Financial Mathematics: I-Liang Chern

FINANCIAL MATHEMATICS
I-Liang Chern
Department of Mathematics
National Taiwan University
and
Chinese University of Hong Kong
December 14, 2016

2
Contents
1 Introduction 1
1.1 Assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Financial Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Forwards contract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.2 Futures (futures contracts) . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.3 Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Payoff functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.1 Premium of an option . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Other kinds of options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Types of traders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Basic assumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Asset Price Model 9

2.1 Efficient market hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 The asset price model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 The discrete asset price model . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.2 The continuous asset price model . . . . . . . . . . . . . . . . . . . . . . 10
2.3 The solution of the discrete asset price model . . . . . . . . . . . . . . . . . . . . 10
2.4 Binomial distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5 Normal distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6 The Brownian motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.6.1 The definition of a Brownian motion . . . . . . . . . . . . . . . . . . . . 14
2.6.2 The Brownian motion as a limit of a random walk . . . . . . . . . . . . . 14
2.6.3 Properties of the Brownian motion . . . . . . . . . . . . . . . . . . . . . 17
2.7 Itô’s Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.8 Geometric Brownian Motion: the solution of the continuous asset price model . . . 21
2.9 Calibrating geometric Brownian motion . . . . . . . . . . . . . . . . . . . . . . . 23
3 Black-Scholes Analysis 25
3.1 The hypothesis of no-arbitrage-opportunities . . . . . . . . . . . . . . . . . . . . . 25
3.2 Basic properties of option prices . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.1 The relation between payoff and options . . . . . . . . . . . . . . . . . . . 26
3
4 CONTENTS
3.2.2 European options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.2.3 Basic properties of American options . . . . . . . . . . . . . . . . . . . . 28
3.2.4 Dividend Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 The Black-Scholes Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3.1 Black-Scholes Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3.2 Boundary and Final condition for European options . . . . . . . . . . . . . 34
3.4 Exact solution for the B-S equation for European options . . . . . . . . . . . . . . 35
3.4.1 Reduction to parabolic equation with constant coefficients . . . . . . . . . 35
3.4.2 Further reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.4.3 Black-Scholes formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.4.4 Special cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.5 Risk Neutrality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.6 Hedging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.6.1 The delta hedging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.7 Time-Dependent r, σ and µ for Black-Sholes equation . . . . . . . . . . . . . . . 44
3.8 Trading strategy involving options . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.8.1 Strategies involving a single option and stock . . . . . . . . . . . . . . . . 45
3.8.2 Bull spreads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.8.3 Bear spreads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.8.4 Butterfly spread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.9 Derivation of heat equation and its exact solution . . . . . . . . . . . . . . . . . . 50
3.9.1 Derivation of the heat equation on R . . . . . . . . . . . . . . . . . . . . . 50
3.9.2 Exact solution of heat equation on R. . . . . . . . . . . . . . . . . . . . . 51
4 Variations on Black-Scholes models 55

4.1 Options on dividend-paying assets . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.1.1 Constant dividend yield . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.1.2 Discrete dividend payments . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2 Futures and futures options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2.1 Forward contracts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2.2 Futures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2.3 Futures options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.4 Black-Scholes analysis on futures options . . . . . . . . . . . . . . . . . . 61
5 Numerical Methods 65
5.1 Monte Carlo method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.2 Binomial Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.2.1 Binomial method for asset price model . . . . . . . . . . . . . . . . . . . 67
5.2.2 Binomial method for option . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.3 Finite difference methods (for the modified B-S eq.) . . . . . . . . . . . . . . . . . 72
5.3.1 Discretization methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.3.2 Binomial method is a forward Euler finite difference method . . . . . . . . 74
5.3.3 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
CONTENTS 5
5.3.4 Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.3.5 Boundary condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.4 Converting the B-S equation to finite domain . . . . . . . . . . . . . . . . . . . . 82
5.5 Fast algorithms for solving linear systems . . . . . . . . . . . . . . . . . . . . . . 83
5.5.1 Direct methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.5.2 Iterative methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6 American Option 89
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.2 American options as a free boundary value problem . . . . . . . . . . . . . . . . . 90
6.2.1 American put option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.2.2 American call option on a dividend-paying asset . . . . . . . . . . . . . . 93
6.3 American option as a linear complementary problem . . . . . . . . . . . . . . . . 94
6.4 *Penalty method for linear complementary problem . . . . . . . . . . . . . . . . . 95
6.5 Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.5.1 Binomial method for American puts . . . . . . . . . . . . . . . . . . . . . 96
6.5.2 Binomial method for American call on dividend-paying asset . . . . . . . 98
6.5.3 *Implicit method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.6 Converting American option to a fixed domain problem . . . . . . . . . . . . . . . 100
6.6.1 American call option with dividend paying asset . . . . . . . . . . . . . . 100
6.6.2 American put option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7 Exotic Options 103

7.1 Binaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
7.2 Compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
7.3 Chooser options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
7.4 Barrier option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
7.4.1 down-and-out call(knockout) . . . . . . . . . . . . . . . . . . . . . . . . . 105
7.4.2 down-and-in(knock-in) option . . . . . . . . . . . . . . . . . . . . . . . . 106
7.5 Asian options and lookback options . . . . . . . . . . . . . . . . . . . . . . . . . 107
8 Path-Dependent Options 109

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
8.2 General Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
8.3 Average strike options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
8.3.1 European calls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
8.3.2 American call options . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
8.3.3 Put-call parity for average strike option . . . . . . . . . . . . . . . . . . . 112
8.4 Lookback Option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
8.4.1 A lookback put with European exercise feature . . . . . . . . . . . . . . . 114
8.4.2 Lookback put option with American exercise feature . . . . . . . . . . . . 114
CONTENTS 1
9 Bonds and Interest Rate Derivatives 115

9.1 Bond Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
9.1.1 Deterministic bond model . . . . . . . . . . . . . . . . . . . . . . . . . . 115
9.1.2 Stochastic bond model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
9.2 Interest models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
9.2.1 A functional approach for interest rate model . . . . . . . . . . . . . . . . 117
9.3 Convertible Bonds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
A Basic theory of stochastic calculus 121

A.1 From random walk to Brownian motion . . . . . . . . . . . . . . . . . . . . . . . 121
A.2 Brownian motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
A.3 Stochastic integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
A.4 Stochastic differential equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
A.5 Diffusion process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
2 CONTENTS
Chapter 1
Introduction
1.1 Assets
Assets represent value of ownership that can be converted into cash. There are two kinds of assets:
tangible and intangible. Commodity, foreign currency, house, building, equipments are tangible,
while copyright, trademarks, patterns, computer programs and financial assets are intangible.
The financial assets include bank deposits, debt instrument, stocks and derivatives. Debt instru-
ments are issued by anyone who borrows money – firms, governments, and households. They in-
clude corporate bonds, government bonds, residential, commercial mortgages and consumer loans.
These debt instruments are also called fixed-income instruments because they promise to pay fixed
sums of cash in the future.
The stocks, shares and equities are all words used to describe what is essentially the same thing.
They are the claim of the ownership of a firm. When someone buys a share, he is buying ownership
of part of a company and becomes a shareholder in that company.
The prices of assets are settled through trading. Many assets are traded in markets. There are
commodity markets as well as financial markets. Financial markets include bond markets, stock
markets, currency markets, financial derivatives markets, etc. The financial market provides a link
between saving and investment. Savers can earn high returns from their saving and borrowers can
execute their investment plans to earn future profits.
The derivatives are contracts derived from some underlying assets. Usually, the prices of assets
fluctuate time by time. In order to reduce the risk of price fluctuation, a corresponding contract
is introduced to make such uncertainty more certain. For instance, suppose today’s price of corn
is US$3 per bushel. You want to buy 5,000 bushels of corn for delivery two months later. You
can sign an agreement with someone who is willing to sell you this amount at such price at such
future time. This agreement is called a forward contract. It certainly costs you some money but
make your price uncertainty more certain. We call that such forward contact hedges the risk of price
fluctuation. Forward contract is one kind of financial derivative. The corn is the underlying asset.
More financial derivatives will be introduced later.
1
2 CHAPTER 1. INTRODUCTION
1.2 Financial Derivatives

1.2.1 Forwards contract
A forward contract is an agreement which allows the holder of the contract to buy or sell a certain
asset at or by a certain day at a certain price. Here,
• the certain day—maturity or expiration date,
• the certain price—delivery price,
• the person who write the contract (has the asset) is called in short position,
• the person who holds the contract is called in long position.
Example 1. Suppose today’s price of corn is US$3 per bushel. You want to buy 5,000 bushels of
corn for delivery two months later. You can sign an agreement with someone who is willing to sell
you this amount at price, say US$3.05, at such future time. Then you make your uncertain risk of
price fluctuation more certain.
Example 2. (quoted from Chan) Suppose it is May 30 now and you need to pay your tuition
to Oxford University £1,000 pound by September 1. Suppose you can earn HK$15,600 in the
summer. The exchange rate now is £1.00 = HK$15.40. What should you do?
• 1. Try your luck: If the exchange rate is less than HK$15.60 by August 31, you earn some
extra cash besides the registration fee. If the exchange rate is higher than HK$15.60, you ... ?
• 2. Buy a forward on British pounds: Look for a forward contract on British pounds that
entitles you to use HK$15,600 to buy £1,000 on August 31. Then you have eliminated, or
hedged, your risk. Thus, a forward contracts eliminate your risk.
1.2.2 Futures (futures contracts)

A futures contract is very similar to a forward contract. Futures contracts are usually traded through
an exchange, or clearinghouse, which standardizes the terms of the contracts. The exchange helps
to eliminate the risk of default of either party through a margin account. Each party has to pay
an initial margin as deposit at the inception of the contract. The profit or loss from the futures
position is calculated every day and the change in this value is paid from one party to the other. The
maintenance margin is the minimum level below which the investor is required to deposit additional
margin.
The difference between forward contract and futures contract are
1.2. FINANCIAL DERIVATIVES 3
Forward Contract Future Contract

nature customized contract standarized contract
Trading over the counter through exchanges
liquidity less liquid highly liquid
counter party risk high negligible
Settlement delivery (at the end) closed out prior to maturity
Margin no margin compulsorily needs to be paid by the parties
Example Mr. Chan takes a long (buy) position of one contract in corn (5,000 bushels) for March
delivery at a price of US$3.682 per bushel at the Chicago Board of Trade (CBOT). See quotation
at http://www.cbot.com/.CBOT It requires maintenance margin of US$700 with an initial
margin markup of 135%, i.e. the initial margin is US$945 which Mr. Chan and the seller each has
to deposit into the broker’s account on the first day they enter the contract. The next day the price
of this contract drops to US$3.652. This represents a loss of US$0.03 × 5,000 = US$150. The
broker will take this amount from Mr. Chan’s margin account and deposit it to the seller’s margin
account. It leaves Mr. Chan with a balance of US$795. The following day the price drops again to
US$3.552. This represents an additional loss of US$500, which is again deducted from the margin
account. As this point the margin account is US$295, which is below the maintenance level. The
broker calls Mr. Chan and tells him that he must deposit at least US$405 in his margin account,
or his position will be closed out, i.e. both sides agree to settle the contract at this point. Once the
position is closed, Mr. Chan will not be able to earn back any money in the future even if the price
rises above US$3.682.
1.2.3 Options
There are two kinds of options — call options and put options. A call (put) option is a contract
between two parties, in which the holder has the right to buy (sell) and the writer has the obligation
to sell (buy) an asset at certain time in the future at a certain price. The price is called the exercise
price (or strike price). The holder is called in long position, while the writer is called in short
position. The underlying assets of an option can be commodity, stocks, stock indices, foreign
currencies, or future contracts.
There are two kinds of exercise features:
• European options : Options can only be exercised at the maturity date.
• American options : Options can be exercised any time up to the maturity date.
Notation
• t current time
• T maturity date
• S current asset price

• ST asset price at time T
• E strike price
• c premium, the price of call option
• r bank interest rate
Examples An investor buys 100 European call options on XYZ stock with strike price $140.
Suppose
E = 140,
St = 138,
T = 2 months,
c = 5 (the price of one call option).
If at time T , ST > E, then he should exercise this option. The payoff is 100 × (ST − E) =
100 × (146 − 140) = 600. The premium is 5 × 100 = 500. Hence, he earns $100. If ST ≤ T , then
he should not exercise his call contracts. The payoff is 0.
The payoff function for a call option is Λ = max{ST − E, 0}. One needs to pay premium (ct )
to buy the options. Thus the net profit from buying this call is
Λ − ct er(T −t) .
Example Suppose
today is t = 03/09/2016,
expiration is T = 31/12/2016 ,
the strike price E = 250
for some stock. If ST = 270 at expiration, which is bigger than the strike price, we should exercise
this call option, then buy the share for 250, and sell it in the market immediately for 270. The payoff
Λ = 270 − 250 = 20. If ST = 230, we should give up our option, and the payoff is 0. Suppose the
share take 230 or 270 with equal probability. Then the expected profit is
1 1
× 0 + × 20 = 10.
2 2
Ignoring the interest of bank, then a reasonable price for this call option should be 10. If ST = 270,
then the net profit= 20 − 10 = 10. This means that the profits is 100% (He paid 10 for the option).
If ST = 230 the loss is 10 for the premium. The loss is also 100%. On the other hand, if the
investor had instead purchased the share for 250 at t, then the corresponding profit or loss at T is
±20. Which is only ±8% of the original investment. Thus, option is of high risk and with high
return.
1.3. PAYOFF FUNCTIONS 5
1.3 Payoff functions

At the expiration day, the payoff of a future or an option is the follows.
Payoff of futures
• Payoff of a future in long position: At the expiration day, the price of the asset is ST . The
holder can buy the asset at price E. He has the obligation to buy it. Thus the payoff of the
holder is ST − E.
• Payoff of a future in short position: at expiration, the writer also has the obligation to sell the
asset at price E. So, he needs to use ST to buy the asset, then sell to the holder at price E.
Thus his payoff is E − ST .
Λ Λ
K K
ST ST
future (long) future (short)
(a) left (b) right
Figure 1.1: Payoff of a future, long position (left) and short position (right)
Payoff of call options
• Long call: Λ = max{ST − E, 0}
– If ST > E, then the holder exercise the call at price E then sell at price ST
– If ST ≤ E, then he gives up the call option.
• Short call: Λ = min{E − ST , 0}
– If ST > E, he has the obligation to sell the asset at price E, thus he needs to buy the
asset at price ST then sell it at price E. He losses ST − E.
– If ST ≤ E, the holder will give up the call option. So the writer losses nothing.
Λ Λ
K K
ST ST
call option (long):max{ST−K,0} call option (short):−max{ST−K,0}=min{K−ST,0}
(a) left (b) right
Figure 1.2: Payoff of a call, long position (left) and short position (right)
Payoff of put options
• Long put: Λ = max{E − ST , 0}
– If ST < E, the holder has the right to sell the asset at E, then buy it back at ST . Thus
the payoff is E − ST .
– If ST ≥ E, the holder just gives up the option.
• Short put: Λ = min{ST − E, 0}
– If ST < E, he has the obligation to buy the asset at E. He then sell it at ST . Thus he
losses E − ST .
– If ST ≥ E, the holder will give up the put option, So the writer losses nothing.
1.3.1 Premium of an option

