Sunteți pe pagina 1din 49

Equity-Based Insurance Guarantees Conference

November 1-2, 2010

New York, NY

Black Holes or Black Scholes: Modeling


Long-dated Options

Paul Staneski
Black--Scholes or Black Holes?
Black

Issues in modeling long-


long-dated options
1 November
N b 2010
1330 – 1415 hrs

Paul Staneski, Ph.D.


Credit Suisse
Head of Derivatives Solutions & Training

Society of Actuaries Equity-Based Guarantees Conference New York

1
It’s a Model!

 “All models are wrong, some are useful.”


 George Box/Gwilym Jenkins

Slide 2

2
Some Issues

 The role of Gamma

 Rho vs. Vega


g
– Relative impacts over time

 Variance Swaps
– Short-term vs. long-term replication

 Realized vs. Implied Vol


– How do you price?
– How do you hedge?

Slide 3

3
Black-Scholes

 The “Black-Scholes” model rivals CAPM (the Capital Asset Pricing Model) as
the most important result in the history of finance.
– Fischer Black and Myron
y Scholes, “The Pricing
g of Options
p and Corporate
p Liabilities”,
Journal of Political Economy, 81 (May-June 1973), pp. 637-659.
– Scholes and Robert Merton won the 1997 Nobel Prize in Economics for this work
(Black died in 1995 and the prizes are not awarded posthumously).

 Historical note: Ed Thorp actually first derived this result in 1967!


– Read “Fortune’s Formula” by William Poundstone, Hill & Wang, 2005.

Slide 4

4
Black-Scholes in a Nutshell

Option ∆Stock

(“balance” with financing)

Slide 5

5
The Greeks are Key!

Slide 6

6
Delta Hedging & Gamma

Value Hedge (Short Stock) Long Call

Gain on Long Call


Gain on Hedge

Loss on Call
Loss on Hedge

Stock Price
Because of gamma
gamma, the hedged position has a net gain regardless of the move in spot
spot.

Slide 7

7
Capturing Gamma

 Hedging & re-hedging a long call …

P/L

S1 S2
Stock Price

Initial Premium

Gamma “captured”

Slide 8

8
Capturing (More) Gamma

 Hedging & re-hedging a long call …

P/L

S3 S1 S2
Stock Price

Initial Premium
Gamma “captured”
captured
Gamma “captured”

Slide 9

9
Model Assumptions

 Under all the model’s assumptions (there are many!) the application of the
model to a 10-day option is not at all different than its application to a 10-year
option.

 However, in the “real world” this time invariance is not the case!

 The “troublesome”
troublesome assumptions …
 Constant interest rates (and dividends)
 Constant volatility

Slide 10

10
Question

 Suppose you buy a 90-day Call option on a stock at an implied volatility of


35% and dynamically hedge it (perfectly, as per the Black-Scholes
assumptions) to expiration. If realized volatility over the 90 days is 40%, will
you make
k money? ?

– Answer: ____________

Slide 11

11
Answer

 Not necessarily!

 You expect
p to make money,
y, but given
g the particular
p price
p path
p of the stock
you might not.

Slide 12

12
Where’s your gamma?

 These two price paths have exactly the same vol …


A B

 … but gamma is not constant over time.


 Which one would rather be hedging over? Answer: _________

Slide 13

13
Gamma and Time
X = 100, vol = 16%, r = 4%
0.05

t = 0.25
0.04
0.03
ma
Gamm

0.02

t=1
01
0.0
0.0

60 80 100 120 140


Spot

Slide 14

14
Non-Constant Gamma

 The “path-dependency” problem induced by non-constant gamma over time


is not a serious issue for short-dated options.

 Large volumes generally ensure that the “expected” nature of hedging profits
prevails.

 However,
However because many fewer long
long-dated
dated options are traded this “averaging”
averaging
out may not occur.

Slide 15

15
More Greeks: Vega and Rho

 The change in an option’s value given a change in volatility is known as vega



chg. in option value C
Vega  
chg in volatility
chg. σ

– Also known as tau or kappa.


– Unit of change in vol is an absolute 1% (e.g., 25% to 26%).

 The change in an option’s value given a change in rates is known as rho …

chg. in option value C


Rho    
chg. in rates r

– Unit of change in rates is usually taken to be an absolute 1% (100 bps).

