TRADING VOLATILITY

AS AN ASSET CLASS

Emanuel Derman
Columbia University

emanuel@ederman.com
www.ederman.com

Page 1 of 37
June 10, 2003
 Summary
Volatility is a useful trading hedge against all kinds of disasters. How
can you trade it?
Calls and puts don’t quite do it. Though calls and puts are sensitive to
volatility, they are not sensitive only to volatility.
How can you do better?

1. WHY TRADE VOLATILITY?

2. VOLATILITY TRADING WITH OPTIONS: THE VALUE OF CURVATURE

3. OPTIONS TRADING: WHAT CAN GO WRONG

4. THE VOLATILITY SMILE VIOLATES BLACK-SCHOLES

5. WHAT CAUSES THE SMILE?

6. IS THE SMILE FAIR?

7. TRADING & PRICING VOLATILITY USING VOLATILITY SWAPS

Page 2 of 37
June 10, 2003
 I

WHY TRADE
VOLATILITY?
WHY TRADE VOLATILITY?

Volatility is the simplest measure of a stock’s riskiness or uncertainty.

Different types: realized volatility, implied volatility, local volatility...

Page 3 of 37
June 10, 2003
 Realized Volatility
The realized daily volatility σd of an equity index Si over a period of N
days is the square root of the variance of the daily returns ri:
∆S i 2 2 1 2
∑ ∑
1
r i ≈ -------- σd = ---- ( r i ) –  ---- r i
Si N N 
i i

WHY TRADE Stock returns are roughly random and

VOLATILITY? normal; variance grows ~ return time. ∆S ≈ σS ∆t

σ annual ∼ 16 × σ daily

Similarly for currencies, commodities, interest rates, volatility itself,

Page 4 of 37
June 10, 2003
 Implied Volatility
Black-Scholes: the fair price of a stock option depends on its future
realized volatility.
C BS = C ( S, K, t, T, r, σ )

Black-Scholes is both a model and a quoting mechanism for options

prices.
I Implied volatility Σ is the value of the volatility σ you have to insert
into the Black-Scholes formula to make it match the market price of
WHY TRADE an option.
VOLATILITY?
You can think of it as the market’s expectation of future realized
volatility, plus a spread.

Page 5 of 37
June 10, 2003
 Bonds and Options /Yields and Volatilities
Bonds
Interest rates are the parameters
people use to quote bond
prices.
Realized daily interest rates:
I actual short-term interest rates
WHY TRADE Yield to maturity of a bond:
VOLATILITY? the average of the future real-
ized rates that make that bond
price fair. It’s the implied yield
based on price.
Forward rates:
the future realized rates,
moment by moment, that must
come to pass to make current
yields of all liquid bonds fair.
Page 6 of 37
June 10, 2003
Options
Volatilities are the parameters
people use to quote options
prices
Realized daily volatility:
the actual volatility of an index
Implied volatility of an option:
the average of future realized
volatilities that make the
options price fair, based on
Black-Scholes.
Local (forward) volatilities:
the future realized index vola-
tilities, index level by index
level and moment by moment,
that must come to pass to make
current implied volatilities fair.
 Why Trade Volatility?
Attractive characteristics:
• It grows when uncertainty increases.
• It reverts to the mean.
• It goes up and tends to stay up when most assets go down.
Types of volatility trading:
I • Speculative trading of volatility levels in various markets.
Index vs. stock, foreign vs. domestic, short-term vs. long-
WHY TRADE term, currencies, rates,...
VOLATILITY? You need views about future uncertainty to trade it.
• Trading the spread between realized and implied volatility levels.
• Hedging implicit volatility exposure.
Hedge funds and risk arbitrageurs are often implicitly short
volatility. They often take short positions in the spread
between stocks of companies planning to merge, assuming
that the spread will narrow if the merger takes place. If
overall market volatility increases, these mergers are less
likely to occur, and the spread may widen.

