakeven, 70;
and composite options, 149;
and call overwriting, 15, 16, 19, 21;
and delta, 214;
and collars, 26, 29, 30;
and forward starting option pricing,
and contingent premium options, 148;
103;
convexity of, 9;
and gamma, 67;
cost of, 31, 32, 135, 144;
implied volatility, 220, 221, 230–31,
definition, 2;
275, 281;
delta of, 6, 7, 8, 9, 214, 223, 224, 258;
implied volatility surface, 222;
and delta hedging, 36, 37, 39, 97;
invention of, 2;
digital, 8;
and pricing of European options, 255–
and dispersion, 170;
56;
downside of using, 9;
and put option pricing, 255;
early exercise of, 11, 12, 13;
and skew, 218, 219, 231;
effect of volatility on, 3;
and square root of time, 198;
European, 2, 12, 13, 36, 37, 274;
and stock prices, 178;
and gamma scalping, 40;
and volatility, 49, 89, 218
high cost of, 31;
"black swan," 60
intrinsic value, 3;
bonus shares, 14
ITM, 12;
borrow cost, 38, 154
and kurtosis, 205;
borrowing shares, 266–68
liquidity of, 7, 111;
bounce back, 17
long, 4, 5, 9, 12, 13, 31, 33, 40, 69, 84,
box spreads, 123
123, 156, 220;
breakevens, dispersion trading, 173, 174
long equity exposure of, 2, 3;
bull market, protection strategies for, 27,
and look-back options, 146, 147;
28
maximum loss with, 31;
bullish strategies, 3, 4, 5, 26, 31
OTM, 15, 23, 24, 26, 31, 54, 63, 97,
butterflies, 34, 202
204, 205, 211, 220, 223, 263, 264, 265;
buy-write, 15–16
and outperformance options, 143, 144;
BXM index, 18, 19 payoff of, 8;
288 Index

call options (continued): as method of reducing volatility, 21;

payout, 12, 70, 135, 140, 276; as method of yield enhancement, 15,
premium of, 69, 70, 195, 220, 223; 16, 21;
pricing of, 3, 44, 251, 252, 255; near-dated options, 20, 75;
and probability of being ITM, 8, 9; optimal strike for, 20–21;
profit conditions for, 2; and overpricing of implied volatility,
as protection strategy, 23, 24; 15, 17;
and put-call parity, 36, 37, 69, 284; popularity of, 19–20, 75;
and put underwriting, 10; profitability of, 19;
and quanto options, 151; and put spread protection, 31;
and relative value trading, 156; similarity to put underwriting, 10, 17;
replacing stock with, 9; on single stocks, 18, 21, 22;
rho for, 260; and skew, 188;
and risk reversals, 31, 84, 85, 220, three-month vs. one-month, 29, 30, 76;
225n.40; vanilla, 22;
and shadow delta, 101; with booster, 18
short, 4, 5, 10, 16, 17, 19, 29, 30, 156, call spreads:
220, 223, 224, 225; 1x2, 17, 18, 31, 32;
and skew, 204, 208, 211, 223, 227; payout of, 195, 251;
and sticky delta volatility, 214; and relative value trading, 156;
and straddles and strangles, 33; and square root of time rule, 195;
strike of, 24, 31, 146, 195; vs. put, 31, 32
and synthetics, 123; capital structure arbitrage, 270–84
theta for, 259; carry:
upside, 31, 32, 54, 85, 111, 204, 263; calculation of, 36, 42, 47–48;
on value of enterprise, 276, 277; and call overwriting, 20;
vanilla, 135, 151, 156; and forward starting products, 100,
and variance swaps, 54, 58, 63, 69, 70; 158;
vega of, 84n.9; and quanto options, 150;
and volatility trading, 36, 70; and relative value volatility trading, 158;
and VXO, 51, 109; and reverse trades, 270
when to use, 9; cash gamma, 67, 226–27
and worst-of/best-of options, 140, 141 CBOE:
call overwriting: and automatic exercise of options, 13;
in bull markets, 19; call overwriting index of, 18;
definition, 10; and correlation plotting, 167, 168;
dependence on market conditions, 19, opening of, 2;
20; skew index, 207;
to enforce disciplined investing, 16; and volatility indices, 110, 113, 118
enhanced, 22; central moments, 201
far-dated options, 20; Chicago Board Options Exchange. See:
and implied volatility imbalances, 198; CBOE
on indices, 18, 21, 22; cliquets, 106
Index 289

close-to-close volatility, 232, 238, 239–40, types of, 166;