The person who hold an option has the right and the counter party has the obligation to fulfill the
contract. Thus, a premium should be paid by the holder to the writer. Below is a portion of a call
option copied from the Financial Times.
the current time t = Feb 3

the expiration T = end of Feb,
T − t ≈ 10 days
St = 2872
E 2650 2700 2750 2800 2850 2900 2950 3000

c 233 183 135 89 50 24 9 3
1.4. OTHER KINDS OF OPTIONS 7
Λ Λ
K K
ST ST
put option (long):max{K−ST,0} put option (short):−max{K−ST,0}=min{ST−K,0}
(a) left (b) right
Figure 1.3: Payoff of a put, long position (left) and short position (right)
1.4 Other kinds of options

• Barrier option: The option only exists when the underlying asset price is in some prescribed
value before expiry.
• Asian option: It is a contract giving the holder the right to buy or sell an asset for its average
price over some prescribed period.
• Look-back option: The payoff depends not only on the asset price at expiry but also its
maximum or minimum over some period price to expiry. For example, Λ = max{J − S0 , 0},
J = max0≤τ ≤T S(τ ).
1.5 Types of traders

1. Speculators (high risk, high rewards)
2. Hedgers (to make the outcomes more certain)
3. Arbitrageurs (Working on more than one markets, p12, p13, p14, Hull).
1.6 Basic assumption

Arbitrage opportunities cannot last for long. Only small arbitrage opportunities are observed in
financial markets. Our arguments concerning future prices and option prices will be based on the
assumption that “there is no arbitrage opportunities”.
250
c
200
150
100
50
−50
2600 2650 2700 2750 2800 2850 2900 2950 3000 3050
Figure 1.4: The FT-SE index call option values versus exercise price.
Chapter 2
Asset Price Model
2.1 Efficient market hypothesis

The asset prices move randomly because of the following efficient market hypothesis:
1. The past history is fully reflected in the present price, which does not hold any future infor-
mation. This means the future price of the asset only depends on its current value and does
not depends on its value one month ago, or one year ago. If this were not true, technical
analysis could make above-average return by interpreting chart of the past history of the asset
price. This contradicts to the hypothesis of no arbitrage opportunities. In fact, there is very
little evidence that they are able to do so.
2. Market responds immediately to any new information about an asset.
2.2 The asset price model

We shall introduce a discrete model and a continuous model. We will show that the continuous
model is the continuous limit of the discrete model.
2.2.1 The discrete asset price model

The time is discrete in this model. The time sequence is n∆t, n ∈ N. Let us denote the asset price
at time step n by Sn . We model the asset price by

Sn+1 u with probability p
= (2.1)
Sn d with probability 1 − p.
Here, 0 < d < 1 < u. The information we are looking for is the following transition probability
P (Sn = S|S0 ), the probability that the asset price is S at time step n with initial price S0 . We shall
find this transition probability later.
9
10 CHAPTER 2. ASSET PRICE MODEL
2.2.2 The continuous asset price model

Let us denote the asset price at time t by S(t). The meaningful quantity for the change of an asset
price is its relative change
dS
,
S
dS
which is called the return. The change S can be decomposed into two parts: one is deterministic,
the other is random.
• Deterministic part: This can be modeled by
dS
= µdt.
S
Here, µ is a measure of the growth rate of the asset. We may think µ is a constant during the
life of an option.
• Random part: this part is a random change in response to external effects, such as unexpected
news. It is modeled by a Brownian motion
σdz,
the σ is the order of fluctuations or the variance of the return and is called the volatility. The
quantity dz is sampled from a normal distribution which we shall discuss below.
The overall asset price model is then given by
dS
= µdt + σdz. (2.2)
S
We shall look for the transition probability density function P(S(t) = S|S(0) = S0 ). Or equiva-
lently, the integral
Z b
P(S(t) = S|S(0) = S0 ) dS
a
is the probability that the asset price S(t) lies in (a, b) at time t and is S0 initially.
2.3 The solution of the discrete asset price model

Let us consider the discrete price model

Sn+1 u with probability 1/2
=
Sn d with probability 1/2.
A sequence of movements (S0 , S1 , ..., Sn ) is called an n-step path. In such a path, it can consist of
` times up movements and n − ` times down movements, where 0 ≤ ` ≤ n. The corresponding
2.4. BINOMIAL DISTRIBUTION 11
values of Sn are S0 u` dn−` . Since the probability of each movement is independent, the probability
of an n-step path with ` up movements (and n − ` down movements) is
` n−` n
1 1 1
= .
2 2 2
There are n` paths with ` up movements in n-step paths, we then obtain the transition probability

of the asset price to be:

n 1 n
` 2 when S = S0 u` dn−` ,
P (Sn = S|S0 ) = (2.3)
0 otherwise.
Such Sn is called a (discrete) random walk. It is hard to deal with the values S0 u` dn−` . Instead, we
take its logarithmic function. Let us denote ln Sn by Xn . This random walk Xn obeyes the rule

ln u with probability 1/2
Xn+1 − Xn =
ln d with probability 1/2.
Let us define
ln u − ln d ln u + ln d
∆x = , ln(1 + r) = .
2 2
Then the rule can be written as

∆x with probability 1/2
Xn+1 − Xn = ln(1 + r) +
−∆x with probability 1/2.
The term ln(1 + r) ≈ r is called a drift term. It measures the growth of Sn . It can be absorpt into
Xn . Indeed, if we define Zn = ln((1 + r)−n Sn ), then Zn satisfies

∆x with probability 1/2
Zn+1 − Zn = (2.4)
−∆x with probability 1/2.
The advantage of this formulation (using Zn now) is that this random walk has equal increment ∆x,
the possible values of the random variable Zn are m∆x, m ∈ Z. This is the random walk with
binomial distribution.
2.4 Binomial distribution

Consider a particle moves randomly on an one dimensional lattice {m∆x|m ∈ Z} according to
the rule (2.4), where Zn denotes the location of this particle at time step n. Suppose the particle is
located at 0 initially. Our goal is to find the transition probability
P (Zn = m∆x|Z0 = 0)
explicitly and also find its properties.

This random walk is equivalent to the following binomial trials: flipping a coin n times. Suppose
we have equal probability to get Head or Tail in each toss. We move the particle to the right or left
adjacent grid point according to whether we get Head or Tail. Let Xk be the result of the kth
experiment. Namely,
1 if we get Head
Xk =
−1 if we get tail
Pn
Then Zn /∆x = k=1 Xk . The random walk Zn is equivalent to the random walk Yn , the number
of Heads we get in n trials. Namely,
P (Zn = m∆x|Z0 = 0) = P (Yn = `)
with
1
m = ` − (n − `) = 2` − n, or equivalently ` = (n + m).
2
Notice that m is even (odd), when n is even (odd). There is a one-to-one correspondence between
{` | 0 ≤ ` ≤ n} ↔ {m | − n ≤ m ≤ n, m + n is even}.
Thus, we can also use ` instead m to label our particle. There are n` paths to reach m∆x from 0

with ` = (n + m)/2, we get

0, 1 n if m + n is odd,
P (Zn = m∆x|Z0 = 0} = n
(n+m)/2 ( 2 ) , if m + n is even.
Or equivalently,
` n−`
n 1 1
P (Yn = `) = .
` 2 2
It is clear that
n n
X 1 1
P (Yn = `) = + = 1.
2 2
`=0
This probability distribution is called the binomial distribution. We denote them by pZn and pYn ,
respectively.
Properties of the binomial distribution We shall compute the moments of pZn or pYn . They are
defined by
X n X n
< `k >:= `k pYn (`), < mk >:= mk pZn (m)
`=0 m = −n
m + n even
Since m = 2` − n, we can find the moments < mk >=< (2` − n)k > by computing < `k >,
which in turn can be computed through the help of the following moments generating function:
X X 1 n n 1 + s n
`
G(s) := s pYn (`) = s` = .
2 ` 2
` `
2.5. NORMAL DISTRIBUTION 13
To compute moments, for instance

n
X
0
G (1) = ` · pYn (`) =< ` > .
`=0
n
X
G00 (1) = `(` − 1)pYn (`) =< `2 > − < ` > .
`=1
Hence, we obtain the mean
n
< ` >= G0 (1) = .
2
The second moment
n(n + 1)
< `2 >= G00 (1)+ < ` >= .
4
We can also compute the mean and variance of pZn as the follows. From m = 2` − n, we have
< m >= 2 < ` > −n = 0.
To compute the second moment < m2 >, from m = 2` − n, we have
< m2 >= 4 < `2 > −4n < ` > +n2 = n.
The mean of this random walk is < m >= 0, while its variance is < (m− < m >)2 >= n.
Exercise
1. Find the transition probability, mean and variance for the case

∆x with probability p
Zn+1 − Zn =
−∆x with probability 1 − p
2.5 Normal distribution

A random variable Z is said to be distributed as standard normal if its probability density is
1 2
pZ (x) = √ e−x /2 .
2π
We denote it as Z ∼ N (0, 1). A random variable X is said to be distributed as normal with mean µ
and variance σ 2 if its probability density function is
1 2 /2σ 2
pX (x) = √ e−(x−µ) .
2πσ 2
We denote it by X ∼ N (µ, σ 2 ). For such X, it is clearly that
X −µ
Z := ∼ N (0, 1).
σ
For
µ+σb b
X −µ
Z Z
1 −(x−µ)2 /(2σ 2 ) 1 2
P (a ≤ < b) = P (µ+σa ≤ X < µ+σb) = √ e dx = √ e−z /2 dz
σ µ+σa 2πσ 2 a 2π
The moments of Z (i.e. E[Z m ]) can be computed through the helps of moment generating function
∞ ∞
sm
Z
X
sZ 1 X
G(s) := E[e ] = sm xm pZ (x) dx = E[Z m ].
m! R m!
m=0 m=0
We have
∞ ∞ ∞
s2k
Z Z
1 sx −x2 /2 s2 /2 1 2 /2 2 /2
X
G(s) = √ e e dx = e √ e−(x−s) dx = es = .
2π −∞ 2π −∞ 2k k!
k=0
By comparing the two expansions above, we obtain

(
0 if m is odd
E[Z m ] = (2k)!
2k k!
if m = 2k.
2.6 The Brownian motion

2.6.1 The definition of a Brownian motion
The definition of the Brownian motion (or the standard Wiener process) z(t) is the following:
Definition 2.1. A time dependent function z(t), t ∈ R is said to be a Brown motion if
(a) ∀t, z(t) is a random variable.
(b) z(t) is continuous in t.
(c) For any u > 0, s > 0, the increment z(t + s) − z(t), z(t) − z(t − u) are independent.
(d) ∀s > 0, the increment zt+s − zt is normally distributed with mean zero and variance s, i.e.
zt+s − zt ∼ N (0, s).
2.6.2 The Brownian motion as a limit of a random walk

We may realize the Brownian motion as the limit of the standard random walk. Let us study Brown-
ian motion in [0, t]. First, we partition the time interval [0, t] into n subintervals evenly. Let us imag-
ine a particle moves√ randomly on lattice points {m∆x|m ∈ Z} at discrete time k∆t, k = 0, ..., n.
We choose ∆x = ∆t. The motion of the particle is:
Z0 = 0,
2.6. THE BROWNIAN MOTION 15
Zk = Zk−1 + Xk ∆x
where
1 with probability 1/2
Xk =
−1 with probability 1/2
The location Zn at time step n obeys the binomial distribution pZn (m) = P (Zn = m∆x|Z0 = 0).
We can connect discrete path (Z0 , Z1 , ..., Zn ) by linear function and form a piecewise linear path
Z n (s), 0 ≤ s ≤ t with Z n (k∆t) = Zk . As n → ∞, we expect that these kinds of paths tend to
a class of zig-zag paths z(·), called Brownian motions. They are time-dependent random variable.
Their properties are:
• z(0) = 0;
• z(·) is continuous;
• Any non-overlapping increments z(t4 ) − z(t3 ) and z(t2 ) − z(t1 ) (t1 < t2 < t3 < t4 ) are
independent.
Let us denote the collection of these “zig-zag” paths by Ω, the sample space of the Brownian mo-
tions.
Next, we find the probability distribution of Brownian motions. The probability mass function
of the random walk is P (Zn = m∆x|Z0 = 0}. Since n + m is even, the corresponding probability
n := P (Z = m∆x|Z = 0)/(2∆x). We claim that
density function on the real line is Pm n 0
Proposition 2.1. The probability density function of the random walk has the limit
n 1 −x2 /(2t)
Pm →√ e
2πt
as n → ∞ with m∆x → x and n∆t = t.
This is the probability density of the standard normal distribution N (0, t). Thus, the 4th property
of the Brownian motion is
• The probability distribution of z(t) is z(t) ∼ N (0, t).
We prove this proposition by using the Stirling formula:
√ 1
n! ≈ 2πnn+ 2 e−n .
Recall that the probability
n
n 1
P (Zn = m∆x|Z0 = 0) = 1 .
2 (m + n) 2
Using the Stirling formula, we have for n, m, n − m >> 1,
n n
n 1 1 n!
1 =
2 (m + n) 2 2 ( 2 (n + m))!( 12 (n − m))!
1
n √ 1
1 2πnn+ 2 e−n
≈ √ 1 1 √ 1 1
2 2π( 12 (n + m)) 2 (n+m)+ 2 e−(n+m)/2 2π( 12 (n − m)) 2 (n−m)+ 2 e−(n−m)/2
1/2
m − 12 (n+m)− 12 m − 12 (n−m)− 12