Page 7 of 37
June 10, 2003
 II

II VOLATILITY TRADING
VOLATILITY
TRADING WITH
WITH OPTIONS
OPTIONS

Page 8 of 37
June 10, 2003
 A Linear Security: A stock or index
ASSUMED STOCK EVOLUTION: P&L OF A STOCK POSITION:
P&L = ∆S

S
S0 ∆S ∆S
II ∆t
VOLATILITY
TRADING WITH
OPTIONS

• If you own a stock or index, you make money if it goes up, lose if it
goes down.
• You have a linear position in ∆S over the next instant ∆t.

• As an investor, how do you profit no matter whether the index goes

up or down?

Page 9 of 37
June 10, 2003
 A Curved Security
To make money irrespective of direction, you want a security which
gives you a curved P&L: a quadratic position in (∆S)2:
P&L = (1/2)Γ(∆S)2

curvature Γ
II

VOLATILITY ∆S
TRADING WITH
OPTIONS To get curvature: delta-hedge away the linear part of a call option.
C

- =
S curved
long call short stock = hedge

Page 10 of 37
June 10, 2003
 P&L of Delta-Hedged Options
gain ~ 1/2Γσ 2S2∆t

The gain due to a move ∆S with

realized vol σ

index
shift ∆S

loss ~ (1/2)Σ2S2∆t
II
The loss in an instant ∆t
VOLATILITY if expected vol is Σ
TRADING WITH
OPTIONS

gain - loss ~(1/2)Γ[σ 2 -Σ2]S2 ∆t

A hedged position makes

money if realized vol exceeds
implied vol, with magnitude
depending on option’s curvature Γ
index
shift ∆S and stock price S2

Page 11 of 37
June 10, 2003
 P&L Depends on Realized Vol vs. Implied Vol

Net P&L = --- ∫ Γ S  σ – Σ  ∆t

1 2 2 2
2  

Long options: you make money if σ > Σ

Short options: you make money if σ < Σ

II

VOLATILITY
TRADING WITH
OPTIONS

Page 12 of 37
June 10, 2003
 III

III

OPTIONS
TRADING: WHAT
OPTIONS TRADING:
CAN GO WRONG WHAT CAN GO WRONG

Page 13 of 37
June 10, 2003
 1. It’s Not A Clean Bet on Volatility Alone

Net P&L = --- ∫ Γ S  σ – Σ  ∆t

1 2 2 2
2  

2
Γ S is irritating. The Γ (Gamma) (Curvature) of an option as stock
price S varies sharply:
III

OPTIONS
TRADING: WHAT
CAN GO WRONG 100 call

If the stock price moves away, you get very little bang for your option
buck.
Page 14 of 37
June 10, 2003
 2. Delta-Hedging In Practice Is Risky
Delta-hedging assumes smooth stock movements, continuous
hedging, no transactions costs, a known future volatility.
Real markets violate Black-Scholes assumptions because of:
• Jumps.
• Transaction costs.
• Stochastic volatility: the future realized volatility which determines
III your hedge ratio is not known.
OPTIONS • You cannot really hedge continuously; you must hedge discretely.
TRADING: WHAT
CAN GO WRONG
All of these imperfections may overwhelm any theoretical gain.
Let’s look at just one example -- discrete hedging.

Page 15 of 37
June 10, 2003
 Example: Error When Hedging is Discrete
Perfect Black-Scholes world; hedge N times for 1-month ATM put
with 20% realized vol. The Black-Scholes value is 2.512 dollars.
Monte Carlo simulation of the P&L:
28 28

Frequency (out of 100%)

Frequency (out of 100%)

24 24
20 20
III 16 16
12 12
OPTIONS 8 8
TRADING: WHAT 4 4
CAN GO WRONG
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5
Final profit/loss Final profit/loss

N=21 (roughly daily) N=84

Mean=0 Mean=0
Standard deviation=0.41 Standard deviation=0.20
About 16% error, 4 vol pts About 8% error, 2 vol pts

Your P&L isn’t guaranteed unless you hedge continuously. But that
introduces large transactions costs and shifts the expected return.
It’s not easy to make money this way.
Page 16 of 37
June 10, 2003
 IV

IV THE VOLATILITY SMILE

THE VOLATILITY
SMILE VIOLATES
VIOLATES
BLACK-SCHOLES BLACK-SCHOLES

The “smile” is the characteristic variation of implied volatility with

strike and expiration. It is inconsistent with the Black-Scholes model
everyone uses.