241, 242 and volatility, 166, 169, 170, 174, 175;
collars: and worst-of/best-of options, 140, 141
calendar, 29, 30; correlation indices tickers, 167
cap on performance with, 24, 26; correlation surface, 192
as low-cost method of protection, 24; correlation swaps, 161, 165, 166, 168–70
put spread, 23, 24, 25, 26, 27, 28; correlation trading, 161–77
and skew, 24, 26; correlation-weighted dispersion trading,
and volatility 24, 26, 28; 171
when to use, 26. counterparty risk, 128, 266, 267
See also: risk reversals covariance, 176
composite options, 133, 149–50 covariance swaps, 165, 176
constant smile rule, 103, 105, 106, 262 crash, protection strategies for, 28
contingent premium options, 133, 148 crash puts, 148
continuous barriers, 138 credit default swaps, 270, 271
convertible bonds, 198 credit-induced skew, 274–75, 280–81
convexity: credit spreads, 270, 273, 280, 281
and call options, 9; crossed vega, 167
gamma as measure of, 40, 258; cross-gamma, 176n.24
of options, 6; cross-holding trading, 154
and options on variance, 69, 70; currencies, options involving, 149–51
of payout, 40;
and profits from delta hedging, 41, 45; D
and relative value trading, 156;
daily profit and loss, calculation of, 45,
and stock replacing with calls, 9;
47–48, 50, 55
and variance, 49;
debt, 275, 276, 277
and variance swaps, 54, 67, 69, 109;
declines, types of, 28
vega, 83, 86, 87, 88;
delta:
and volatility indices, 109 anchor, 217n.38, 262, 263–65;
correction, protection strategies for, 27,
barrier options, 134, 136;
28
and borrow cost, 38;
correlation:
call options, 8, 9, 258;
and composite options, 149, 150;
debt, 277;
index, 163–64;
definition, 6, 7, 39, 257, 258;
methods of trading, 164–65;
and dividend risk, 7, 37, 38;
and outperformance options, 143, 144,
intraday, 59;
145, 157, 165; ITM options, 6;
plotting of, 167;
and probability of option being ITM at
and quanto options, 150;
expiry, 8–9;
in relative value trading, 154, 155; put options, 8, 9, 258;
and skew, 158; shadow, 100–101;
and theta-weighted dispersion trading, stock, 9;
174;
290 Index

delta (continued): Dodd-Frank Act, the, 82

volatility futures, 119 dollar gamma, 47, 67, 226–27
delta-1 products, 4, 154, 155, 156 double barriers, 138
delta hedging: drift, 232, 234, 238, 241, 243
continuous, 40, 52, 89–93; DTB (now Eurex), 118
daily profit and loss from, 45, 47–48, dual listing trading, 154
50;
definition, 39; E
discrete, 89, 93–98;
efficiency, 239, 241
and dividend risk, 37;
enhanced call overwriting, 22
and equity exposure, 257, 258;
enterprise value, 275, 276, 277, 279
every 5 percent move in spot, 94;
EONIA, 128
every 10 percent move in spot, 94; equity risk premium, 19, 20, 79
example of, 39;
ETN/ETF, 122–23, 125, 126, 127
gamma, 225;
Eurex, 13, 118
on key dates, 41;
Euronext-Liffe, 13
as method of locking in profit, 40;
European options:
profit from, 41, 45–47, 49, 52, 90–96;
automatic exercise of, 11;
skew, 225;
barrier options, 134;
and trending markets, 41;
contingent premium options, 148;
via synthetics, 123; definition, 2, 11;
when to do, 41
look-back options, 146;
delta re-hedging, 40
maintenance of, 11;
delta skew, 207, 208, 209
options on variance, 68;
diffusive volatility, 178, 179, 180, 280
outperformance options, 143;
digital options, 8, 138
payout of, 92;
directional investors, 36, 39
put-call parity and, 37;
discrete barriers, 138
worst-of/best-of, 140
discrete sampling, 110, 112, 113 event-driven trading, 155
dispersion trading, 159, 161, 164–65,
Everest, 164
170–74, 175
excess return indices, 128
dividend net present value (NPV), 11, 12,
execution risk, 154
13
expected daily return, 245–47
dividend risk, 7, 37
expected value, 201
dividends:
expected volatility, 97
discrete, 58;
expiry:
fixed, 101; choice of as enforcing investor
impact on early exercise of calls, 11, 12;
discipline, 4;
ordinary, 14;
correlation indices, 167;
proportional, 58, 101;
far-dated, 4;
special, 14; key dates in choice of, 4, 5;
and variance swaps, 58
dividend yield assumption, 149
Index 291

expiry (continued): and implied jump trading, 180;

linear interpolation between, 110, 113, maturity of, 101;
114; modeling of, 99, 100;
look-back options, 146; and outperformance options, 144;
near-dated, 4; pricing of, 100, 103, 262;
non-triple witching, 123; relative value trading of, 158;
rolling, 110, 113; straddles with, 102;
strategies for choosing, 4–5; strike of, 101, 102;
triple witching, 167; vega hedging of, 102;
volatility futures, 118; zero gamma of, 100, 102;
and volatility surfaces, 184 zero theta of, 99, 100
exponentially weighted volatility, 232, forward starting variance swaps:
237–38 constructing, 107
hedging of, 102, 108;
F maturity of, 101–2;
fair price, 109, 110, 166 payout of, 106;
fair value, 3, 115, 142, 164 pricing of, 100, 102, 108;
fat-fingered trade, 238 spread of, 108;
fixed annuities, 81 and term structure trading, 200
fixed dividends, 101 forward variance, 107
fixed leg, 57 forward volatility:
flat market, protection strategies for, 27 and diffusive volatility, 179;
floating leg, 57 methods of calculating, 103–6, 180;
forwards: methods of trading, 100, 101–2;
and borrow cost, 38; and pricing of forward starting options,
correlation swaps as, 168; 103;
definition, 6, 36; underpricing of, 106
delta of, 6; FTSE, 158
excluding dividends, 37; futures, on volatility indices, 2, 102, 108,
as first moment of a probability 158
distribution, 51, 201, 202;
gamma swaps as, 58; G
potential loss on, 6; gamma:
profitability of, 6; Black-Scholes, 67;
variance swaps as, 58, 168; cash, 67, 226–27;
volatility swaps as, 58 definition, 40, 257, 258–59;
forward skew, 103, 105 and delta hedging, 52, 90, 91, 225;
forward starting options: dollar, 47, 67, 226–27;
average implied volatility of, 100; and forward starting options, 100;
and carry, 158; definition, 99, 100, 101; long position, 41;
delta of, 100, 101, 102; as measure of convexity of payout, 257,
hedging of, 100, 102; 258;
292 Index