2
= 1+ 1−
πn n n
We shall use the formula
x n
→ ex , as n → ∞.
1+
n
We may treat m as a real number
√ because the functions above are smooth functions. Now, we fix
x, t, define m = x/∆x = Cn, where C = x2 /t. We take n → ∞. Then
−n/2
m2 C −n/2

m −n/2 m −n/2
1+ 1− = 1− 2 = 1− → eC/2
n n n n
√
√ !− Cn/2
m −m/2 C
1+ = 1+ √ → e−C/2
n n
√
√ ! Cn/2
m m/2 C
1− = 1− √ → e−C/2
n n
−1/2
m2 C −1/2

m −1/2 m −1/2
1+ 1− = 1− 2 = 1− →1
n n n n
Thus, we obtain
m − 21 (n+m)− 21 m − 12 (n−m)− 21 2
1+ 1− → e−C/2 = e−x /(2t) .
n n
We also have q
2
πn 1 1
=√ √ =√ .
2∆x 2πn ∆t 2πt
Summarizing above, we obtain
n
n 1 n 1 1 −x2 /(2t)
Pm = 1 →√ e .
2∆x 2 (m + n) 2 2πt
Remarks.
1. We can also show this limiting process as a corollary of the central limit theorem. Recall that
n
X
Zn = Xk ∆x,
k=1
where Xk ∼ X are iid and

1 with probability 1/2
X=
−1 with probability 1/2
2.6. THE BROWNIAN MOTION 17
√ p
Since ∆x = ∆t = t/n, we get
n
1 X√
Zn = √ tXk .
n
k=1
√ √
The mean < tX >= 0, the variance
√ < ( tX)2 >= t. Thus, by the central limit theorem,
the random variable √1n nk=1 tXk converges in distribution to a normal N (0, t):
P
n
1 X√
√ tXk → N (0, t).
n
k=1
2. We can consider more general random walk

n
X
Zn = Xk σ∆x,
k=1
where Xk ∼ X are i.i.d. The parameter σ, called volatility, measures the variation of Z in
each movement. In this case,
n
1 X√
Zn = √ tσXk → N (0, σ 2 t).
n
k=1
2.6.3 Properties of the Brownian motion

By (d) of the definition
1 − x2
P(z(t) = x|z(0) = 0) = √ e 2t .
2πt
Let us consider the Brownian motion starting from y at time s and reaches x at later time t. The
probability density is denoted by p(s, y, x, t), called the transition probability density of the stochas-
tic process z(t):
p(s, y, t, x) = P(z(t) = x|z(s) = y) = P(z(t) = x − y|z(s) = 0)

1 2
= P(z(t − s) = x − y|z(0) = 0) = p e−(x−y) /2(t−s) .
2π(t − s)
Let us simplify the notation p(0, 0, t, x) by p(t, x). This transition probability density is the proba-
bility density function for Brownian motions z(t) starting from 0 at time 0. Any function f (z), its
expectation is defined to be
Z ∞
E[f (z(t))] := f (x)p(t, x) dx.
−∞
This transition probability function has the following properties.

1. E[z(t)] = 0 Z ∞
1 −x2 /(2t)
x√ e dx = 0.
−∞ 2πt
2. E[z(t)2 ] = t Z ∞
1 −x2 /(2t)
x2 √ e dx = t.
−∞ 2πt
3. Let us consider an infinitesimal change dz := z(t + dt) − z(t). Here, dt is an infinitesimal
time increment. Then we have
(dz)2 = dt with probability 1. (2.5)
This formula is very important for the stochastic calculus below. It is interpreted by integrat-
ing the above equation in t: Z t
(dz)2 = t.
0
Pn
The left-hand side is defined to be the limit of the Riemann sum k=1 (z(tk ) − z(tk−1 ))2 .
Proposition 2.2. Let z(·) be the Brownian motion. Then
n
X
lim (z(tk ) − z(tk−1 ))2 = t with probability 1
n→∞
k=1
as n → ∞. Here tk := kt/n.
Proof. 1. We shall apply the strong law of large numbers. Let us consider the random variables:
Yk := n(z(tk ) − z(tk−1 ))2 . We see that
n n
X 1X
(z(tk ) − z(tk−1 ))2 = Yk .
n
k=1 k=1
The random variables Yk are i.i.d. Yk ∼ Y , where
Y = n(z(t/n) − z(0))2 .
2. The mean of Y is
Z
1 2
E[Y ] = nx2 p e−x /(2t/n) dx = n(t/n) = t.
2π(t/n) 2
The variance of Y is also finite (check). That is
E[(Y − t)2 ] < ∞, (2.6)
where Z ∞
1 2
E[f (Y )] = f (nx2 ) p e−x /(2t/n) dx
−∞ 2π(t/n)
2.7. ITÔ’S LEMMA 19
3. By the strong law of large numbers
n
1X
Yk → t with probability 1.
n
k=1
Exercise
1. Find E[z(t)m ].
2. Check (2.6).
2.7 Itô’s Lemma

In this section, we shall study differential equations which consist of deterministic part: ẋ = b(x),
and stochastic part σ ż(t). Here, z(t) is the Brownian motion. We call such an equation a stochastic
differential equation (s.d.e.) and expressed as
dx(t) = b(x(t))dt + σ(x(t))dz(t). (2.7)
An important lemma for finding their solution is the following Itô’s lemma.
Lemma 2.1 (Itô). Suppose x(t) satisfies the stochastic differential equation (2.7), and f (x, t) is a
smooth function. Then f (x(t), t) satisfies the following stochastic differential equation:

1 2
df = ft + bfx + σ fxx dt + σfx dz. (2.8)
2
Proof. According to the Taylor expansion,
1 1
df = ft dt + fx dx + ftt (dt)2 + fxt dx dt + fxx (dx)2 + · · · .
2 2
Plug (2.7) into this equation. The term
(dx)2 = b2 (dt)2 + 2bσdt · dz + σ 2 (dz)2 .
In the Taylor expansion of df , the terms (dt)2 , dt · dz are relative unimportant as comparing with the
dt term and (dz)2 term. Using (2.7) and noting (dz)2 = dt with probability 1, we obtain (2.8).
Example 1 The stochastic process
x(t) := x0 + at + σz(t)
solves the s.d.e.

dx = adt + σdz,
where a and σ are constants. By letting y = x − at, then y is a function of x and t. From Itô’s
lemma, y(x(t), t) is also a stochastic process and y satisfies dy = σdz. Since σ is a constant, we
then have y(t) = y0 + σz(t). The transition probability density function for y is
1 2 /2σ 2 t
P(y(t) = y|y(0) = y0 ) = √ e−(y−y0 ) .
2πσ 2 t
Or equivalently, the transition probability density function for x is
1 2 /2σ 2 t
P(x(t) = x|x(0) = x0 ) = √ e−(x−at−x0 ) .
2πσ 2 t
Example 2. Let y(t) = y0 + z(t)2 . Then y(t) solves
dy = dt + 2z(t)dz(t).
To show this, we apply Itô’s lemma. Let y = f (x, t) = x2 , and x = z. Then

1
dy(t) = fx dx + fxx dt = 2z(t)dz(t) + dt.
2
This means
d(z 2 ) = 2zdz + dt.
Example 3. Let dx = dz and y = ln x. Then fx = 1/x and fxx = −1/x2 . Thus, we have
1 1
d ln x = dx − 2 (dx)2 .
x 2x
In terms of Brownian motion, it can be written as
dz dt
d ln z = − 2.
z 2z
Thus, ln z solves the above s.d.e.
Example 4. Let y = eσz , σ is a constant. We choose dx = σdz and y = ex . Then

1 1
dex = ex dx + ex (dx)2 = ex σdz + ex σ 2 dt.
2 2
Thus
1
deσz = eσz σdz + eσz σ 2 dt.
2
2.8. GEOMETRIC BROWNIAN MOTION: THE SOLUTION OF THE CONTINUOUS ASSET PRICE MODEL21
2.8 Geometric Brownian Motion: the solution of the continuous asset

price model
In this section, we want to find the transition probability density function for the continuous asset
price model:
dS = µS dt + σS dz. (2.9)
with initial data S(0) = S0 . We apply Itô’s lemma with x = f (S) = ln S. Then, fS = 1/S and
fSS = −1/S 2 . By Itô’s lemma, x satisfies the s.d.e.
σ2

1
dx = fS dS + fSS (dS)2 = µ− dt + σ dz.
2 2
and x(0) = x0 := ln S0 . From Example 1 of the previous section, we obtain
σ2
x(t) = x0 + (µ − )t + σz(t).
2
Since S = ex , we obtain
σ2

S(t) = S0 exp (µ − )t + σz(t) . (2.10)
2
The probability density of x(t) is
1 σ2
)t)2 /2σ 2 t
P(x(t) = x|x(0) = x0 ) = √ e−(x−x0 −(µ− 2 .
2πσ 2 t
That is,
σ2
x(t) ∼ N (x0 + (µ − )t, σ 2 t).
2
From
P(S(t) = S|S(0) = S0 )dS = P(x(t) = x|x(0) = x0 )dx

= P(x(t) = x|x(0) = x0 )dS/S
we obtain that the transition probability density function for S(t) is
2
1 −(ln S
−(µ− σ2 )t)2 /2σ 2 t
P(S(t) = S|S(0) = S0 ) = √ e S0 . (2.11)
2πσ 2 tS
This is called the log-normal distribution. The stochastic process x(t) obeys the normal distribution,
while S(t) = ex(t) is called to obey the geometric Brownian motion.
Remark. Alternatively, we may also use probability distribution function to derive the probability
density function as below. The probability distribution function of S(t) is
Z S
FS (S) = P(S(t) = S1 |S(0) = S0 ) dS1
−∞
On the other hand, Z x

Fx (x) = P(x(t) = x1 |x(0) = x0 ) dx1
−∞
We notice that
FS (S) = Fx (x) for x = ln S, x0 = ln S0 .
Thus,
d d dx d
P(S(t) = S|S(0) = S0 ) = FS (S) = Fx (x) = Fx (x)
dS dS dS dx
2
1 1 −(ln SS −(µ− σ2 )t)2 /2σ 2 t
= P(x(t) = ln S|x(0) = ln S0 ) = √ e 0 .
S 2πσ 2 tS
S0 S
Figure 2.1: The log-normal distribution.
Properties of the geometric normal distribution S(t) Let us denote the transition probability
density P(S(t) = s|S(0) = S0 ) by pS(t) (s) for short. That is
2
1 −(ln s
−(µ− σ2 )t)2 /2σ 2 t
pS(t) (s) = √ e S0
2πσ 2 ts
We compute its mean and variance below.
2.9. CALIBRATING GEOMETRIC BROWNIAN MOTION 23
Proposition 2.3. The mean and variance of the geometric Brownian motion are
R∞
(a) E[S(t)] = −∞ spS(t) (s) ds = S0 eµt ,
2
(b) Var[S(t)] = S02 e2µt [eσ t − 1].
Proof. (a)
Z ∞ Z ∞ 2
1 −(ln s
−(µ− σ2 )t)2 /2σ 2 t
E[S(t)] = spS(t) (s) ds = √ e S0 ds
Z0 ∞ 0 2πσ 2 t
1 σ2
)t)2 /2σ 2 t
= √ ex e−(x−x0 −(µ− 2 dx
2πσ 2 t
Z−∞
∞
1 σ2 2 2
= √ ex+x0 +µt e−(x+ 2 t) /2σ t dx
−∞ 2
2πσ t
Z ∞
1 σ2 2 2
= S0 eµt √ ex−(x+ 2 t) /2σ t dx
2
2πσ t
Z−∞
∞
1 σ2 2 2
= S0 eµt √ e−(x− 2 t) /2σ t dx
−∞ 2πσ 2 t
µt
= S0 e .
(b)
Z ∞ Z ∞ 2
2 2 1 −(ln s
−(µ− σ2 )t)2 /2σ 2 t
E[S(t) ] = s pS(t) (s) ds = s√ e S0 ds
Z0 ∞ 0 2πσ 2 t
1 σ2
)t)2 /2σ 2 t
= √ e2x e−(x−x0 −(µ− 2 dx
2πσ 2 t
Z−∞
∞
1 σ2 2 2
= √ e2(x+x0 +µt) e−(x+ 2 t) /2σ t dx
−∞ 2
2πσ t
Z ∞
1 σ2 2 2
= S02 e2µt √ e2x−(x+ 2 t) /2σ t dx
−∞ 2πσ 2 t
Z ∞
1 3σ 2 2 2 2
= S0 eµt √ e−(x− 2 t) /2σ t+σ t dx
−∞ 2
2πσ t
2
= S02 e2µt+σ t .
2
Var[S(t)] = E[S(t)2 ] − E[S(t)]2 = S02 e2µt [eσ t − 1].
2.9 Calibrating geometric Brownian motion

How do we obtain µ and σ of an asset? We estimate them from the historical data Si , at discrete
time ti := i∆t, i = 0, ..., n. The procedure is as below. First, let x(t) = ln S(t). Let ∆xi :=
x(ti+1 ) − x(ti ). The variables ∆xi are independent random variables with identical distribution:
σ2
∆xi ∼ ∆x ∼ N ((µ − )∆t, σ 2 ∆t).
2
The historical data we collect are Si which is a sample of S(ti ). We define xi = ln Si . We estimate
the mean and variance of ∆x by using these historical data xi :
n
σ̂ 2

1X
µ̂ − ∆t = (xi+1 − xi ) := m̂,
2 n
i=1
n
1 X
σ̂ 2 ∆t = ((xi+1 − xi ) − m̂)2 .
n−1
i=1
The µ̂ and σ̂ are the estimators of µ and σ. 1
1
The variance estimator has n − 1 in denominator instead of n. This is called unbiased estimator because the expec-
tation of this estimator is the correct variance. This is called Bessel’s correction.
Chapter 3
Black-Scholes Analysis
3.1 The hypothesis of no-arbitrage-opportunities

The option pricing theory was introduced by Black and Scholes. The fundamental hypothesis of
their analysis is that ”there is no arbitrage opportunities in financial markets”.
For simplicity, we shall also assume
1. There exists a risk-free investment that gives a guaranteed return with interest rate r. ( e.g.
government bond, bank deposit.)
2. Borrowing or lending at such riskless interest rate is always possible.
3. There is no transaction costs.
4. All trading profits are subject to the same tax rate.
We will use the following notations:
S current asset price

E exercise price
T expiry time
t current time
µ growth rate of an asset
σ volatility of an asset
ST asset price at T
r risk-free interest rate
c value of European call option
C value of American call option
p value of European put option
P value of American put option
Λ the payoff function
25
26 CHAPTER 3. BLACK-SCHOLES ANALYSIS
Remark.
1. In assumption 1, if the annual interest rate is r and the compound interest is counted daily,
then the principal plus the compound interest in one year are
r 365
A 1+ ≈ Aer .
365
2. Assumption 1 implies that an amount of money A at time t has value Aer(T −t) at time T if it
is deposited into bank. This is called the future value of A. Similarly, a strike price E at time
T should be deducted to Ee−r(T −t) at time t. This is called its present value.
3. The option price c, p, C and P are functions of (St , t). Usually, we only write c, which means
the current value c(St , t). We use the same convention for S, an abbreviation of the current
value St .
3.2 Basic properties of option prices

3.2.1 The relation between payoff and options
1. The payoff of a contract is the return from the deal. For European option, it can only be
exercised on expiry date. Therefore, its payoff Λ is only meaningful at the expiry date T .
However, for American option, the option can be exercised at anytime between t and the
expiration date T . It is therefore meaningful to define the payoff function at t. For a person
who longs an American call option, his payoff is Λ(t) = max(St −E, 0), while for American
put option, Λ(t) = max(E − St , 0).
2. c(ST , T ) = Λ(T ) = max(ST − E, 0).