Slide 16

16
Vega and Rho

Question: Why does the put value rise then begin to decline? Slide 17

17
Vega vs. Rho over Time

 Rho increases absolutely with time but vega actually starts to decline …

Slide 18

18
Vega/Put vs. Rho/Put

 Rho will eventuallyy exceed the value of the pput!

Slide 19

19
Hedging Rho

 For short-dated options, and especially in this low rate and low rate-vol
environment, rho is not a big concern for equity-option traders.

 Rho risk that is generated by trading equity options can be hedged with FRAs
and Eurodollar Futures.

 For long-dated
long dated options,
options as we have seen
seen, rho risk becomes significant and this
long-dated rate risk is not as easily hedged.
– A 100 bp change in rates is very unlikely in the next 3 months, but is virtually certain
over the next 10 years.

Slide 20

20
Stochastic Interest Rates

 Interest-rate sensitive derivatives can be priced with a model that allows for the
evolution of rates (“stochastic” interest rates) over time.

 Modelling rates is fundamentally more complex than modelling equities.


– What rate to model?
 Short rate (LIBOR)?
 Ri k f
Risk-free (e.g.,
( Treasury)?
T )?
 Forward rate(s)?
– What distributional assumption?
 Normal?
 Lognormal?

 This has lead to a plethora of models.

Slide 21

21
More Questions

 Do rates mean revert?

 Is vol dependent
p on the level of rates?
– Are the vols of spot rates and forward rates different?

 How important is “fitting the market”?


– What is to be “fit”?
 Zero-coupon bond prices?
 Forward rates?
 Option prices (caps
(caps, floors
floors, swaptions)?

Slide 22

22
Basic Model Structure

 Is a model a model of what is (that is, calibrated to market prices) or what


should be (equilibrium models)?
– Equilibrium
q models: term structure of rates is an output.
p
– Market (“no-arbitrage”) models: term structure is an input.

 Models that “fit the market” all suffer from the same problems as Black-
S h l
Scholes: h
hedging
d i llong-dated
d d iinstruments iis diffi
difficult.
l

 Equilibrium models often yield “unsatisfactory” prices.

Slide 23

23
One-Factor Interest Rate Models

 All of these models do a reasonable job of pricing, but alone do not provide
adequate hedging strategies.
strategies

 Multi-factor models also exist.

Slide 24

24
Variance Swaps

 Variance Swap: a contract that pays the difference between the realized
variance and a fixed level of variance (the “strike”).
– Payoff = M (Realized Variance Ӎ Strike Variance)
M*(Realized

– The “multiplier” M = notional size of the contract (“Variance Units”)


– Really just a Forward.

 Note 1: The strike is quoted in vol terms (20%, for example) even though the
contract is valued as the difference between variances.

 Note 2: Despite how the strike is quoted, M is chosen to achieve a specific


“vega notional” or vega amount (change in value of swap given a 1% change in
vol).

Slide 25

25
Replicating/Hedging a Variance Swap

 The payoff of a variance swap is replicated, and hence hedged, by a weighted


portfolio of out-of-the-money (with respect to the Forward) puts and calls.

Imp. Vol The weights are in inverse


proportion to the square of strike.

Buy all these options …

Strike
Out-of-the-money Puts F Out-of-the-money Calls

Slide 26

26
Short vs. Long-Dated Variance Swaps

 Given the replicating portfolio of a variance swap, it is clear that a reasonably


wide, liquid strip of options is necessary.

 For short-dated swaps ( < 1 year), such options trade with sufficient liquidity on
major indices.

 For long-dated
long dated swaps,
swaps it is hard to hedge properly and market makers thus try
to match buyers with sellers.

Slide 27

27
Valuing an Option: Inputs

 In order to value an option, we need to know …


– Spot price (S)
– Strike Price (X)
– Interest rates (r) (and dividend yield, if applicable)
– Time to expiration (t)
– σ
(Expected) Volatility ( )

S
 X
 
r  Valuation Model
Option Value
V l
  (Black-Scholes?)
t 
σ 

Slide 28

28
Option Pricing Models ?

 Models yield values … markets give us prices!

Slide 29

29
Implied Volatility

 The “character” of the 5 inputs is different: S, X, r, and t are all observable


(and essentially unequivocal), volatility is not.

 What we can observe in the market, however, is the price of an option.

 We can “reverse-engineer” the model to solve for the volatility that would
produce this price (along with the other known inputs) …

Implied Vol Valuation Model Option Price


(Black-Scholes)

Slide 30

30
Realized Volatility

 Realized volatility is a statistical calculation; it is the observed (annualized)


standard deviation of historical returns …

σR  A
r i
2

 A  Periodic Vol
N

– Where ri are log periodic returns (usually daily); e.g. ri = log(Pi /Pi – 1)
– N is the number of returns in the calculation.
– A is the number of periods in a year (A = 252 for daily data).