gamma (continued): of option pair trades, 159;

proportionality to volatility, 47, 202; shadow, 67, 100–101, 262–65;
share, 67; and skew, 257;
short position, 41; and time decay, 257;
and term structure trading, 200; of variance swaps, 67;
and theta, 90; and volatility exposure, 257;
and volatility pair trading, 159; and volatility of volatility, 257;
volga as measure of, 205 of volatility swaps, 67
gamma scalping, 40, 41
gamma swap dispersion, 171 H
gamma swaps:
hedge funds:
advantages of, 55–56;
increasing popularity of, 2;
constant share gamma of, 67; maturity of, 7
and dispersion trading, 171, 172;
hedging error, 93, 94, 95
fair price of, 52;
high frequency trading (HFT), 236
as forwards, 58;
Himalayas, 164
Greeks of, 67;
historical volatility:
hedging of, 59, 61, 63;
calculation of, 232–44,
invention of, 52;
close-to-close, 232, 238, 239–40, 241,
liquidity of, 56;
242;
payoff, 56, 58; exponentially weighted, 232, 237–38;
payout, 55, 57, 62, 171, 240;
frequency of, 234, 235–36;
single stocks, 56;
Garman-Klass, 232, 238, 239, 241, 243;
spot weighting of, 55–56;
length of time for, 234–235;
as square root of variance, 52;
Parkinson, 232, 238, 239, 241, 242;
vega profile of, 62
Rogers-Satchell, 232, 238, 239, 241,
gamma theta, 226–27, 228
243, 244;
gamma trading, 225
Yang-Zhang, 232, 238–39, 241, 243
gamma-weighted dispersion trading, 171, hurdles, 13, 143
172, 172, 174
gamma weighting, 159, 160
gap risk premium, 15 I
Garman-Klass, 232, 238, 239, 241, 243 implied correlation:
gearing, 77 and dispersion trading, 164–65, 166,
geometric Brownian motion, 178 167, 170, 174;
Greeks: calculating, 96, 97; and outperformance options, 143, 144;
and convexity of payout, 257; ............... overpricing of, 170;
definition, 257–61; sticky, 192;
of dispersion trading, 172; and worst-of/best-of options, 140, 141,
and equity exposure, 257; 142
of gamma swaps, 67; implied correlation surface, 162, 184
and interest rate exposure, 257; implied diffusive volatility, 275
and moments, 202; implied jump, 5, 178–82, 245–47
Index 293

implied no-default volatility, 270, 275, 284 intraday volatility, 236–37

implied variance, 50 intrinsic value, 3
implied volatility: investors:
absolute, 215, 216; directional, 36, 39;
Black Scholes, 220, 221, 230–31, 275, volatility, 36, 39
281; ITM options:
and calculating delta, 92, 96, 97, 98; automatic exercise of, 13;
and calculating implied jump, 178; cost of, 3;
and diffusive volatility, 179; definition 3, 6;
discrete sampling of, 110, 112, 113; delta of, 6;
as estimate of future volatility, 96; early exercise of, 11, 12, 13;
flooring and capping of, 263; intrinsic value, 3, 6;
and forward starting options, 100; liquidity of, 7;
index, 18, 163, 170; and "normal" market moves, 6;
overpricing of, 15, 17, 18, 29, 53, 74, pricing of, 12, 230;
75, 78; return on, 6;
overshoot of, 83, 86; similarity to forward, 6
and put spread pricing, 26;
and realised volatility, 186; J
relative, 215, 216;
Jensen's inequality, 164, 242
removal of cap on, 60, 61; jump diffusion, 270, 274–75, 280–81, 284
single stock, 18, 163, 170;
jump volatility, 180, 181, 245
and square root of time rule, 194;
jumpy volatility regimes, 210, 212, 213,
sticky, 189–91, 192;
214, 215, 221–23, 225
term structure of, 117, 125, 186;
undershoot of, 83;
and variance, 49, 50; K
volatility of, 130, 186; key dates:
and volatility surfaces, 184 and choosing expiry, 4, 5, 178;
implied volatility premium, 15 and delta hedging, 41
implied volatility surfaces, 162, 215, 216, knock-in options, 134, 135, 136, 137
222 knock-out options, 134, 135, 136, 137,
indexation flow, 154 138, 139
indices: kurtosis, 201, 202, 203, 205, 206
call overwriting on, 18, 21, 22;
implied volatility in, 18; L
implied volatility premium for, 15; ladders, 32, 33, 123
liquidity of, 7; leptokurtic distributions, 202, 205
options on, 11 leverage, 7, 188, 201, 272, 279
instantaneous volatility, 230–31 leveraged buyout, 272
in-the-money options. See: ITM options Levy processes, 206n.34
interest rates, 11, 13
intraday delta, 59
294 Index

light exotics: forward starting variance swaps, 101–2;