Otherwise, there is a chance of arbitrage. For instance, if c(ST , T ) < Λ(T ), then we can buy
a call on price c, exercise it immediately. If ST > E, then Λ = ST − E > 0 and c < Λ by
our assumption. Hence we have an immediate net profit ST − E − c > 0. This contradicts
to our hypothesis. If ST ≤ E, then Λ(T ) = 0 and c < Λ(T ) = 0 leading to c has negative
value. This means that anyone who can buy this call option with negative amount of money.
This again contradicts to our hypothesis.
If c(ST , T ) > Λ(T ), we can short a call and earn c(ST , T ). If the person who buy the call
does not claim, then we have net profit c. If he does exercise his call, then we can buy an asset
from the market on price ST and sell to that person with price E. The cost to us is ST − E.
By doing so, the net profit we get is c(ST , T ) − (ST − E) > 0. Again, this is a contradiction.
3. p(ST , T ) = Λ(T ) = max(E − ST , 0).

If p(ST , T ) < Λ(t), then an arbitrageur longs a put at T , exercises it right away, and get net
profit Λ(T ) − p(ST , T ). If p(ST , T ) > Λ(T ), then an arbitrageur shorts a put. If the person
who buys the put exercise it, then the arbitrageur earns P (ST , T ) − Λ(T ); if he gives up this
put, then the arbitrageur earns p(ST , T ).
3.2. BASIC PROPERTIES OF OPTION PRICES 27
4. We show C(St , t) = Λ(t) := max(St − E, 0).

If C(St , t) > Λ(t), then an arbitrageur can short the call and earn C(St , t). If the person who
exercises it right away, the arbitrageur need to pay Λ(t). He still has net profit C(St , t)−Λ(t).
If C(St , t) < Λ(t). Then an arbitrageur buy C and exercises it right away and get net profit
Λ(t) − C(St , t).
5. P (St , t) = Λ(t) := max(E − St , 0). This can be shown similarly.
3.2.2 European options

Theorem 3.1. The following statements hold for European options
max{S − Ee−r(T −t) , 0} ≤ c ≤ S (3.1)

−r(T −t) −r(T −t)
max{Ee − S, 0} ≤ p ≤ Ee (3.2)
and the put-call parity

p + S = c + Ee−r(T −t) (3.3)
To show these, we need the following definition and lemmae.
Definition 3.2. A portfolio is a collection of investments.
For instance, a portfolio I = c − ∆S means that we long a call and short ∆ amount of an asset
S.
Remark. The value of a portfolio depends on time. Suppose a portfolio I has A amount of cash
at time t, its value is Aer(T −t) at time T . On the other hand, a portfolio involves E amount of cash,
its present value is Ee−r(T −t) . We should make such deduction, otherwise there is an arbitrage
opportunity.
Lemma 3.2. Suppose I(t) and J(t) are two portfolios containing no American options. Then under
the hypothesis of no-arbitrage-opportunities, we can conclude that
I(T ) ≤ J(T ) ⇒ I(t) ≤ J(t), ∀t ≤ T,
I(T ) = J(T ) ⇒ I(t) = J(t), ∀t ≤ T.
Proof. Suppose the conclusion is false, i.e., there exists a time t ≤ T such that I(t) > J(t). An
arbitrageur can buy (long) J(t) and short I(t) and immediately gain a profit I(t) − J(t). Its value
at time T is (I(t) − J(t))er(T −t) . Since I and J containing no American options, nothing can
be exercised before T . At time T , since I(T ) ≤ J(T ), he can use J(T ) (what he has) to cover
I(T ) (what he shorts) and gains another profit J(T ) − I(T ). This contradicts to the hypothesis of
no-arbitrage-opportunities.
The equality case can be proven similarly.
Proof of Theorem 3.1.

1. Let I = c and J = S. At T , we have
I(T ) = cT = max{ST − E, 0} ≤ max{ST , 0} = ST = J(T ).
Hence, I(t) ≤ J(t) holds for all t ≤ T .

Remark. The equality holds when E = 0. In this case c = S
2. Consider I = c and J = 0. At time T , I(T ) = Λ(T ) ≥ 0 = J(T ), hence I(t) ≥ 0 for all
t ≤ T . Similarly, we also get p ≥ 0.
3. Consider I = c + Ee−r(T −t) and J = S. At time T ,
I(T ) = max{ST − E, 0} + E = max{ST , E} ≥ ST = J(T ).
This implies I(t) ≥ J(t).
4. Let I = p and J = Ee−r(T −t) . At time T ,
I(T ) = max{E − ST , 0} ≤ E = J(T ).
Hence, p(t) ≤ Ee−r(T −t) . The equality holds only when ST = 0.
5. Consider I = p + S and J = Ee−r(T −t) . At time T ,
I(T ) = max(E − ST , 0) + ST = max{ST , E} ≥ E = J(T ).
Hence, I(t) ≥ J(t).
6. For put-call parity, we consider I = c + Ee−r(T −t) and J = p + S. At time T ,
I(T ) = c + E = max{ST − E, 0} + E = max{ST , E},

J(T ) = p + S = max{E − ST , 0} + ST = max{E, ST }
Hence, I(t) = J(t).
3.2.3 Basic properties of American options

Theorem 3.2. For American options, we have
(i) The optimal exercise time for American call option is T , and we have C = c.
(ii) The optimal exercise time for American put option is as earlier as possible, and we have
P (t) ≥ p(t).
(iii) The put-call parity for American option is
C + Ee−r(T −t) ≤ S + P ≤ C + E. (3.4)

To prove these properties, we need the following lemma.
Lemma 3.3. Let I or J be two portfolios that contain American options. Suppose I(τ ) ≤ J(τ ) at
some τ < T . Then I(t) ≤ J(t), for all t ≤ τ .
Proof. Suppose I(t) > J(t) at some t ≤ τ . An arbitrageur can long J(t) and short I(t) at time t to
make profit I(t) − J(t) immediately. At later time τ , he can use J(τ ) to cover I(τ ) with additional
profit J(τ ) − I(τ ), in case the person who owns I exercises his American option.
Remark. The equality also holds , that is, if I(τ ) = J(τ ) for τ < T , then I(t) = J(t) for t ≤ τ .
Proof of Theorem 3.2.
1. Firstly, we show Ct ≥ ct for all t < T . If not, then c(τ ) > C(τ ) for some time τ < T , we
can buy C and sell c at time τ to make a profit c(τ ) − C(τ ). The right of C is even more than
that of c. This is an arbitrage opportunity which is a contradiction.
Secondly, we show ct ≥ Ct . Consider two portfolios I(t) = Ct + Ee−r(T −t) and J(t) = St .
The portfolio I is an American call plus an amount of cash Ee−r(T −t) . Suppose we exercise
C at some time τ < T , then I(τ ) = Sτ − E + Ee−r(T −τ ) and J(τ ) = Sτ . This implies
I(τ ) < J(τ ). By our lemma, I(t) ≤ J(t) for all t ≤ τ . Since τ ≤ T arbitrary, we conclude
I(t) ≤ J(t) for all t < T . Combine this inequality with the inequality
ct + Ee−r(T −t) ≥ St
in section 3.2,we conclude ct = Ct for all t < T . Further, early exercise results in C(τ ) +
Ee−r(T −τ ) = S(τ ) − E + Ee−r(T −τ ) < S(τ ). But S(τ ) ≤ cτ + Ee−r(T −τ ) . Thus, early
exercise leads to decrease the value of C(τ ), if τ < T . Hence, the optimal exercise time for
American option is T .
2. We show p(t) ≤ P (t). Suppose p(t) > P (t). Then we can make an immediate profit by
selling p and buying P . We earn p − P and gain more right. This is a contradiction.
Next, we show that we should exercise American puts as early as possible (the first time
E > Sτ ). We consider a portfolio I(t) = Pt + St . This is called a covered put. That is,
we use S to cover P when we exercise P . When we exercise P at τ < T , the payoff of I
is E − Sτ + Sτ = E. By the time T , its value is Eer(T −τ ) . Thus, in order to maximize the
profit, we exercise time should be as earlier as possible.
3. The inequality C + Ee−r(T −t) ≤ S + P follows from the put-call parity (3.3) and the facts
that c = C and P ≥ p. To show the inequality S + P < C + E, we consider two portfolios:
I(t) = c(t) + E and J(t) = P (t) + St . Suppose P is not exercised before T . Then
J(T ) = max(E − ST , 0) + ST = max(E, ST ), and
I(T ) = c(T ) + Eer(T −t) = max(E, ST ) + E(er(T −t) − 1) > J(T ).

Suppose P is exercised at some time τ , with τ < T . Then we sell Sτ at price E. Thus
J(τ ) = E. On the other hand,
I(τ ) = c(τ ) + Eer(τ −t) ≥ E = J(τ ).
From lemma, we have I(t) > J(t). That is, c(t) + E ≥ P (t) + St . Since c(t) = C(t), we
also have C(t) + E ≥ P (t) + St .
Remark. P − p is called the time value of a put. The maximal time value is E − Ee−r(T −t) .
Examples.
1. Suppose S(t) = 31, E = 30, r = 10%, T − t = 0.25 year, c = 3, p = 2.25. Consider two
portfolios:
I = c + Ee−r(T −t) = 3 + 30 × e−0.1×0.25 = 32.26,

J = p + S = 2.25 + 31 = 33.25.
We find J(t) > I(t). It violates the put-call parity. This means that there is an arbitrage
opportunity.
Strategy : long the security in portfolio I and short the security in portfolio J. This results a
cashflow: −3 + 2.25 + 31 = 30.25. Put this cash into a bank. We will get 30.25 × e0.1×0.25 =
31.02 at time T . Suppose at time T , ST > E, we can exercise c, also we should buy a
share for E to close our short position of the stock. Suppose ST < E, the put option will be
exercised. This means that we need to buy the share for E to close our short position. In both
cases, we need to buy a share for E to close the short position. Thus, the net profit is
31.02 − 30 = 1.02.
2. Consider the same situation but c = 3 and p = 1. In this case
I = c + Ee−r(T −t) = 32.25

J = p + S = 1 + 31 = 32.
and we see that J is cheaper.

Strategy: We long J and short I. To long J, we need an initial investment 31 + 1, to short c,
we gain 3. Thus, the net investment is 31 + 1 − 3 = 29 initially. We can finance it from the
bank, and we need to pay 29 × e0.1×0.25 = 29.73 to the bank at time T . Now, at T , we must
have that either c or p be exercised. If ST > E, then c is exercised. We need to sell the share
for E to close our short position for c. If ST < E, we exercise p. That is, we sell the share
for E. In both cases, we sell the share for E. Thus, the net profit is 30 − 29.73 = 0.27.
3.2.4 Dividend Case

Many stocks pay out dividends. These are payments to shareholders out of the profits made by the
company. Since the company’s wealth does not change after paying the dividends, the stock price,
the strike prices fall as the dividends being paid. If a company declared a cash dividend, the strike
price for options was reduced on the ex-dividend day by the amount of the dividend.
The inequalities below are all at the present time t. Thus, let D be the present value of the
dividend, no matter when they are paid. Certainly its value at time T is Der(T −t) . Thus, a company
pays dividend before T , its asset at time T becomes ST + Der(T −t) .
Theorem 3.3. Suppose a dividend will be paid during the life of an option. Let D denotes its present
value. Then we have for European option
S − D − Ee−r(T −t) ≤ c ≤ S (3.5)
− S + D + Ee−r(T −t) ≤ p ≤ Ee−r(T −t) . (3.6)

and the put-call parity:
c + Ee−r(T −t) = p + S − D. (3.7)
For the American options, we have (i)
S − D − E ≤ C − P ≤ S − Ee−r(T −t) (3.8)
provided the dividend is paid before exercising the put option, or (ii)
S − E ≤ C − P ≤ S − Ee−r(T −t) (3.9)
if the put is exercised before the dividend being paid.
Proof. 1. Proof of c ≥ S − Ee−r(T −t) − D.

We consider two portfolios:
I = c + D + Ee−r(T −t) ,
J = S.
Then at time T ,
I(T ) = max{ST − E, 0} + D + E = max{ST , E} + Der(T −t)

J(T ) = ST + Der(T −t) .
Hence I(T ) ≥ J(T ). This yields I(t) ≥ J(t) for all t ≤ T . This proves
c ≥ S − D − Ee−r(T −t) .
In other word, c is reduced by an amount D.