Page 17 of 37
June 10, 2003
 Implied Volatility Σ Violates Black-Scholes
• Implied volatility is the single value of the volatility you have to
insert into the Black-Scholes model to match the market price of an
option.
• It is the future volatility the index must have to make the Black-
Scholes price fair.
• If the Black-Scholes model is correct, all options would have the
IV same implied volatility.
• That isn’t the case.
THE VOLATILITY
SMILE VIOLATES
BLACK-SCHOLES Black-Scholes is wrong in principle, not just in practice, and therefore
hedging is even more difficult.

Page 18 of 37
June 10, 2003
 Pre- and Post-Crash Implied Volatilities:
Since the ‘87 crash there has been a persistent skewed structure in
Black-Scholes implied volatilities in most world equity option
markets.
Representative implied volatility skews of S&P 500 options. (a) Pre-crash.
(b) Post-crash. Data taken from M. Rubinstein, “Implied Binomial Trees” J. of
Finance, 69 (1994) pp 771-818.
20

IV
18

V olatility
THE VOLATILITY
SMILE VIOLATES 16

BLACK-SCHOLES
14
0.95 0.975 1 1.025 1.05
Strike/Index

(b)
20

18
V olatility

16

14
0.95 0.975 1 1.025 1.05
Strike/Index

Page 19 of 37
June 10, 2003
 A Persistent Negative Global Skew/Smile
A persistent large skew, almost linear, and inconsistent with Black-
Scholes.

Global Three-Month Volatility Skews

Mar 99

50 Nikkei 225
IV
45 S&P 500
THE VOLATILITY g
Hang Sendg
40
SMILE VIOLATES 35
FTSE 100
BLACK-SCHOLES DAX
30
CAC 40
25 MIB 30
20 SMI
25D Put Atm 25D Call AEX

Σ ( K ) = Σ atm – b ( K – S 0 )

Page 20 of 37
June 10, 2003
 The Implied Volatility Surface of Indexes
The implied volatility surface for S&P 500 index options as a function of
strike level and term to expiration on September 27, 1995.

IV

THE VOLATILITY
SMILE VIOLATES
BLACK-SCHOLES

• Out-of-the-money puts always have higher implied volatilities than

out-of-the-money calls.
• Short-term volatilities are usually more volatile.
• Implied volatility is usually greater than recent historical volatility.

Page 21 of 37
June 10, 2003
 More recent S&P 500 smiles
SPX Imp Vol (12/26/01)

70%
60%
50% slope ~ 1 vol pt. per 1% change in strike 1 Mon

Imp Vol
40% 3 Mon
30% 6 Mon
20% 12 Mon
10%
0%
IV 0.5 0.75 1 1.25 1.5
Strike/Spot

THE VOLATILITY
SMILE VIOLATES Single stock smile
BLACK-SCHOLES
VOD Imp Vol (12/27/01)

70%

60% 1 Mon
Imp Vol

3 Mon
50%
6 Mon
40% 12 Mon

30%
0.5 0.75 1 1.25 1.5
Strike/Spot

Page 22 of 37
June 10, 2003
 Some currency smiles....

USD/EUR MXN/USD JPY/USD

IV

THE VOLATILITY
SMILE VIOLATES
BLACK-SCHOLESATM Strike = 0.90 9.85 123.67

Page 23 of 37
June 10, 2003
 Negative Correlation Between Implied Vols and
Index Levels

Three-Month Implied Volatilities of SPX Options

65 1200
60 1150
55 1100
IV 50 1050
1000
45
950
40
THE VOLATILITY 900
35 INDEX
SMILE VIOLATES 30
850
800
BLACK-SCHOLES 25 750
ATM

20 700
15 650
09-01-97

10-01-97

11-03-97
12-01-97

01-02-98
02-02-98
03-02-98

04-01-98

05-01-98

06-01-98

07-01-98

08-03-98

09-01-98

10-01-98

11-02-98
But be careful - ATM vol isn’t something you own. You own a specific
strike vol.