barrier options as most popular, 142; hedge funds, 7;
definition, 133; look-back options, 147;
as European options, 134, 143, 146; outperformance options, 143;
as OTC, 134, 143, 146 and smirk, 206;
liquidity: structured products, 7, 169;
call options, 7; worst-of/best-of options, 140
correlation swaps, 170; maximum anniversary value, 81
covariance swaps, 176; mean reversion, 119, 131, 154, 158, 159,
and demand for protection, 7; 184, 193, 199
as factor in choosing strike, 5, 7 mergers and acquisitions, effect on value
gamma swaps, 56; of options, 14
indices, 7; Merton, Robert 2
ITM options, 7; Merton model, 270, 274, 275–79
OTM options, 7; mesokurtic distributions, 202
outperformance options, 157; model risk, 66
put options, 7; moments, 201–5
single stock options, 7; moneyness, 103
single stock variance market, 171;
and straddle dispersion trading, 170; N
variance swaps, 52, 53, 55, 170 naked put, selling of, 10, 17
local volatility, 212, 218, 219, 230–31
near-dated options:
log contract, 50, 62, 63, 115, 248
and call overwriting, 15, 20;
long strategies, 2, 4, 5
and capital structure arbitrage
long-short trading, 155
strategies, 284;
Long-Term Capital Management, collapse
and implied volatility, 199, 251;
of, 52
and multiple-expiry protection
long volatility strategies, 4, 5, 78, 80
strategies, 28, 29;
look-back options, 81, 133, 146–47 and volatility selling strategies, 75
NKY, 166
M noise, 77, 94, 95, 110, 113, 154, 155
Margrabe's formula, 144 "normal" market moves, 6
mark-to-market, 136, 157, 170, 267 notional:
market disruption, 57, 58 and relative value trading, 155, 157;
market risk, 93 variance, 54, 56, 62, 63, 64, 65, 66, 68,
market skew, 213 69, 106–7;
maturity: variance swap, 250;
choosing, for options, 157; volatility, 53, 54, 62, 64;
correlation swaps, 169–70; volatility swap, 250
as driven by duration of likely decline,
28;
forward starting options, 101;
Index 295

O single stock options; vanilla options

option adjusted spread (OAS), 270 options on variance:
option combos, 31 breakeven for, 70;
option on bankruptcy, 281 capped, 71;
option structures, 31–32, 33 and convexity, 69;
options: inverted term structure of, 71;
advantages to using, 4, 6; payout of, 69, 70;
Asian, 68; poor liquidity of, 68, 71;
basket, 14, 68, 159, 164, 165, 167, 175– price of, 68, 71;
76; and put-call parity, 69;
cheap, 92, 93, 95, 96; skew of, 70, 71;
choosing expiry of, 4; straddles on, 72;
choosing maturity of, 157; and volatility of volatility, 68
choosing strike of, 157; options on volatility, 70
composite, 133, 149–50; options on volatility futures, 72
contingent premium, 133, 148; OTM options:
convexity of, 6; and "abnormal" market moves, 7;
digital, 8, 138; and call overwriting, 20;
involving currencies, 149–51; definition, 3;
on indices, 11; delta of, 6, 7, 223;
knock-in, 134, 135, 136, 137; and implied volatility, 201, 202;
knock-out, 134, 135, 136, 137, 138, and ladders, 32;
139; leverage of, 7;
look-back, 81, 133, 146–47; liquidity of, 7, 51, 53, 109, 111;
mergers and acquisitions effect on, 14; method for calculating premium, 41,
outperformance, 133, 142, 143–45, 44;
154, 155, 156, 157, 165; pricing with additive variance rule, 105;
potential loss on, 6; and put spread collars, 24;
as preventing inertia, 4, 16; and put underwriting, 10;
as protection, 15, 23–30; time value of, 117, 199;
quanto, 133, 149, 150–51; and variance swaps, 53, 54, 117;
rainbow, 175n.23; vega of, 86, 87;
rolling, 11, 14, 29, 77, 78, 102, 110, 157; and volatility of volatility, 202;
spinoffs' effect on, 14; volga of, 205, 261
time value of, 3, 14; outperformance, 25
worst-of/best-of, 133, 140–42, 164, outperformance options, 133, 142, 143–
165. 45, 154, 155, 156, 157, 165
See also: American options; ATM out-the-money options. See: OTM
options; barrier options; call options; options
European options; forward starting
options; ITM options; near-dated
options; OTM options; put options;
296 Index

P pinning, 42, 43, 59

pair trading, weighting of, 159, 160 platykurtic distributions, 202
Parkinson, 232, 238, 239, 241, 242 power vega, 199
payoff: preference shares, 154
barrier options, 135; premium:
call options, 8; of ATM options, 157;
correlation swaps, 168–69, 176–77; and contingent premium options, 148;
corridor variance swaps, 55; equity risk, 74, 78, 79;
dispersion trading, 170; importance of to directional investors,
gamma swaps, 56, 58; 39;
variance swaps, 54, 58, 115; of knock-in/knock-out options, 134;
volatility swaps, 54, 58; methods for calculating, 41–42, 44,
volatility futures, 115 198;
payout: as method of paying for protction, 27;
barrier options, 134, 135, 136, 138; and near-dated options, 251;
basket options, 176; and put spreads, 195;
call spreads, 195, 251; as restriction on choice of strikes, 28;
and close-to-close prices, 240; return of, 81;
and correlation, 166; and risk reversal, 220;
correlation swaps, 168–69, 175; and strike reset options, 146
delta-hedged options, 47, 91, 94; pricing:
and dispersion trading, 168, 170, 175; barrier options, 138;
European options, 92; basket options, 176;
forward starting variance swaps, 106; calendars, 198, 199;
gamma swaps, 55, 57, 62, 171, 240; forward starting options, 100, 103, 262;
look-back options, 146, 147; outperformance options, 143, 157, 165;
options on variance, 69, 70; quanto options, 151;
outperformance options, 143; straddle dispersion, 170;
put spreads, 195, 251, 252; worst-of/best-of options, 140, 165
quanto options, 150; probability distribution, 201
relative value trading, 154; profit:
structured products, 176; from delta hedging, 41, 45–47, 49, 52,
variance swaps, 52, 53, 54, 55, 57, 58, 90–96;
59, 64, 69, 110, 119, 240; path independence of, 90, 91–92;
vega-weighted pair trades, 160; proportionality to volatility, 41, 45–47,
volatility futures, 115; 49;
volatility swaps, 52, 53, 55, 57, 64, 240, from theta-weighted variance
250; dispersion, 166
proportional dividends, 58, 101
with zero volatility, 3;
protection:
worst-of/best-of options, 140
index vs. single stock, 15, 25;
performance caps with collars, 24, 26
median maturity of, 28;
perpetual structured products, 122
methods of paying for, 27;
Index 297