2. Proof of p ≥ D + Ee−r(T −t) − S. That is, p increases by an amount D.

We consider
I =p+S
J = Ee−r(T −t) + D.
At time T , I(T ) = p(T ) = max(E − ST , 0) + ST + Der(T −t) , while J(T ) = E + Der(T −t) .
Thus, I(T ) ≥ J(T ), we then obtain p + S ≥ Ee−r(T −t) + D.
3. For the put-call parity, we consider
I = c + D + Ee−r(T −t)
J = S + p.
At time T ,
I = J = max{ST , E} + Der(T −t) .
This yields the put-call parity for all time.
4. When there is no dividend, we have shown that
C − P ≤ S − Ee−r(T −t) .
When there is dividend payment, we know that
CD ≤ C, PD ≥ P
Hence,
CD − PD ≤ C − P ≤ S − Ee−r(T −t) .
Here, CD and PD refer to Amperican call and put with paying dividends. For the American
call option, we should not exercise it early, because the dividend will cause the stock price
to jump down, making the option less attractive. We should exercise it immediately after the
ex-dividend date.
5. For the American put option, we consider
I = C + D + E, J = P + S.
If we exercise P at τ ≤ T after the dividend being paid, then Sτ < E and
I(τ ) = Der(τ −t) + Eer(τ −t) ,
J(τ ) = E + Der(τ −t) .
We have J(τ ) ≤ I(τ ). Hence J(t) ≤ I(t) for all t ≤ τ .
6. If the put option is exercised before the dividend being paid, then we should consider I =
C + E and J = P + S. At τ ,
I(τ ) = Eer(τ −t) ,
J(τ ) = E.
Again, we have J(τ ) ≤ I(τ ). Hence J(t) ≤ I(t) for all t ≤ τ .
3.3. THE BLACK-SCHOLES EQUATION 33
3.3 The Black-Scholes Equation

3.3.1 Black-Scholes Equation
The fundamental hypothesis of the Black-Scholes analysis is that there is no arbitrage opportunities.
Besides, we make the following additional assumptions:
(1) The asset price follows the log-normal distribution. That is, the asset price S satisfies the
s.d.e.
dS = µSdt + σSdz.
(2) There exists a risk-free interest rate r.
(3) No transaction costs.
(4) No dividend paid.
(5) Shorting selling is permitted.
Our purpose is to value the price of an option (call or put). Let V (S, t) denotes for the price of an
option. The randomness of V (S(t), t) would be fully correlated to S(t). Thus, we consider a port-
folio which contains only S and V , but in opposite position in order to cancel out the randomness.
Then this portfolio becomes deterministic. To be more precise, let the portfolio be
Π = V − ∆S,
that is, long an option and short ∆ amount of the underlying asset. In a small time step dt, the
change of this portfolio is
dΠ = dV − ∆dS.
Here ∆ is held fixed during the time step. From Itô’s lemma
1
dΠ = Vt dt + VS dS + VSS σ 2 dt − ∆dS
2
∂V
= σS − ∆ dz
∂S
1 2 2 ∂2V

∂V ∂V
+ µS − µS∆ + + σ S dt.
∂S ∂t 2 ∂S 2
Now, we can eliminate the randomness by choosing
∂V
∆= (3.10)
∂S
at the starting time of each time step. The resulting portfolio
Π = V − VS S
has the property that its change

∂2V

∂V 1
dΠ = + σ2S 2 2 dt
∂t 2 ∂S
dΠ
is wholly deterministic. From the hypothesis of no arbitrage opportunities, the return, Π , should
be the same as investing Π in a riskless bank with interest rate r, i.e.
dΠ
= rdt.
Π
Otherwise, there would be either a net loss or an arbitrage opportunity. Hence we must have
1 2 2 ∂2V

∂V ∂V
rΠdt = µS − µS∆ + + σ S dt
∂S ∂t 2 ∂S 2
∂2V

∂V 1
= + σ 2 S 2 2 dt,
∂t 2 ∂S
or
1 2 2 ∂2V

∂V ∂V
+ σ S =r V −S . (3.11)
∂t 2 ∂S 2 ∂S
This is the Black-Scholes partial differential equation (P.D.E.) for option pricing.
σ2 2
• Its left-hand side Vt + 2 S VSS is the return from the hedged portfolio per unit time,
• while its right-hand side r(V − SVS ) is the return from bank deposit per unit time.
Note that the equation is independent of µ.
Remark. Notice that the Black-Scholes equation is invariant under the change of variable S 7→ λS.
3.3.2 Boundary and Final condition for European options

• Final condition:
c(S, T ) = max{S − E, 0}
p(S, T ) = max{E − S, 0}.
In general, the final condition is

V (S, T ) = Λ(S),
where Λ is the payoff function. These final conditions are derived from the “no-arbitrage
opportunities” assumption.
• Boundary conditions:
(i) On S = 0:
c(0, τ ) = 0, ∀t ≤ τ ≤ T.
This means that you wouldn’t want to buy a right whose underlying asset costs nothing.
3.4. EXACT SOLUTION FOR THE B-S EQUATION FOR EUROPEAN OPTIONS 35
(ii) On S = 0:
p(0, τ ) = Ee−r(T −τ ) .
This follows from the put-call parity and c(0, t) = 0.
(iii) For call option, at S = ∞:
c(S, t) ∼ S − Ee−r(T −t) , as S → ∞.
Since S → ∞, the call option must be exercised, and the price of the option must be
closed to S − Ee−r(T −t) .
(iv) For put option, at S = ∞:
p(S, t) → 0, as S → ∞
As S → ∞, the payoff function Λ = max{E − S, 0} is zero. Thus, the put option is
unlikely to be exercised. Hence p(S, T ) → 0 as S → ∞.
3.4 Exact solution for the B-S equation for European options
3.4.1 Reduction to parabolic equation with constant coefficients
Let us recall the Black-Scholes equation
∂V 1 ∂2V ∂V
+ σ 2 S 2 2 + rS − rV = 0, (0, T ] × (0, ∞). (3.12)
∂t 2 ∂S ∂S
This P.D.E. is a parabolic equation with variable coefficients. Notice that this equation is invariant
under S → λS. That is, it is homogeneous in S with degree 0. We therefore make the following
change-of-variable:
S
x = log .
E
The fraction S/E makes x dimensionless. The domain S ∈ (0, ∞) becomes x ∈ (−∞, ∞) and
∂V ∂S ∂V ∂V
= =S ,
∂x ∂x
∂S ∂S
∂2V

∂ ∂V
= S
∂x2 ∂x ∂S
∂S ∂V ∂S ∂ 2 V
= +S
∂x ∂S ∂x ∂S 2
∂V ∂2V
= S + S2 2
∂S ∂S
∂V ∂ 2V
= + S2 2 .
∂x ∂S
Next, let us reverse the time by letting
τ = T − t.
Then the Black-Scholes equation becomes
1 ∂2V

∂V 1 ∂V
= σ2 2 + r − σ2 − rV.
∂τ 2 ∂x 2 ∂x
We can also make V dimensionless by setting v = V /E. Then v satisfies
1 ∂2v

∂v 1 ∂v
= σ2 2 + r − σ2 − rv. (3.13)
∂τ 2 ∂x 2 ∂x
The initial and boundary conditions for v becomes
v(x, 0) = Λ̄(x) = Λ(Eex )/E,
• call option: Λ̄(x) = max{ex − 1, 0}, v(−∞, τ ) = 0, v(x, τ ) → ex − e−rτ as x → ∞;
• put option: Λ̄(x) = max{1 − ex , 0}, v(−∞, τ ) = e−rτ , v(x, τ ) → 0 as x → ∞.
Our goal is to solve v for 0 ≤ τ ≤ T .
3.4.2 Further reduction

Let us divide the v equation by σ 2 /2:
∂2v

2 ∂v 2r ∂v 2r
2
= + −1 − v.
σ ∂τ ∂x2 σ2 ∂x σ 2
Let us make a change of variable and define the following new constants:
1
s = τ /( σ 2 )
2
1
a = 1 − r/( σ 2 )
2
1 2
b = r/( σ )
2
Then the equation becomes
vs + avx + bv = vxx . (3.14)
The part, vs + avx is call the advection term. The term bv is called the source term, and the term
vxx is called the diffusion term. The advection part:
vτ + avx = (∂τ + a∂x ) v
is a direction derivative along the curve (called characteristic curve)

dx
= a.
dτ
This suggests the following change-of-variable:
y = x − as
s0 = s.
Then the direction derivative become
∂s0 = ∂s + a∂x
∂y = ∂x
Hence the equation is reduced to

vs0 + bv = vyy .
Since s0 = s, we will still use s below instead of s0 . Therefore, the equation becomes
vs + bv = vyy .
Next, the equation vs + bv suggests that v behaves like ebs along the characteristic curves. Thus, it
is natural to make the following change-of-variable
v = e−bs u.
Then
∂s u = ∂s (ebs v) = ebs ∂s v + bebs v = ebs (∂s + b)v = ebs ∂y2 v = ∂y2 u.
Thus, the equation is reduced to
us = uyy , y ∈ (−∞, ∞), s > 0.
This is the standard heat equation on the real line. Its solution can be expressed as
Z ∞
1 (y−z)2
u(y, s) = √ e− 4s f (z) dz, s > 0,
−∞ 4πs
where f is the initial data.
3.4.3 Black-Scholes formula

Let us return to the Black-Scholes equation (5.6). Thus, the solution v
Z ∞ 1 σ 2 )τ )2
1 (x−z+(r− 2
v(x, τ ) = e −rτ
√ e− 2σ 2 τ Λ̄(z) dz (3.15)
−∞ 2πσ 2 τ
In terms of the original variables, with the change of variable S 0 = Eez , we have the following
Black-Scholes formula:
Z ∞ 1 σ 2 )(T −t))2
(ln( S0 )+(r− 2
−r(T −t) 1 − S
V (S, t) = e p e 2
2σ (T −t) Λ(S 0 ) dS 0 (3.16)
2
2πσ (T − t)S 0
0
We may express it as
Z ∞
−r(T −t)
V (S, t) = e P(S 0 , T, S, t)Λ(S 0 ) dS 0 . (3.17)
0
Here,
0
!
1 [ln( SS )−(r − 21 σ 2 )(T − t)]2
P(S 0 , T, S, t) := p exp − . (3.18)
2πσ 2 (T − t)S 0 2σ 2 (T − t)
This is the transition probability density of an asset price model with growth rate r and volatility
σ. In other words, V is the present value (deducted by e−r(T −t) ) of the expectation of the payoff Λ
under an asset price model with volatility σ and growth rate r. It is important to note that it depends
on r, not on the growth rate of the underlying asset. We shall come back to this point later.
3.4.4 Special cases

1. European call option. The rescaled payoff function for a European call option is
Λ̄(z) = max{ez − 1, 0}.
Then Z ∞
−rτ 1 1 2 −r)τ )2 /(2σ 2 τ )
v(x, τ ) = e √ e−(x−z−( 2 σ (ez − 1) dz. (3.19)
0 2πσ 2 τ
This can be integrated. Finally, we get the exact solution for the European call option
c (S, t) = SN (d1 ) − Ee−r(T −t) N (d2 ), (3.20)
where
Z y
1 z2
N (y) = √ e− 2 dz, (3.21)
2π −∞
S
log( E ) + (r + 12 σ 2 )(T − t)
d1 = √ , (3.22)
σ T −t
S
log( E ) + (r − 12 σ 2 )(T − t)
d2 = √ . (3.23)
σ T −t
The calculation is done as below. Let us use the following notations for abbreviation.
√ √
a = r − σ 2 /2, D = σ 2 τ, d1 D = x + aτ + D, d2 D = x + aτ
We have
Z ∞
1 1
u(x, τ ) = √ exp − (z − x − aτ ) (ez − 1) dz = I − II.
2
0 2πD 2D
Z ∞
1 1
I= √ exp − (z − x − aτ ) ez dz
2
0 2πD 2D
Z ∞
D2 + 2D(x + aτ )

1 1 2
= √ exp − (z − (x + aτ + D)) + dz
0 2πD 2D 2D
Z ∞ √ √

1 1
= √ exp − (z − d1 D)2 + (d1 + d2 ) D/2 dz
0 2πD 2D
√ Z ∞ √ 2

(d1 +d2 ) D/2 1 1
=e √ exp − (z − d1 D) dz
0 2πD 2D
Z ∞ √ √

S 1 1
= erτ √ exp − (z/ D − d1 )2 dz/ D
E 2π 2
Z0 ∞
S 1 −z 2 /2 S
= erτ e dz = erτ N (d1 ).
E −d1 2π E
Z ∞
1 1
II = √ exp − (z − x − aτ )2 dz
2πD 2D
Z0 ∞
√ 2

1 1
= √ exp − (z − d2 D) dz
0 2πD 2D
Z ∞ √ √

1 1
= √ exp − (z/ D − d2 )2 dz/ D
2π 2
Z0 ∞
1 2
= √ e−z /2 dz = N (d2 ).
−d2 2π
V (S, t) = Ee−r(T −t) (I − II) = SN (d1 ) − Ee−r(T −t) N (d2 ).
Remarks
• We have seen that a lower bound of c is c ≥ S − Ee−r(T −t) .

• The formula is a modification of the previous lower bound formula.
√ The function N is
the error function. N (+∞) = 1. The two parameters d1 − d2 = σ T − t, the standard
deviation of the asset price model.
Exercise. Show that
(a) N (0) = 1/2, N (x) → 1 as x → ∞;

(b) N (x) + N (−x) = 1 for any x ∈ R
Exercise. Prove the formula (3.20).
2. European put option. Recall the put-call parity
c + Ee−r(T −t) = p + S.
We can obtain the price for p from c:
p(S, t) = Ee−r(T −t) N (−d2 ) − SN (−d1 ). (3.24)
Exercise. Prove (3.24).
3. Forward contract Recall that a forward contract is an agreement between two parties to buy
or sell an asset at certain time in the future for certain price. Its value V also satisfies the B-S
equation. The payoff function for such a forward contract is
Λ(S) = S − E.
In terms of the rescaled variable, Λ̄(z) := Λ(Eez )/E = ez −1. Let x = ln(S/E), τ = T −t)
and u(x, τ ) = erτ V (Eex , t)/E. Then u satisfies the advect heat equation. Its solution with
initial data Λ̄(x). Its solution is given by
Z ∞
1 (x−z−(r−σ 2 /2)τ )2
u(x, τ ) = √ e− 2σ 2 τ (ez − 1) dz
2πσ 2 τ −∞
= ex+rτ − 1.
Hence,
V (S, t) = S − Ee−r(T −t) . (3.25)
This means that the current value of a forward contract is nothing but the difference of S and
the discounted E. Notice that this value is independent of the volatility σ of the underlying
asset.
Exercise. Show that the payoff function of a portfolio c − p is S − E. From this and the
Black-Scholes formula (3.16), show the formula of the put-call parity.
4. Cash-or-nothing. A contact with cash-or-nothing is just like a bet. If ST > E, then the
reward is B. Otherwise, you get nothing. The payoff function is

B if S > E
Λ(S) =
0 otherwise.
Using the Black-Scholes formula (3.16), we obtain the value of a cash-or-nothing contract to
be
V (S, t) = Be−r(T −t) N (d2 ). (3.26)
Exercise. Verify this formula.
5. Supershare. Supershare is a binary option whose payoff function is defined to be

B if E1 < S < E2
Λ(S) =
0 otherwise.
3.5. RISK NEUTRALITY 41
One can show that the value for this binary option is
V (S, t) = Be−r(T −t) (N (d2 (E1 )) − N (d2 (E2 ))) ,
where d2 (E) is given by (3.23).