 “Realized vol” can be different with different choices of the above, especially
p y
the periodicity.

Slide 31

31
What was realized volatility?

 For the year ended August 31, 2010 we get …


– Daily realized = 18.9%
– Weekly realized = 18
18.4%
4%
– Monthly realized = 17.4%

Slide 32

32
Question

 Is implied volatility the expectation of future realized volatility?

Slide 33

33
Realized vs. Implied

Source: Credit Suisse Equity Derivatives Research/Locus Slide 34

34
Implied Vol is “More” than Vol

 Implied volatility, as the only number you can “change” in the Black-Scholes
Model, embeds all the deviations from the assumptions of the model …
– Skewness ((out-of-the-moneyy pput vol is usuallyy > otm call vol))
 Aka “volatility smiles”
– Kurtosis (“fat tails”).
– Inabilityy to hedge
g continuously.
y
– We own options “continuously” but calculate realized vol at a discrete periodicity (you
can buy and sell an option intra-day, making day-to-day realized vol immaterial).

 Implied
I li d voll iis ““contaminated”
t i t d” b by allll off th
the above.
b

Slide 35

35
Questions

 1. If the 2500 largest realized vol days from the last 50 years were
concatenated into a single 10-year period, what would be the realized vol of
this “wild decade”?

– Answer: ____________

 2. What is the realized vol of the past decade (which includes two of the most
volatile markets in history)?

– Answer: ____________

 3.
3 What is the a erage long-term
average long term level
le el of reali
realized
ed vol?
ol?

– Answer: ____________

Slide 36

36
Answers

 1. If the 2500 largest realized vol days from the last 50 years were
concatenated into a single 10-year period, what would be the realized vol of
this “wild decade”?

– Answer: _____29%____

 2. What is the realized vol of the past decade (which includes two of the most
volatile markets in history)?

– Answer: _____22%_____

 3.
3 What is the a erage long-term
average long term level
le el of reali
realized
ed vol?
ol?

– Answer: _____16%_____

Slide 37

37
Pricing on a 29 Vol …

Slide 38

38
Pricing on a 22 Vol …

Slide 39

39
Pricing on a 16 Vol …

Slide 40

40
2005: C3P2-Driven Rise in 10-year Implied Vol

Slide 41

41
10-Year SPX Vol

1999 -------------------------------------------------------------------------- 2006

Slide 42

42
Constant Volatility

 As an alternative to the constant vol input to Black-Scholes, it has often been


proposed that volatility be treated as stochastic.

 The standard geometric Brownian motion process for a stock price, S, as


assumed by Black-Scholes is …

dS  μSdt  σSdW

– Where …
 “mu” is the drift (g
(growth rate)) of the stock.
 “sigma” is the (constant) volatility.
 “dt” is an increment of time.
 “dW” is a random normal process (Weiner Process).

Slide 43

43
Geometric Brownian Motion

 We can “view” the equation on the previous slide …

dS/S

μ (deterministic component)

+
σ (random component, which is “normal”)

time

dt

Slide 44

44
Stochastic Volatility: Heston Model

 Stochastic volatility process …

dS  μSdt  σ t SdW1 Now have a time subscript on vol.

dσ 2t  λ(θ  σ 2t )dt  ησ t dW2 The “volatility process”.

– Where
Wh (all
( ll else
l as b
before)
f )…
 “lambda” is the mean reversion rate for the variance.
 “theta” is the long-run level to which variance reverts.
 “eta” is the volatilityy of variance.
 dW1 and dW2 are correlated Weiner Processes.

 Heston, S. L., “A Closed-Form Solution for Options with Stochastic Volatility with Applications to Bond
and Currency”, The Review of Financial Studies, 1993.

Slide 45

45
Stochastic Vol

 With Heston’s model, we can fit the vol skew and account for fat tails.

 Closed form solutions exist for pplain vanilla European


p options.
p

 Replication is tricky … no truly risk-neutral portfolio.

Slide 46

46
Slide 47

47
Contact Information

 Paul G. Staneski, Ph.D.


 Head of Derivatives Solutions & Training
 212-325-2935
 paul.staneski@credit-suisse.com

Slide 48

48

S-ar putea să vă placă și