protection (continued): 251, 283;

option buying for, 15, 23–30; and look-back options, 146, 147;
overpricing of, 29; maximum loss with, 23;
partial, 24, 25; maximum value of, 8;
strategy comparison, 27; methods of reducing cost of, 23;
volatility as determination of strategy multiple expiry strategies, 29;
for, 28 and no-fault implied volatility, 284;
put-call parity, 36, 37, 69 OTM, 12, 23, 24, 26, 31, 54, 63, 83, 88,
put options: 148, 204, 211, 212, 220, 223, 225, 263,
American, 13; 264;
and anchor delta, 264; payout of, 12, 13, 26, 70, 135, 148, 252,
ATM, 26, 28, 69, 81, 85, 136, 211, 283; 282;
and barrier options, 134, 135, 136, 137, premium of, 41, 69, 195, 251;
138; price of, 3, 61, 119, 252, 255, 283;
as bearish or bullish strategy, 3, 5; probability expires ITM, 8;
and borrow cost, 38; profit conditions for, 2;
breakeven, 70; as protection strategy, 23, 24, 25, 27,
and calendar collar, 30; 28, 29, 31, 75, 83, 136, 186, 188;
and contingent premium options, 148; and put-call parity, 36, 37, 69;
cost of, 23, 24, 25, 26, 28, 29, 31, 227; and put underwriting, 17;
crash, 148; and quanto options, 151;
definition of, 2; replacing stock with, 9, 10;
delta of, 8, 9, 39, 207, 208, 214, 224, rho of, 260;
258, 277; and risk reversals, 31, 84, 85, 202, 220,
and delta hedging, 36, 39, 97; 225n.40;
digital, 8; and shadow delta, 101;
disadvantages of, 23; short, 2, 4, 5, 9, 17, 26, 30, 31, 69, 83,
downside, 23, 54, 85, 111, 119, 212, 84, 123, 138, 220, 277;
263; short-dated, 28,29;
early exercise of, 11, 12, 13; short equity exposure of, 2;
effect of volatility on, 3; and skew, 204, 206, 211, 212, 223, 227;
on the enterprise, 274; and smile, 206;
European, 2, 12, 13, 36, 37; and stock replacing, 9, 10, 17;
and gamma scalping, 40; and straddles and strangles, 30, 33, 36,
hedging with, 80; 38;
and implied volatility, 119, 211, 263, strike of, 24, 145;
284; and structured product guarantee, 88;
intrinsic value, 3; and synthetics, 123;
ITM, 81, 97; theta of, 259;
and kurtosis, 205; time value of, 61;
liquidity of, 7, 111; and value of jump to zero stock price,
long, 2, 4, 5, 12, 13, 23, 24, 26, 29, 31, 281;
40, 138, 202, 220, 223, 224, 225n.40, vanilla, 137, 156;
298 Index

put options (continued): relative value volatility trading, 158–60

and variable annuities, 81; re-marking of volatility, 100, 217, 218,
and variance swaps, 54, 58, 69, 70, 119; 220, 221, 222, 225, 262
and volatility trading, 36, 70; repo, 38
vs CDS, 270, 275, 281, 282, 283; return of premium, 81
and VXO, 51, 109; reverse trades, 270
and worst-of/best-of options, 140 rho, 257, 260
put spreads: risk-free rate, 74, 78, 79, 80
1x2, 26, 27, 28, 32, 33; risk-on assets, 141
calendar, 30; risk reversals, 24, 31, 32, 211, 220, 223.
cost-effectiveness of, 24, 26; See also: collars
outperformance with, 25; Rogers-Satchell, 232, 238, 239, 241, 243,
as partial protection, 24, 27, 31; 244
payout of, 195, 251, 252; roll-down cost, 125
price of, 26, 254; rolling expiries, 110, 113
and relative value trading, 156; rolling options, 11, 14, 29, 77, 78, 102,
risk-reward profile of, 25; 110, 157
and square root of time rule, 195; roll-up, 81
and volatility, 28 Russell 2000, 118
put spread collars 23, 24, 25,26,27, 28
put spread vs call, 23 S
put vs CDs, 270, 275, 281–84
S&P100, 109
put vs strangle/straddle, 30
S&P500:
put underwriting, 10, 17, 31
call overwriting index on, 18;
and correlation, 166;
Q implied volatility of, 158;
quantitative easing, 141 OTC market of, 123–24;
quanto options, 133, 149, 150–51 price return index (SPX), 18;
QE2, 141 size of options market, 122, 123;
and skew index, 207;
R term structure, 117, 125;
total return index (SPXT), 18;
rainbow options, 175n.23
and VIX index, 109
range trading, 16, 146, 147
Scholes, Myron, 2
ratchet, 81
shadow delta, 100–101
ratio put spreads, 283
shadow Greeks, 67, 100–101, 262–65
realised dkew, 222–23
share class trading, 154
rebates, 139, 268
recovery rate, 277, 282 share gamma, 67
shark fin, 138
relative dimensions, 215
Sharpe ratio, 269
relative time, 103, 228
short rebate, 268
relative value trading, 143, 144, 154–58,
178
Index 299