Exercise. Verify this formula.
6. Deterministic case (σ = 0). In this case, the Black-Scholes equation is reduced to
Vt + rSVS − rV = 0.
Or in τ, x and u variables:
uτ − rux = 0
with initial data
u(x, 0) = Λ(Eex )/E,
Thus, its solution is given by
u(x, τ ) = u(x + rτ, 0) = Λ(Eex+rτ )/E = Λ(Serτ )/E.
Or
V (S, t) = e−r(T −t) Λ(Ser(T −t) ).
This means that when the process is deterministic, the value of the option is the payoff func-
tion evaluated at the future price of S at T (that is Ser(T −t) ), and then discounted by the factor
e−r(T −t) .
3.5 Risk Neutrality

Notice that the growth rate µ of the underlying asset does not appear in the Black-Scholes equation.
The option may be valued as if all random walks involved are risk neutral. This means that the drift
term (growth rate) µ in the asset pricing model can be replaced by r. The option is then valued
by calculating the present value of its expected return at expiry. Recall the lognormal probability
density function with growth rate r, volatility σ is
S0
!
1 2 2
1 [ln( ) − (r − σ )(T − t)]
P(S 0 , T, S, t) := p exp − S 2
. (3.27)
2πσ 2 (T − t)S 0 2σ 2 (T − t)
This is the transition probability density of an asset price model in a risk-neutral world:
dS
= rdt + σdz. (3.28)
S
The expected return at time T in this risk-neutral world is
Z
P(S 0 , T, S, t)Λ(S 0 )dS 0 .
At time t, this value should be discounted by e−r(T −T ) :

Z
−r(T −t)
V (S, t) = e P(S 0 , T, S, t)Λ(S 0 )dS 0 .
We may reinvestigate the function N and the parameters di in the Black-Scholes formula. After
some calculation, we find Z ∞
N (d2 ) = P(S 0 , T, S, t)dS 0 . (3.29)
E
This is the probability of the event {S̃ ≥ E}, where S̃ obeys the risk-neutral pricing model:
dS̃
= rdt + σdz.
S̃
Similarly, one can show that
R∞
E P(S 0 , T, S, t)S 0 dS 0
N (d1 ) = . (3.30)
Ser(T −t)
is the expectation of S̃ at T when S̃ = 1 at t and under the condition that S̃ ≥ E at T .
Exercise. Check formulae (3.29) and (3.30).
3.6 Hedging
Hedging is the reduction of sensitivity of a portfolio to the movement of the underlying of asset
by taking opposite position in different financial instruments. The Black-Scholes analysis is a dy-
namical hedging strategy. The delta hedge is instantaneously risk free. It requires a continuous
rebalancing of the portfolio and the ratio of the holdings in the asset and the derivative product. The
delta (∆) for a whole portfolio is ∆ = ∂Π∂S . This is the sensitivity of Π against the change of S. By
taking dΠ − ∆ · dS = 0, the sensitivity of the portfolio to the asset price change is instantaneously
zero.
Besides the delta hedge, there are more sophisticated hedging strategies such as:
∂2Π
Gamma: Γ = ,
∂S 2
∂Π
Theta: Θ = ,
∂t
∂Π
Vega: = ,
∂σ
∂Π
rho: ρ = .
∂r
Hedging against any of these dependencies requires the use of another option as well as the asset
itself. With a suitable balance of the underlying asset and other derivatives, hedgers can eliminate
the short-term dependence of the portfolio on the movement in t, S, σ, r.
3.6. HEDGING 43
3.6.1 The delta hedging

Suppose a financial institution write 1, 000 European calls on 10,000 shares of a stock. How does
this institution hedge the risk from the price fluctuation of the underlying stock? The Delta-hedge
is such a strategy. It is a dynamic hedging strategy. The institution long ∆ shares of the underlying
stock. The ∆ is chosen to be
∂c
∆= .
∂S
If ∆ = 0.5, this means that it should long 0.5 × 10, 000 = 5, 000 shares. In this case, the total
portfolio Π = c − ∆S is instantaneously neutral. That is dΠ = dc − ∆dS = 0 under a small change
of dS. We call Π is delta neutral. However, this delta neutral holds only for a short period of time.
It should be adjusted periodically. This is called a dynamic hedging.
For the Delta-hedge for the European call and put options, we have the following propositions.
Proposition 1. For European call options, its ∆ hedge is given by
∆ = N (d1 ).
Proof. By definition, ∆ = ∂c/∂S. Since c = SN (d1 ) − Ee−r(T −t) N (d2 ), we get
∂c
= N (d1 ) + S · N 0 (d1 ) · d1S − Ee−r(T −t) N 0 (d2 )d2S .
∂S
Since
S 2 2
log( E ) + (r + σ2 )(T − t) S
log( E ) + (r − σ2 )(T − t)
d1 = √ , d2 = √ ,
σ T −t σ T −t
we get
1 1 2
d1S = d2S = p , N 0 (di ) = √ e−di /2 .
Sσ (T − t) 2π
Hence,
∂c √
= N (d1 ) + SN 0 (d1 ) − Ee−r(T −t) N 0 (d2 ) /(Sσ T − t)
∂S √
≡ N (d1 ) + I/(Sσ T − t).
We claim that I = 0. Or equivalently,
S N 0 (d1 )
= e−rτ
E N 0 (d2 )
This follows from the computation below.
S N 0 (d1 ) 2 2
0
= ex · e−(d1 −d2 )/2 .
E N (d2 )
From (3.22), (3.23),

1 σ 2 σ 2
d21 − d22 = (x + rτ + τ ) − (x + rτ − τ)
σ2τ 2 2
= 2(x + rτ )
Hence,
S N 0 (d1 )
= ex · e−x−rτ = e−rτ .
E N 0 (d2 )
Proposition 2. For European put options, its ∆ hedge is given by
∆ = N (−d1 ).
Proof. From the put-call parity,

∂p ∂c
∆= = − 1 = N (d1 ) − 1 = −N (−d1 ),
∂S ∂S
3.7 Time-Dependent r, σ and µ for Black-Sholes equation

We can extend Black-Schole analysis to the case when r, σ, µ are functions of t, but still determin-
istic. We use the change-of-variables:
S = Eex , V = Ev, τ = T − t.
The Black-Scholes equation is converted to
σ 2 (τ ) σ 2 (τ )
vτ = vxx + (r(τ ) − )vx − r(τ )v (3.31)
2 2
We look for a new time variable τ̂ such that
dτ̂ = σ 2 (τ )dτ
For instance, we can choose Z τ

τ̂ = σ 2 (τ ) dτ.
0
Then the equation becomes
1
vτ̂ = vxx + a(τ̂ )vx − b(τ̂ )v. (3.32)
2
To handle the term a(τ̂ ), we consider the following characteristic equation:
dx
= −a(τ̂ )
dτ̂
3.8. TRADING STRATEGY INVOLVING OPTIONS 45
This can be integrated and yields

Z τ̂
x=− a(τ 0 )dτ 0 + y,
0
where y is an integration constant. The lines with constant y’s are called characteristic curves.
Along the characteristic curves, the derivative ∂τ̂ + a(τ̂ )∂x is just the derivative in time with fixed
y. In fact, we introduce the following change-of-variable:
R τ̂
y = x + 0 a(τ 0 )dτ 0 ≡ x + A(τ̂ )

x y
→ with
τ̂ τ̂1 τ̂1 = τ̂
Then,
∂ ∂y ∂ ∂
|τ̂ = |τ̂ = ,
∂x ∂x ∂y ∂y
and
∂ ∂ τ̂ ∂ ∂x ∂ ∂ ∂
|y = + = − a(τ̂ ) .
∂ τ̂1 ∂ τ̂1 ∂ τ̂ ∂ τ̂1 ∂x ∂ τ̂ ∂x
The equation (3.32)is transformed to
1
vτ̂1 = vyy − b(τ̂1 )v.
2
R τ̂1
Let B(τ̂1 ) = 0 b(τ 0 )dτ 0 and u = eB(τ̂1 ) v, then uτ̂1 = 12 uyy . And we can solve this heat equation
explicitly.
3.8 Trading strategy involving options

An investor can design his/her payoff by using options and underlying stocks. The options whose
payoff are max{ST − E, 0} or max{E − ST , 0} are called vanilla option. They are the simplest
payoff functions. In this section, we shall discuss how to design more general payoff functions to
fit an investor’s anticipation.
3.8.1 Strategies involving a single option and stock

There are four ways:
a. Π = S − c (writing a covered call option). When an investor short a call, we say he write
a nake call. The investor has cash inflow at the beginning. However, short selling is highly
risky. If the stock rises ST > E, then the investor needs to buy the share on ST and sell to the
call holder on E. The investor loses ST − E. This loss can be unlimited because ST may rise
very high. To limit this risk, the investor can long a share St at time t. That is, his portfolio
is Π = −c + S. At expiry, if ST > E, then he can use this share to sell to the call owner on
E, instead of ST . His risk is limited. We say that this share covers the call, and say the writer
writes a covered call option. The payoff of Π is Λ = ST − max(ST − E, 0) = min{ST , E}.
b. Π = c − S (reverse of a covered call). Suppose an investor anticipates the stock price will
decrease. So he shorts a share and receive money St at time t. If ST does decrease, he earns
money. If unfortunately, the stock price increases, then he loses ST − St at time T . This risk
is unlimited. Thus, short selling has unlimited risk. However, he can buy a call to limit this
risk. That is, his portfolio Π = −S + c. Since c < S, he still receive some money from Π at
beginning. At expiry, if the price goes beyond E, then he can exercise this call to cover the
shorted share. The payoff now is Λ = − min{S, E} ≥ −E. Thus, his loss is at most E.
c. Π = p + S (protective put). When an investor anticipates a stock will increase, he set up a

portfolio Π = S. This costs him St . If the prices ST goes up, he earns ST − St . However,
if the price goes down, he sells it and loses St − ST . If he wants to limit his loss, he can set
up a portfolio Π = S + p. This costs him more at beginning. But at expiry, if the price goes
down, he can sell the share on E instead of a lower price ST . Thus, he limits his loss. In this
portfolio, the payoff is Λ = S + max{E − S, 0} = max{S, E}. The downside is limited (by
E) and the upside is unlimited (ST can be arbitrary large).
d. Π = −p − S (reverse of a protective put). In this portfolio, the investor anticipates the stock
price will decrease. He shorts a share St and also a put. He receives money from shorting
these. At expiry, the payoff is − max{S, E}. This means that he pays E if ST ≤ E. However,
if ST is large, then his risk is unlimited.
Below are the payoff functions for the above four cases.
3.8.2 Bull spreads

In this strategy, an investor anticipates the stock price will increase. However, he would like to give
up some of his right if the price goes beyond certain price, say E2 . Indeed, he does not anticipate
the stock price will increase beyond E2 . Therefore he does not want to own a right beyond E2 .
Such a portfolio can be designed as
Π = CE1 − CE2 , E1 < E2 ,
where CEi is a European call option with exercise price Ei and CE1 , CE2 have the same expiry.
The payoff
Λ = max{ST − E1 , 0} − max{ST − E2 , 0}

 0 if ST < E1
= ST − E1 if E1 < ST < E2
E2 − E1 if ST > E2

Since E1 < E2 , we have CE1 > CE2 . A bull spread, when created from CE1 − CE2 , requires an
initial investment. We can describe the strategy by saying that the investor has a call option with a
strike price E1 and has chosen to give up some upside potential by selling a call option with strike
price E2 > E1 . In return, the investor gets E2 − E1 if the price goes up beyond E2 .
E S
Λ
Λ
E S
(a) (b)
E S
E S
(c) (d)
Example: CE1 = 3, CE2 = 1 and E1 = 30, E2 = 35. The cost of the strategy is 2. The payoff is

 0 if ST ≤ 30
ST − 30 if 30 < ST < 35
5 if ST ≥ 35

The bull spread can also be created by using put options

Π = PE1 − PE2 , E1 < E2 .
3.8.3 Bear spreads

An investor entering into a bull spread is hoping that the stock price will increase. By contrast, an
investor entering into a bear spread is expecting the stock price will go down. The bear spread is
Π = CE2 − CE1 , E1 < E2 .
There is cash flow entered (CE2 − CE1 ). The payoff is
Λ = − max{ST − E1 , 0} + max{ST − E2 , 0}

 0 if ST < E1
= −ST + E1 if E1 < ST < E2
−E2 + E1 if ST > E2

E2−E1
E1 E2 S
3.8.4 Butterfly spread

If an investor anticipate the stock price will stay in certain region, say, E1 < ST < E3 , he or she
can have a butterfly spread such that the payoff function is positive in that region and he or she gives
up the return outside that region.
1. Butterfly spread using calls: Define the portfolio:
Π = CE1 − 2CE2 + CE3 , with E1 < E2 < E3 .
where E3 = E2 + (E2 − E1 ). Its payoff function is a piecewise linear function and is
determined by Λ(E1 ) = Λ(E3 ) = 0, Λ(E2 ) = E2 − E1 . Below is the graph of its payoff
function.
E2−E1
S−E E2−S
1
E E2 E3 S
1
Example: Suppose a certain stock is currently worth 61. A investor who feels that it is
unlikely that there will be significant price move in the next 6 month. Suppose the market of
6 month calls are
E C
55 10
60 7
65 5
55 60 65 S
The investor creates a butterfly spread by
Π = CE1 − 2CE2 + CE3 .
The cost is 10 + 5 − 2 × 7 = 1. The payoff is the figure below (Figure (1)).
2. Butterfly spread using puts.

E1 + E3
PE1 + PE3 − 2PE2 , E1 < E3 , E2 = .
2

 linear
ϕE2 = ∆E if S = E2
0 if S < E2 − ∆E, or S > E2 − ∆E

Remark 1. Suppose European options were available for every possible strike price E, then any
payoff function could be created theoretically:
X
Λ(S) = Λi ϕ(S − Ei )
where Ei = i∆E, Λi is constant and


 1 if S − Ei = 0
ϕ(S − Ei ) = 0 if |S − Ei | ≥ ∆E
linear for |S − Ei | ≤ ∆E.