Page 24 of 37
June 10, 2003
 Patterns of Change in the Volatility Surface
Volatility surfaces fluctuate in modes:
∆Σ = β 1 ( volatility level mode ) + β 2 ( term structure mode ) + β 3 ( skew mode ) :
+ other modes

Modes, or factors, are movements of the entire surface.

IV As with interest rates, there are parallel shifts, “steepening”, and
“twists.”
THE VOLATILITY
SMILE VIOLATES
BLACK-SCHOLES Modes are useful when:
• They can be understood intuitively.
• Few modes explain most of the variation.
• Modes are historically stable.

One finds about 85% of moves are parallel, 10% changes in slope with
respect to time, and 5% related to out-of-the-money short term puts.

Page 25 of 37
June 10, 2003
 V

WHAT CAUSES
THE SMILE?
WHAT CAUSES THE SMILE?

Page 26 of 37
June 10, 2003
 Causes of The Smile
Behavioral causes:
• Supply and demand - purchases of collars
• Expectation of changes in volatility
• Fear of crashes (down for equities, up for gold
After a crash, realized and implied volatilities will be higher
correlation of individual stocks in a jump down increases
V • Level- and time-dependent effects
resistance levels in stocks
WHAT CAUSES support levels in currencies
THE SMILE? change of volatility behavior at low interest rates
Also:
• Fat tails in distributions
• Leverage effects
• Transactions costs?
• Uncertain future volatility

All of these effects make Black-Scholes wrong: the underlier doesn’t

carry out simple Brownian motion.

Page 27 of 37
June 10, 2003
 Models for the Smile

Classic Black-Scholes:
constant volatility

V
Local volatility models:
WHAT CAUSES correlated volatility
THE SMILE?

Stochastic volatility models: random volatility

Jump-diffusion models:
jump
small probability of a large
jump

S
diffuse
Page 28 of 37
June 10, 2003
 Which model you use depends on what market
you deal with.

• Currencies tend to have stochastic volatilities.

• Interest rates have volatilities that depend on interest levels.
• Stock markets tend to jump in the short run, diffuse over longer
times.
V •
WHAT CAUSES There is no single correct replacement for Black-Scholes. It’s hard to
THE SMILE? test options models under the best of circumstances.

Page 29 of 37
June 10, 2003
 VI

VI IS THE SMILE FAIR?

IS THE SMILE
FAIR?

Page 30 of 37
June 10, 2003
 Pre- and Post-Crash SPX Return Distributions
Three-month, S&P 500 index, observed return distributions.
(a) pre-1987 crash (1/70 to 1/87); (b) post-1987 crash (6/87 to 6/99)

mean =1.8% mean = 3.3%

std. div. = 7.3% std. div. = 7.8%

6
5
3
Probability (%)

Probability (%)
4
2

3
VI

2
IS THE SMILE 1

1
FAIR?
0

0
-20 0 20 -20 0 20
Index Return (%) Index Return (%)
24

Fair estimated volatility skew.

22

post-crash
pre-crash
20
Volatility (%)
16 18 14
12
10

90 95 100 105 110

Strike Level

Page 31 of 37
June 10, 2003
 VII

VII TRADING & PRICING VOLATILITY

TRADING &
PRICING
USING
VOLATILITY VOLATILITY SWAPS
USING
VOLATILITY
SWAPS The cleanest and easiest way to trade volatility is through a volatility
swap, which is a forward contract on realized volatility.
Swaps have become very popular because dealers have access to a
theory for pricing them relative to the options market.
The theory provides a basis for defining volatility indexes like the VIX
in terms of the price of a basket of options that can be traded, and for
creating your own volatility swaps.