short volatility strategies: realised, 222–23;

expiry for, 5; rise in, 82;
when to use, 4 as risk indicator, 191;
single stock options: selling, 26;
as American, 11; and smirk, 206;
call overwriting on, 18, 21, 22; and square root of time rule, 194, 195;
implied volatility in, 18; sticky delta, 103;
implied volatility premium for, 15; strike, 207, 208, 209;
liquidity of, 7; and term structure, 183, 184, 188–92;
volatility of, 20 theoretical, 213;
skew: as third moment of probability
and bankruptcy, 189, 201; distribution, 201, 202, 204, 206;
Black-Scholes, 218, 231; upper and lower bound for, 252–54;
buying, 26; vanna as measure of, 86, 204, 257, 261;
and collars, 24, 26; and variable annuities, 81;
credit-induced, 274–75, 280–81; and variance, 49;
decay of, 106; and variance swaps, 54, 59, 68, 101,
definition of, 184; 109, 117;
delta, 207, 208; and volatility indices, 109
delta hedging, 225; skew index, 207
dependence on correlation, 158; skew profit, 223
with diverse indices, 192; skew theta, 211, 212, 217, 218, 221, 222,
fair pricing of, 218, 222–23; 223, 225, 226, 228
fixed strike, 207, 208; skew trading, 183, 193, 210–28
forward, 103, 105; smile, 183, 201, 202, 206; measuring, 183
and forward starting options, 101, 103, smirk, 201, 206
105; Sortino ratio, 21, 269
index vs. single stock, 192; spinoffs: effect on value of options, 14
index for, 207; spreads:
inverted, 187; bid-offer, 25, 41, 100, 101, 102;
and local volatility, 218, 219, 231; box, 123;
and log contracts, 63; call, 23, 31, 195, 251;
market, 213; call 1x2, 17, 18, 32;
as measure of value of skew, 86; call/put, 122;
measuring, 183, 207–8; credit, 270, 273, 280, 281;
and options of different maturity, 187; option adjusted, 275;
and options on variance, 70, 71; put, 23, 24, 25, 26, 27, 28, 31, 32, 195,
and options on variance swaps, 129; 251, 252, 254;
and options on volatility futures, 129, ratio put, 283
131; SPX, 166
overpricing of, 10, 81, 212; square root of time rule, 119, 183, 193–
and put spread pricing, 26; 97, 199, 251
in put underwriting, 10;
300 Index

standard deviation, 21, 42, 58, 232, 233, maturity of, 7, 169;
234, 242, 269 perpetual, 122;
sticky delta regimes, 210, 212, 213, 214, vicious circle, 83, 87
215, 216–17, 218, 221, 223, 225 stub trades, 154
sticky delta skew, 103 supply and demand imbalances, 15, 74,
sticky local volatility regimes, 210, 212, 117, 125, 184, 186, 198
214, 215, 218–21, 223, 225 Swisscom, pinning of, 43
sticky moneyness regimes, 210, 216 SX5E:
sticky strike, 184, 185, 187, 194 and 1x2 put spreads, 26;
sticky strike regimes, 210, 212, 214, 215, and ATM puts, 28;
217–18, 221, 223 and call overwriting, 20, 21;
stochastic local volatility model, 116 implied correlation of, 166;
stock replacing, 9, 10 implied volatility of, 158, 191;
straddle dispersion, 170–71 and liquidity of variance market, 55;
straddles: and skew, 82, 188, 195;
and butterflies, 34; term structure, 125, 188, 197;
delta hedging of, 59; variance indices of, 118;
and dispersion trading, 165, 170; and variance swaps, 77, 79;
and forward starting options, 102; vega, 126, 127;
on options on variance, 72; vs. 5-year CDS, 274;
put vs., 30; and vStoxx, 74, 78, 125
similarity to strangles, 33; synthetics, 123
vanilla, 102;
and volatility pair trading, 159; T
zero delta, 38, 40
tail risk, 53, 66
strangles, 30, 33, 34, 40, 102, 159, 170
tails, 109, 111
strike:
takeovers, 155, 187, 271, 272
liquidity as factor in choosing, 5, 7;
term structure:
premium as factor in choosing, 28; ATM, 199;
in put underwriting, 10;
and bankruptcy, 189;
sticky, 184, 185, 187, 194;
definition of, 184, 196;
strategies for choosing, 5–7;
and forward starting options, 100, 103;
and volatility surfaces, 184;
implied correlation, 192;
zero delta straddles, 38
implied no-default implied volatility,
strike reset, 146, 147
275, 284;
strike skew, 207, 208, 209
implied variance, 71;
structured products: implied volatility, 75, 87, 125, 130, 186,
and implied correlation, 164–66;
199, 275;
capital guarantee of, 88;
index, 181, 182;
hedging of, 7;
inaccurate pricing of, 81, 105;
and implied volatility imbalances, 198;
inverted, 29, 68, 71, 185, 186, 187;
lifting of index implied volatility, 15,
modeling of, 196;
18, 83, 86;
Index 301