Then Λ(Ei ) = Λi and Λ is linear on every interval (Ei , Ei+1 ) and Λ is continuous. As ∆E → 0,
we can approximate any payoff function by using butterfly spreads.
Remark 2. One can also use cash-or-nothing to create any payoff function:
X
Λ(S) = Λi ψ(S − Ei ),
where
1 if 0 ≤ S < ∆E
ψ(S) := H(S) − H(S − ∆E) =
0 otherwise.
and H is called the Heaviside function. The value for such a portfolio is
Z
−r(T −t)
V = e P(S 0 , T, S, t)Λ(S 0 )dS 0 ,
= e−r(T −t) ΣΛi P(Ei ≤ S ≤ Ei+1 ).
3.9 Derivation of heat equation and its exact solution

3.9.1 Derivation of the heat equation on R
The heat equation on R reads

ut = uxx , x ∈ R, t > 0,
u(x, 0) = f (x).
This equation models heat propagation on the line. Here, u(·, t) represents the temperature distribu-
tion at time t and f the initial temperature distribution. The derivation of this equation is based on
a physical law – the conservation of energy, and the Fourier law of heat flux. Given a temperature
distribution u(x, t), the Fourier law describes that the temperature flows from high temperature to
low temperature at a rate q called heat flux. It is the energy passing through a point (per unit area)
per unit time. The Fourier law states that
q = −κux .
Here, the constant κ is called thermal conductivity. It is a positive constant. It differs for different
material. For instant, κ of cooper is much larger than that of woods. According to thermal dynamics,
the energy density e is related to temperature by e = cv u. The constant cv is called specific heat (at
constant volume). The conservation of energy states that the change of energy in an interval (a, b)
per unit time is the same as the heat flux flows into (a, b) from boundaries. Mathematically, it is
Z b Z b
∂t e dx = q(a) − q(b) = − qx dx.
a a
This gives
Z b Z b
cv ∂t u dx = (κux )x dx.
a a
This holds for any interval (a, b). Thus, the integrands much be the same. This gives
∂t u = Duxx .
where D = κ/cv > 0 is the heat conductivity.

3.9. DERIVATION OF HEAT EQUATION AND ITS EXACT SOLUTION 51
3.9.2 Exact solution of heat equation on R.

If we rescale time by t0 = Dt, then the heat equation becomes
∂t0 u = uxx .
Thus, without loss of generality, we consider the heat equation with D = 1. We shall show that its
solution is given by
Z ∞
1 −x2 /(4t)
u(x, t) = G(x − y, t)f (y) dy, G(x, t) := √ e .
−∞ 4πt
The function G(·, t) is the Gaussian distribution. It corresponds to the temperature distribution
at time t when the initial temperature distribution is a hot spot at x = 0. The derivation of this
formulation is decomposed into the following steps.
1. Superposition principle: If u and v are solutions, so is au + bv, where a and b are constants.
This is called superposition principle for linear partial differential equations.
Now, we imagine f can be approximated by piecewise step function: let h be a small mesh
size, xj = jh and f can be approximated by fh defined by
∞
X
f ≈ fh , fh (x) := f (xj )δh (x − xj )h.
j=−∞
Here,
1 1 for − h/2 < x < h/2
δh (x) := χh (x), χh (x) :=
h 0 otherwise.
Suppose each solution corresponding to f (xj )δh (x) is f (xj )Gh (x − xj ). Because the equa-
tion is linear, then the general solution is just the linear superposition of these solutions:
X Z ∞
u(x, t) ≈ Gh (x − xj , t)f (xj ) · h ≈ G(x − y, t)f (y) dy.
j −∞
Here, Gh (x−xj , t) is the solution of the heat equation corresponding to the initial data δh (x−
xj ). Notice that we have used another important of heat equation, namely, the translation
invariance: if u(x, t) is a solution of the heat equation with initial data u(x, 0), then u(x−a, t)
is a solution of heat equation with initial data u(x − a, 0). This is because the equation is
unchanged under the translation transformation x 7→ x − a. Now if Gh (x, t) is the solution
with initial data δh (x), then Gh (x − xj , t) is the solution corresponding to the initial data
δh (x − xj ).
2. Translational invariance. Because the equation is translationally invariant, we can solve the
equation with initial data to be χh (x)/h. This initial function tends so-called δ-function as
h → 0. Thus, what we need is to solve the case when the initial function is a δ function. A
δ-function is a generalized function having the property:

0 x 6= 0
δ(x) =
∞ x=0
Z ∞
δ(x) dx = 1.
−∞
The δ-function is considered to be the limit of δh (x) as h → 0+ in the sense that
Z ∞ Z ∞
1 h/2
Z
f (x)δ(x) dx = lim f (x)δh (x) dx = f (x) dx = f (0),
−∞ h→0 −∞ h −h/2
for any smooth function f .
3. Parabolic scaling. To solve the heat equation with an initial hot spot at x = 0, we use
the parabolic invariance property of the heat equation. Namely, the equation and the initial
condition are both invariant under the transformation
x 7→ λx, t 7→ λ2 t,
λ > 0.
√
This suggests that the√solution is a function of ξ = x/ t. So, let us look a solution of the
form u(x, t) = U (x/ t). Plug it into the equation, we obtain

1 x 1
0
ut = U (ξ) − 3/2 , uxx = U 00 (ξ) .
2t t
This leads to
ξ
− U 0 (ξ) = U 00 (ξ).
2
0 ξ
ln U 0 (ξ) = − .
2
Integrate it,
ξ2 2
ln U 0 (ξ) = − + Const., U 0 (ξ) = Ce−ξ /4 .
4
Hence Z ξ
2
U (ξ) = Ce−η /4 dη.
−∞
√
√ a solution has the property: U (−∞) = 0 and U (+∞) = C 4π. We choose C =
Such
1/ 4π for normalization.
√ However, this is still not what we want. But we notice that if
v(x, t) = U (x/ t) is a solution, so is vx . Thus, we let

x 1 x2 /(4t)
G(x, t) = ∂x U √ = √ e
t 4πt
Then G is a solution and satisfies
Z ∞
G(x, t) dx = 1.
−∞
3.9. DERIVATION OF HEAT EQUATION AND ITS EXACT SOLUTION 53
4. Verification. Finally, one check

R∞
• u(x, t) := −∞ G(x − y, t)f (y) dy is a solution of the heat equation
• u(x, t) → f (x) as t → 0+ for any bounded continuous function f .
To show the first one, we check that Gt = Gxx . This is left to you.
For the second one, the integral
Z ∞ Z ∞ Z ∞
G(x − y, t)f (y) dy = G(−y, t)f (x + y) dy = G(y, t)f (x + y) dy
−∞ −∞ −∞
represents the average of f about x with

√ weight G(y, t). We notice that G(y, t) is a Gaussian
distribution with standard deviation t. As t → 0, this weight is more√
and more concentrated
at 0. Therefore, this average becomes the average of f roughly in a t-neighborhood of x.
Its limit is f (x) for any bounded smooth function f .
We can also give a more rigorous proof. We want to show
Z ∞

lim |u(x, t) − f (x)| =
G(x − y, t)f (y) dy − f (x) = 0.
t→0 −∞
Notice that Z ∞
G(x − y, t) dy = 1 for all x ∈ R, t > 0.
−∞
Hence Z ∞ Z ∞
G(x − y)f (y) dy − f (x) = G(x − y) (f (y) − f (x)) dy.
−∞ −∞
Therefore, Z ∞
|u(x, t) − f (x)| ≤ G(x − y, t) |f (y) − f (x)| dy.
−∞
Here, we have used G(x, t) ≥ 0. To estimate this integral, we break it into two parts
!
Z ∞ Z Z
G(x−y, t) |f (y) − f (x)| dy = + G(x−y, t) |f (y) − f (x)| dy = I+II,
−∞ |x−y|<δ |x−y|≥δ
where δ > 0 is a small parameter to be chosen later. Since f is continuous, we have for any
> 0, there exists a δ > 0 such that
|f (y) − f (x)| < , whenever |x − y| < δ.
Thus,
Z Z
I= G(x − y, t)|f (y) − f (x)| dy ≤ G(x − y, t) dy < .
|x−y|<δ |x−y|<δ
For II, since we assume f is bounded, there exists a constant M > 0 such that |f (x)−f (y)| ≤
2M . Thus, Z Z
II ≤ 2M G(x − y, t) dy = 2M G(y, t) dy.
|x−y|≥δ |y|≥δ
This integral is the tail estimate of a Gaussian distribution:

Z Z Z
1 −y2 /(4t) 1 −z 2
G(y, t) dy = 2 √ e dy = √ √ e dz.
|y|≥δ y≥δ 4πt z≥δ/(2 t) π
As t → 0+, this integral goes to 0. Therefore, we can choose a small t1 > 0 such that
II ≤ for all 0 < t < t1 .
Combining I and II, we have for any > 0, there exists t1 > 0 such that for any 0 < t < t1
we have I + II < 2. This shows I + II → 0 as t → 0.
Chapter 4
Variations on Black-Scholes models
4.1 Options on dividend-paying assets

Dividends are payments to the shareholders out of the profits made by the company. We will con-
sider two “deterministic” models for dividend. One has constant dividend yield. The other has
discrete dividend payments.
4.1.1 Constant dividend yield

Suppose that in a short time dt, the underlying asset pays out a dividend D0 Sdt, where D0 is a
constant, called the dividend yield. This continuous dividend structure is a good model for index
options and for short-dated currency options. In the latter case, D0 = rf , the foreign interest rate.
As the dividend is paid, the return dSS must fall by the amount of the dividend payment D0 dt. It
follows the s.d.e. for the asset price is
dS
= (µ − D0 )dt + σdz.
S
To value the corresponding option V , we consider the portfolio : Π = V − ∆S as before. In one
time step, the change of the portfolio is
dΠ = dV − ∆dS − ∆D0 Sdt,
where the last term −∆D0 Sdt is the dividend our assets received.1 Thus
− ∆(dS +D0 Sdt)

dΠ = dV
∂V
= σS − ∆ dz
∂S
1 2 2 ∂2V

∂V
+ (µ − D0 )S + σ S + Vt − (µ − D0 )∆S − ∆D0 S dt
∂S 2 ∂S 2
1
If you own a share, after dt, you will receive SD0 dt dividend.
55
56 CHAPTER 4. VARIATIONS ON BLACK-SCHOLES MODELS
∂2V

∂V 1
= Vt − D0 S + σ2S 2 2 dt,
∂S 2 ∂S
∂V
Here, we have chosen ∆ = ∂S to eliminate the random term. From the absence of arbitrage
opportunities, we must have
dΠ = rΠdt.
Thus,
1 2 2 ∂2V

∂V ∂V
V t − D0 S + σ S =r V −S .
∂S 2 ∂S 2 ∂S
i.e.,
1 ∂2V ∂V
Vt + σ 2 S 2 2 + (r − D0 )S − rV = 0
2 ∂S ∂S
This is the Black-Scholes equation when there is a continuous dividend payment.
The boundary conditions are:
c(0, t) = 0,
c(S, t) ∼ Se−D0 (T −t)
The latter is the asset price S discounted by e−D0 (T −t) from the payment of the dividend. The
payoff function c(S, T ) = Λ(S) = max{S − E, 0}.
To find the solution, let us consider
c(S, t) = e−D0 (T −t) c1 (S, t).
Then c1 satisfies the original Black-Scholes equation with r replaced by r − D0 and the same final
condition. The boundary conditions for c1 are
c1 (0, t) = 0,
c1 (S, t) ∼ S as S → ∞
Hence,
c1 (S, t) = SN (d1,0 ) − Ee−(r−D0 )(T −t) N (d2,0 )
where
S
ln E + (r − D0 + 12 σ 2 )(T − t)
d1,0 = √ ,
σ T −t
√
d2,0 = d1,0 − σ T − t.
The original
c(S, t) = Se−D0 (T −t) N (d1,0 ) − Ee−r(T −t) N (d2,0 ).
This means that the call option on a dividend-paying asset is just like a call option in risk-free
environment with rate r − D0 and discounted by e−r(T −t) .
4.1. OPTIONS ON DIVIDEND-PAYING ASSETS 57
Remark. We notice that c & as D0 %. This is natural because paying dividend will lower the
value of S, hence c. If the dividend is paid more, the value of call becomes lower.
Exercise. Derive the put-call parity for the European options on dividend-paying assets.
4.1.2 Discrete dividend payments

Suppose our asset pays just one dividend during the life time of the option, say at time td . At td +,
the asset holder receiver a payment dy S(td −). Hence,
S(td +) = S(td −) − dy S(td −) = (1 − dy )S(td −).
We claim that across the jumps, V should be continuous, i.e.,
V (S(td −), td −) = V (S(td +), td +).
Otherwise, there is a net loss or gain from buying V before td then sell it right after td . To find
V (S, t), here is a procedure.
1. Solve the Black-Scholes from T to td + to obtain V (S, td +) (using the payoff function Λ)
2. Adjusting V by
V (S, td −) = V ((1 − dy )S, td +)
3. Solve Black-Scholes equation from td to t with the final condition V ((1 − dy )S, td +).
For the case of call option, we can find its explicit expression. Let cd be the European option for
this dividend-paying asset. Then
cd (S, t) = c(S, t, E) for td + ≤ t ≤ T
cd (S, td −) = cd (S(1 − dy ), td +) = c(S(1 − dy ), td , E)

Note that
c(S(1 − dy ), T, E) = max{S(1 − dy ) − E, 0} = (1 − dy ) max{S − (1 − dy )−1 E, 0}
and the linearity of the Black-Scholes equation, we obtain
c(S(1 − dy ), td , E) = (1 − dy )c(S, td , (1 − dy )−1 E).
Therefore,
cd (S, t) = (1 − dy )c(S, t, (1 − dy )−1 E).
That is, the call option price is reduced to its original value with strike price E replaced by E/(1 −
dy ).
4.2 Futures and futures options