Page 32 of 37
June 10, 2003
 Analogy: Credit Default Swap Market

Corporate Bonds Hedged Options

Bond prices are influenced by
interest rates and credit spreads
Credit default swaps allow you
to simply bet on credit spreads
VII
Options prices are influenced
by stock prices and volatility
Volatility swaps let you trade
pure volatility.

TRADING & Why isn’t there a market for volatility alone?

PRICING
VOLATILITY
USING
VOLATILITY
SWAPS

Page 33 of 37
June 10, 2003
 Volatility Contracts : Volatility and Variance Swaps
A Volatility Swap is a forward contract on realized volatility:

Payoff: ( σ
R – K vol ) × N
where N is the notional amount.

σ R : realized volatility realized by the index until expiration;

VII K vol : previously agreed upon “delivery” volatility.

TRADING & A variance swap is a forward contract on realized variance. It pays
PRICING
VOLATILITY  σ2 – K  × N
USING
VOLATILITY
 R var
SWAPS
Must specify the precise method for calculating realized volatility, the
source and observation frequency of prices, and the annualization
factor.
Variance swaps are easier to price and hedge using options.
A dealer will sell you a variance swap. You can trade volatility in one
market or sector against another easily, without the hassle and errors
of hedging options. But you want to understand how to price it!
Page 34 of 37
June 10, 2003
 Dealers create a variance swap by hedging a Log
Contract

Net P&L = --- ∫ Γ S  σ – Σ  ∆t

1 2 2 2
Trading a single option:
2  

2
An option whose ΓS = 1 would exactly earn the realized volatility
VII 2
over the life of the contract, with a delivery price equal to Σ .
TRADING & • What kind of option has a
Γ 1/S2
PRICING 2
VOLATILITY Γ∼1⁄S ? call
USING
VOLATILITY
SWAPS
S
• An option whose payoff is the call
payoff
natural logarithm of S: – ln(S) ln(S)
S

Page 35 of 37
June 10, 2003
 Dealers Can Create a Log Contract Out of
A Basket of Options with Many Different Strikes
A portfolio of vanilla options with weights inversely proportional to
their strike squared can replicate the payoff of a Log Contract. It will
be equally sensitive to volatility at all spot level S.
strikes: 80,100,120
equally

1/K2
weighted

A density (a) (b)

weighted
inversely
proportional

VII of puts and calls

to square
of strike

will give the same 20 60 100 140 180 20 60 100 140 180

TRADING & payoff as a strikes 60 to 140

spaced 20 apart

PRICING Log Contract. (c) (d)

VOLATILITY
USING 20 60 100 140 180 20 60 100 140 180

VOLATILITY It will be equally strikes 60 to 140

sensitive to volatility spaced 10 apart

SWAPS at all spot levels (e) (f)

20 60 100 140 180 20 60 100 140 180

strikes 20 to 180
spaced 1 apart

(g) (h)

20 60 100 140 180 20 60 100 140 180

Page 36 of 37
June 10, 2003
 Pricing of Variance Swaps
• Dealers replicate the log contract by buying and hedging a portfolio
of puts and calls (and a forward contract).
• The fair price of the variance swap is given by the market price of
the basket of puts and calls, a price which depends on the smile.
You can create your own variance swaps like this to do vol arbitrage.
Difficulties to be aware of
VII
• You need a continuum of puts and calls of all strikes to replicate it
TRADING & exactly.
PRICING • In practice you cannot buy very out-of-the-money strikes. So, if the
VOLATILITY stock price moves too far, your replication will fail.
USING • If the stock jumps rather than moves smoothly, there are additional
VOLATILITY replication failures
SWAPS
Volatility swaps are more complex.
Volatility is the square root of variance, and is a more complex
derivative of variance. Its value depends not just on volatility, but on
the volatility of volatility, and you have to dynamically hedge it by
trading variance swaps as the underlier. This is possible too, but needs
a model for the volatility of volatility

Page 37 of 37
June 10, 2003