term structure (continued): U

normalising, 178, 196, 197;
upside participation strategies, 31–32
and options of different strikes, 185;
options on variance swaps, 129;
risk, 34; V
single stock, 182; V1X, 109, 132
and skew, 183, 184, 188–92, 226; V2X, 118, 132
slope of, 75, 77, 106, 113, 117, 184, vanilla options:
185, 187, 190, 227; compared to barrier options, 134, 135,
and square root of time rule, 193, 194, 136, 137;
196, 226; compared to composite options, 149;
SX5E, 125, 188, 197; compared to contingent premium
trading, 198–200; option, 148;
vanilla equity options, 130; compared to forward starting options,
variance, 51, 199; 100, 103;
volatility, 68, 110; compared to look-back options, 146;
volatility futures, 77, 122, 124, 126, 129 compared to quanto options, 15, 151;
theoretical skew, 213 and constant smile rule, 262;
theta: and dispersion trading, 170;
call options, 259; implied volatility term structure for,
definition, 257, 259; 130;
and dispersion trading, 172; increase in value as strike approaches,
and forward starting options, 99, 100; 136;
and gamma, 90, 91; and outperformance options, 157;
of near-dated options, 20, 29; pricing of, 230, 231;
put options, 259; and relative value trades, 155, 156, 157,
and term structure trading, 200; 158;
and time decay, 257, 259; and removal of implied volatility cap,
volatility slide, 106, 228; 61;
and volatility pair trading, 159 similarity to options on variance, 72;
theta-weighted dispersion trading, 171, size of market for, 122–23
172, 173, 174 vanna:
theta weighting, 159, 160, 166 definition, 85, 86, 225, 257, 261;
tick value, 119 and determining profitability of skew
time decay, 14, 257, 259 trades, 223;
time value, 3, 90, 117, 135, 178, 199 as measure of third moment of
total return indices, 128 probability distribution, 202, 203, 204;
trending markets: and skew, 202, 204, 257, 261
definition, 41; variable annuities:
and delta hedging, 41, 94; definition, 81, 204;
and sticky delta regimes, 213, 216; downside to, 81;
and volatility measurement, 235 hedging of, 81, 123;
and skew, 81;
302 Index

variable annuities (continued): index, 54, 55, 58, 71;

types of downside protection with, 81 influence on market movement, 59, 60;
variance: liquidity of, 52, 53, 55, 56, 68, 170;
as additive, 58, 67, 106, 180, 232, 233– maximum loss of, 66;
34, 245; notional, 250;
and avoiding deviation cancelling, 50; one-year, 54;
and delta hedging payout, 47; options on, 129;
fair price of, 49; path independence of, 50;
and implied volatility, 49; payoff, 54, 58, 115;
index, 53; payout of, 52, 53, 54, 55, 57, 58, 59, 64,
as measure of deviation, 41, 45, 49–51, 69, 110, 119, 240;
109; price of, 65, 66, 68, 109, 110;
options on, 68; put on, 70;
overpricing of, 74, 75; and removal of cap on implied
as second moment of probability volatility, 61;
distribution, 201, 202, 203; S&P500, 55;
single stock, 53, 66; single stock, 53, 54, 55, 71;
term structure of, 117; and skew, 54, 59, 63, 67, 68, 117;
and volatility, 42, 47, 49, 50, 51, 53, 54, SX5E, 55;
104 and tail risk, 53;
variance notional, 54, 56, 62, 63, 64, 65, and term structure trading, 200;
66, 69 up, 55;
variance swap dispersion, 171 vanilla, 54, 55, 71, 102, 107, 108;
variance swaps: vega portfolio of, 62, 63, 64, 65, 67;
as alternative to strangles or straddles, and volatility of volatility, 52, 66;
30; and volatility pair trading, 159
calculation agents for, 58; variance term structure, 51
call on, 70; VDAX (V1X), 51, 109
capped, 54, 71; vega:
constant cash gamma of, 67; and ATM options, 203;
and convexity, 54, 67, 69, 70, 115; convexity, 83, 86, 87, 88;
corridor, 55; crossed, 167;
delta of, 67, 101; decay of with time, 67;
and dispersion trading, 165, 171; definition, 86, 123, 203, 257, 259–60;
down, 55; and dispersion trading, 172;
effect of dividends on pricing of, 58; and forward starting options, 100, 102,
fair price of, 109; 106;
forward starting, 100, 101–2; of gamma swaps, 62;
as forwards, 58, 168; and implied volatility, 200, 203;
Greeks of, 67; as measure of second moment of
hedging by log contract, 248–49; probability distribution, 202, 203, 204;
hedging of, 54–55, 59, 61, 62, 63, 115; of options of different strikes, 61;
history of, 2, 52, 53; peak, 87;
Index 303

vega (continued): as determinator of protection strategy,

of and straddle dispersion, 170; 28;
and variance swap notional, 250; diffusive, 178, 179, 180;
variance swaps, 62, 63, 64, 65; effect on call options, 3;
and volatility exposure, 257, 259–60; effect on put options, 3;
and volatility pair trading, 159; expected, 97;
of volatility swaps, 62, 64, 65 exponentially weighted, 232, 237–38;
vega flat risk reversal, 84, 85 historical, 41, 232–44;
vega hedging, 102 implied vs. realized, 130;
vega-weighted dispersion trading, 171, index, 161–64;
172, 173, 174 instantaneous, 230–31;
vega weighting, 159, 160 intraday, 236–37;
VFTSE, 51, 109 jumpy, 210;
VHSI, 51, 109 local, 212, 218, 230–31;
VIMEX, 51, 109 long strategies, 4, 5;
VIX: 77, 111; as measure of deviation, 41, 45, 49,
ETN/ETF, 125; 109;
forward starting, 99, 100, 101; overnight, 241;
futures delta, 120, 121; overpricing of, 74, 75, 77, 78, 79, 81;
imbalance in market, 127–28; and pricing of options, 3;
liquidity of, 122; proportionality to profit, 41, 45–47, 49;
listing of futures on, 115, 118; realised, 222;
and relative value volatility trading, 158; re-marking of, 100, 217, 218, 220, 221,
settlement price of, 119; 222, 225, 262;
size of futures market, 122, 123; as risk measure, 191n.27;
term structure, 117; seasonality of, 117;
and variance-based calculations 49, 51, as second moment of a distribution, 51;
109 short strategies, 4, 5;
VNKY, 51, 109 sticky local, 210;
volatility: term structure, 68;
annualized, 42; and variance, 42, 47, 49, 50, 51.
ATM, 222; See also: implied volatility
Black-Scholes, 90, 218; volatility cone, 185, 186
calculating, 202; volatility futures:
of cash, 14; constant maturity, 122;
close-to-close, 232, 238, 239–40, 241, delta of, 119, 120;
242; delta hedging of, 122;
and correlation, 166, 169, 170, 174, eighth month, 127;
175; expiry of, 118;
correlation with equity, 74, 77, 78; fair price of, 115;
and credit, 189; fourth month, 122, 127;
daily, 241; front month, 122, 127;
definition, 232; hedging of, 102;
304 Index

volatility futures (continued): 262n.49

liquidity of market for, 102; volatility surfaces:
and mean reversion, 131; and additive variance rule, 104, 105;
options on, 129–32; chopping tails of, 109, 111, 112;
overpricing of, 102, 116; chopping wings of, 110, 111;
payoff of, 115; and constant smile rule, 105;
payout of, 115; convexity of, 115;
probability distribution of options on, curvature of, 63n.6, 112, 117;
131; discrete sampling of, 112, 113;
and relative value volatility trading, 158; and forward starting products, 100;
skew of options on, 129, 131; idealised regimes for, 210;
term structure of, 117, 122, 124; and implied correlation surfaces, 162;
volatility of, 130; implied volatility, 60, 87, 184, 192;
and volatility of volatility, 102, 115, modeling of, 184, 251–54, 263;
117, 131 movement at front end of, 100;
volatility indices: parallel movement of, 119, 194;
as average of 8 different options, 109; pinning of, 59;
fair price of, 109; and realised volatility, 59, 193;
futures on, 2, 102, 108, 115–21; re-marking of, 211, 212;
and linearly interpolating between removal of cap on, 60;
expiries, 113, 114; and skew, 82, 212;
methods of calculating, 109, 110; slope of, 110, 126;
pricing of, 109, 110–14; and square root of time rule, 114, 198,
providers, 51; 199;
and relative value volatility trading, 158; and sticky delta and relative time
variance-based calculation of, 109; method, 103, 104, 105;
volatility of, 130 and sticky strike, 185, 187;
volatility investors, 36, 39 term structure of, 103, 110;
volatility notional, 53, 54, 62, 64 and theta, 67;
volatility of volatility: three dimensions of, 184
and bias, 241; volatility swaps:
calculating level of, 202; and dispersion trading, 165, 171;
and dispersion trading, 164, 166, 170, fair price of, 109;
175; as forwards, 58;
and gamma swaps, 52, 56; Greeks of, 67;
and options on variance, 68; hedging of, 59, 61, 63, 72, 115;
underpricing of, 116; history of, 2, 52, 53;
and variance swaps, 52, 66; index, 53;
and volatility futures, 102, 115, 117, notional, 250;
131; path independence of, 53;
and volatility swaps 52, 53, 56, 64, 65, payoff, 54, 58;
66, 67, 102, 115 payout, 52, 53, 55, 57, 64, 240, 250;
volatility slide theta, 106, 228, 262, price of, 65, 66;
Index 305

volatility swaps (continued): settlement price of, 119;

single stocks, 52, 53, 55; and SX5E term structure, 125;
trade of, 53; total return indices on, 128;
vega profile of, 62, 64, 65; variance-based calculation of, 51, 109
and volatility of volatility, 53, 56, 64, VXD, 118
65, 66, 67, 72, 115 VXN, 118
volatility trading: VXO, 49, 51, 109
carry of, 158; VXTH, 118
forward starting options: 99, 100; VXX, 127, 132
history of, 2; VXZ, 132
payout of, 89;
put-call parity in, 36, 37 W
volatility weighting, 157 weighting in dispersion trading, 171–74
VOLAX futures, 118
wings, 63, 75, 110, 111, 227, 263
Volga:
worst-of/best-of options, 133, 140–42,
definition, 63n.6, 87, 257, 261;
164, 165
and dispersion trading, 164, 168, 170,
175;
and log contracts, 63n.6; X
as measure of fourth moment of XXV, 127
probability distribution, 202, 205;
and OTM options, 205; Y
and volatility of volatility, 202, 257, 261 Yang-Zhang, 232, 238–39, 241, 243
Volker Rule, the, 82 yield curve, 184
vomma, 87
VSMI (V3X), 51, 109
VST1ME, 128 Z
VST1MT, 128 Zero cost:
vStoxx (V2X): 1x2, 17, 18, 26, 27 31, 32
correlation with SX5E, 74; call spread vs put, 32
ETN/ETF, 125; collar, 24, 25, 26, 27
excess return indices on, 128; put spread collar, 24, 25, 27
liquidity of, 122; risk reversal, 32
listing of forwards on, 99, 100, 101;
and relative value volatility trading, 158;
size of market, 125, 128;