4.2.1 Forward contracts
Recall that a forward contract is an agreement between two parties to buy or sell an underlying asset
on a certain price E at a certain future time T . Here, E is called the delivery price. The payoff
function for this forward contract is Λ = ST − E. Based on the no arbitrage opportunity, we have
seen that the price for this forward contract is (3.25)
f = S − Ee−r(T −t) .
Definition 4.3. The forward price F for a forward contract is defined to be the delivery price E
which would make that contract have zero value, i.e.,
ft = 0, or F = St er(T −t) .
One can take another point of view. Consider a party who shorts the contract. He can borrow an
amount of money St at time t to buy an asset and use it to close his short position at T . The money
he received at expiry, F , is used to pay the loan. If no arbitrage opportunities, then
F = St er(T −t) .
If the delivery price E of a forward contract is not equal to the forward price F , then the value of
this forward contract is f = St −Ee−r(T −t) . This is the result we obtained in the last chapter. Thus,
in order to have f = 0, we should take the delivery price to be St er(T −t) .
4.2.2 Futures
Futures are very similar to the forward contracts, except they are traded in an exchange, thus, they
are required to be standardized. This includes size, quality, price, expiry, ... etc. Let us explain the
characters of a future by the following example.
1. Trading future contracts
• Suppose you call your broker to buy one July corn futures contract (5,000 bushels) on
the Chicago Board of Trade (CBOT) at current market price.
• The broker send this signal to traders on the floor of the exchange.
• The trader signal this to ask other traders to sell, if no one wants to sell, the trader who
represents you will raise the price and eventually find someone to sell
• Confirmation: Price obtained are sent back to you.
2. Specification of the futures: In the above example, the specification of this future is
• Asset : quality
• Contract size: 5,000 bushel
4.2. FUTURES AND FUTURES OPTIONS 59
• Delivery arrangement: delivery month is on December

• price quotes
• Daily price movement limits: these are specified by the exchange.
• Position limits: the maximum number of contracts that a speculator may hold.
3. Operation of margins
• Marking to market: Suppose an investor who contacts his or her broker on June 1, 2016,
to buy two December 2016 gold futures contracts on New York Commodity Exchange.
We suppose that the current future price is $400 per ounce. The contract size is $100
ounces, the investor want to buy $200 ounces at this price. The broker will require
the investor to deposit funds in a “margin account”. The initial margin, say is $2,000
per contract. As the futures prices move everyday, the amount of money in the margin
account also changes. Suppose, for example, by the end of June 1, the futures price has
dropped from $400 to $397. The investor has a loss of $200×3=600. This balance in
the margin account would therefore be reduced by $600. Maintaining margin needs to
deposit. Certain account of money to keep that futures contract.
• Maintenance margin: To insure the balance in the margin account never becomes nega-
tive, a maintenance margin, which is usually lower than the initial margin, is set.
We tabulate the futures prices, the main/loss in each day. Suppose there are n days to the
maturity date. Let δ = r(T − t)/n be the riskless interest rate per date. Let Fi be the future
price at the end of day i. Suppose the margin has money M initially. The gain/loss in each
day is the following table.
Day 0 1 2 ··· n−1 n

futures price F0 F1 F2 ··· Fn−1 Fn = ST
Margin (compounded) M M eδ M e2δ ··· ··· M enδ
gain/loss 0 (F1 − F0 ) (F2 − F1 ) ··· ··· (Fn − Fn−1 )
gain/loss compounded
to day n 0 (F1 − F0 )e(n−1)δ (F2 − F1 )e(n−2)δ ··· ··· (Fn − Fn−1 )enδ
At the end of day i, the margin account should have

i
X
iδ
Me + (Fk − Fk−1 )e(i−k)δ .
k=1
This account is required to be kept an amount money greater than certain maintenance fee. If
it is lower than this maintenance fee, it will be closed.
Here, the futures prices Fi are the market future prices. What are the futures prices if there is no
arbitrage opportunities? We show below that under the no-arbitrage assumption, the future price at
time t is indeed identical to the forward price. That is Ft = St er(T −t) .
Theorem 4.4. Forward price and futures price are equal when the interest rates are constant.
Proof. Suppose a futures contract lasts for n days. Let the future prices are
F0 , · · · , Fn
at the end of each business day. Let δ be the risk-free interest rate per day. That is, δ = r(T − t)/n.
Consider the following two strategies:
• Strategy 1: Suppose the forward price is G0 .
– We invest G0 in a risk-free bond;

– take a long position of amount enδ forward contract.
At day n, G0 enδ is used to buy the underlying asset at price ST enδ . The payoff is ST enδ .
• Strategy 2: Suppose the futures price at day 0 is F0 .
– We invest F0 amount of money in a risk-free bond;

– take a long position of eiδ amount of future at the end of day i − 1, i = 1, ..., n.
Day 0 1 2 ··· n−1 n

futures price F0 F1 F2 ··· Fn−1 Fn
position eδ e2δ e3δ ··· enδ 0
δ 2δ
gain/loss 0 e (F1 − F0 ) e (F2 − F1 ) ··· ··· enδ (Fn − Fn−1 )
compound 0 eδ (F1 − F0 )e(n−1)δ e2δ (F2 − F1 )e(n−2)δ ··· ··· enδ (Fn − Fn−1 )
The total gain/loss from the long position of the futures is

n
X
(Fi − Fi−1 )eiδ · e(n−i)δ = (Fn − F0 )enδ = (ST − F0 )enδ .
i=1
At time T , we receive F0 enδ from the initial investment F0 . Thus, the total payoff of strategy 2 is
F0 enδ + (ST − F0 )enδ = ST enδ .
Since both strategies have the same payoff, we conclude their initial investments must be the same,
i.e., F0 = G0 = ST er(T −t) .
4.2.3 Futures options

Options on futures are traded in many different exchanges. They require the delivery of an underly-
ing futures contract when exercised. When a call futures option is exercised, the holder acquires a
long position in the underlying futures contract plus a cash amount equal to the current futures price
minus the exercise price. Suppose the expiry date of the call is T1 , which is T1 ≤ T , the expiration
of the underlying future. The payoff of the call option on future is thus
Λ = max(FT1 − E, 0).
In addition of this payoff, the holder also receive a future contract, yet it costs zero to enter the long
position of a future.
Example An investor who has a September futures call option on 25,000 pounds of copper
with exercise price E = 70 cents/pound. Suppose the current future price of copper for deliv-
ery in September is 80 cents/pound. If the option is exercised, the investor received 10 cents
×25, 000+long position in futures contract to buy 25,000 pound of copper in September at price
80 cents/pound.
The maturity date of futures option is generally on, or a few days before, the earlist delivery
date of the underlying futures contract.
Futures options are more attractive to investors than options on the underlying assets when it is
cheaper or more convenient to deliver futures contracts rather than the asset itself. Futures options
are usually more liquid and involved lower transaction costs.
4.2.4 Black-Scholes analysis on futures options

Future Price Model As we have seen that the futures price is identical to the forward price when
the interest rate is a constant, i.e., F = Ser(T −t) . From Itô’s lemma, we obtain a pricing model for
F:
1 ∂2F 2 2

∂F ∂F
dF = + 2
σ S dt + dS
∂t 2 ∂S ∂S
dS
= (−rer(T −t) S)dt + Ser(T −t)
S
= (−rF )dt + F (µdt + σdz)
= ((µ − r)F )dt + F σdz.
Hence,
dF
= (µ − r)dt + σdz. (4.1)
F
We conclude this discussion by the following proposition.
Proposition 3. The futures price is the same as a stock paying a dividend yield at rate r.
Next, we study the value V of a futures option. For a call option on a future, it is the right to
buy the future on strike price E. Its value V is a function of F , t and E. We apply Black-Sholes
analysis to value V .
Futures Option Price Consider a portfolio
Π = V − ∆F.
∂V
As discussed before, we choose ∆ = ∂F to eliminate randomness of dΠ. Then
dΠ = dV − ∆dF
∂V ∂V 1 ∂2V 2 2 ∂V ∂V
= ( µF F + + 2
σ F )dt + σF dz − (µF F dt + σF dz)
∂F ∂t 2 ∂F ∂F ∂F
∂V 1 ∂2V 2 2
= ( + σ F )dt.
∂t 2 ∂F 2
Here, µF := µ − r, is the growth rate of the undelying future. Since it costs nothing to enter into a
future contract, the cost of setting up the above portfolio is just V . 2 Thus, based on the no arbitrage
opportunity,
dΠ = rV dt,
and we obtain
1 ∂2V
Vt + σ 2 F 2 = rV.
2 ∂F 2
The payoff function for a call option is Λ = max{F − E, 0}.
To solve this equation, we recall the option price equation for stock paying dividend is
1 ∂2V ∂V
Vt + σ 2 S 2 2 + (r − D0 )S − rV = 0
2 ∂S ∂S
In our case, D0 = r, so the futures call option
c(F, t) = e−r(T −t) c1 (F, t)
where c1 satisfies Black-Scholes equation with r replaced by r − r = 0. This gives
c1 (F, t) = F N (d1 ) − EN (d2 ),

F
ln E − 21 σ 2 (T − t)
d2 = √ ,
σ √T − t
d1 = d2 + σ T − t
Notice that this option price V (i.e. c in the present case) is the same as Ṽ (S(F, t), t), where Ṽ is
the solution of the option corresponding to the underlying asset S. That is
Ṽ (S, t) = SN (d˜1 ) − Ee−r(T −t) N (d˜2 )

S
ln E + (r − 12 σ 2 )(T − t)
d˜2 = √ ,
√σ T − t
d˜1 = d˜2 + σ T − t
2
When the futures Ft = er(T −t) St , it is the same as the forward contract price. The price f of corresponding contract
is 0.
We can write this Ṽ in terms of F by S = F e−r(T −t) . Plug this into the above equation to obtain
−r(T −t)
ln F e + (r − 12 σ 2 )(T − t)
d˜2 = E √
σ T −t
F
ln E − 12 σ 2 (T − t)
= √ = d2 ,
σ T −t
Similarly, d˜1 = d1 . Thus
Ṽ (S(F, t), t) = SN (d1 ) − Ee−r(T −t) N (d2 )

= F e−r(T −t) N (d1 ) − Ee−r(T −t) N (d2 )
= V (F, t)
We conclude the above discussion by the following proposition.
Proposition 4. The price for future options is the same as the price for options on the underlying
assets.
Finally, let us find the put-call parity for futures options.
Proposition 5. Let c and p be the call and put options on a future. Then
c + Ee−r(T −t) = p + F e−r(T −t) . (4.2)
Proof. Consider two portfolios:
A = c + Ee−r(T −t)
B = p + F e−r(T −t) + a futures contract
At time T ,
ΛA = max{FT − E, 0} + E = max{FT , E},

ΛB = max{E − FT , 0} + F + (FT − F ) = max{E, FT }.
Hence we obtain A = B.
Chapter 5
Numerical Methods
To evaluate the option price numerically, we need to discretize our option price model. We recall
that there are two equivalent versions of option price models:
• The Black-Scholes equation with initial and boundary conditions:
σ2 2
Vt + S VSS = r(V − SVS ), S > 0, 0 ≤ t ≤ T,
2
V (S, T ) = Λ(S).
• The exact expression in terms of the transition probability of the risk-free asset price model:
Z ∞
−rT
V (S, 0) = e P(S 0 , T, S, 0)Λ(S 0 )dS 0 , (5.1)
0
where P is the transition probability for the risk-free asset price model
dS̃ = rS̃dt + σ S̃dz, 0≤t≤T
S̃(t) = S.
That is, P(S 0 , T, S, 0) is the probability density of S̃(T ) with S̃(0) = S.

The first one is a P.D.E. model. The second one is the exact solution of the P.D.E. It can also be
viewed as a solution of a stochastic model in a risk-free world. The finite difference method is a
numerical procedure for solving the Black-Scholes P.D.E., whereas the Monte-Carlo method is a
numerical procedure for the second expression. The binomial method can be derived from both
versions.
5.1 Monte Carlo method

We recall that the value of a European option is given by
Z
−rT
V (S, 0) = e P(S̃, T, S, 0)Λ(S̃)dS̃ (5.2)
65
66 CHAPTER 5. NUMERICAL METHODS
where Λ is the payoff function, P is the transition probability density function of S̃ which satisfies
dS̃
= rdt + σdz, ( initial state S(0) = S) (5.3)
S̃
i.e., it is the asset price model in the risk-neutral world. The Monte Carlo simulation is a numerical
procedure to evaluate V numerically based on the random simulation of the distribution of S̃(T ).
There are two ways to find an approximation of probability distribution of S̃(T ).
• Use the exact solution formula for S̃. Recall (2.10)
σ2

S̃(T ) = S exp (r − )T + σz(T ) , (5.4)
2
where z(T ) is the Brownian motion. We have seen that

√
z(T ) = T , ∼ N (0, 1).
We can simulate the distribution of S̃(T ) by using this formula. We generate M random
numbers i with distribution N (0, 1). We approximate
σ2 √

S̃i (T ) := S exp (r − )T + σ T i .
2
• Solving the s.d.e. numerically. We simulate M paths (for example, M = 10, 000) from
(5.3) to obtain S̃i (T ), i = 1, ..., M . To sample M paths from (5.3), we divide the interval
T
[0, T ] into N subintervals with equal length ∆t = N . We sample M N random numbers ik ,
i = 1, ..., M , k = 1, . . . , N with distribution N (0, 1) We then discretize (5.3) by using the
forward Euler method:
S̃i (k∆t) − S̃i ((k − 1)∆t) √

= r∆t + σik ∆t, k = 1, ..., N, i = 1, ..., M
S̃i ((k − 1)∆t)
S̃i (0) = S.
This approach certainly costs more. But it is for general s.d.e.
The option price at time 0 is the expectation of the payoff according the probability distribution of
S̃i (T ), then discounted by e−rT . Thus,
M
−rT 1 X
V (S, 0) ≈ e Λ(S̃i ).
M
i=1
5.2. BINOMIAL METHODS 67
Remarks.

1. The error from the numerical integration by Monte-Carlo method is O √1M . This can be
proved by the central limit theorem. The error from the numerical integration of the s.d.e.
1

for a sample path by using forward Euler method is O N . This can be proved by standard
error analysis of numerical ordinary differential equation. Thus, the totalerror for Monte-
Carlo method by using forward Euler integrator for s.d.e. is O √1M + N1 . If there is only
one underlying asset, the Monte Carlo does not have any advantage. However, if there are
many underlying assets, say more than three, the corresponding Black Scholes equation is a
diffusion equation in high dimensions. In this case, finite difference method is very difficult
and the Monte Carlo method wins.
2. The calibration of the volatility σ of an asset from market data has been demonstrated in
Section 2.9.
Below we provide a C code for Monte-Carlo method by using the exact formula of S̃(T ).
Monte_Carlo_European_option(V,S,E,r,sigma,T,M)
{
double St, sd, R;
double payoff(), randn(); //randn() is the random number generator obeying normal
distribution
int i;
R = (r-sigmaˆ2 /2)*T;
sd = sigma * sqrt(T);
V = 0.;
for (i=1; i <= M; i++)
{
St = S*exp(R+sd*randn(0,1));
V += payoff(St,E);
}
V *= exp(-r*T)/M;
}
double payoff(S,E)
{
return max(S-E,0);
}
5.2 Binomial Methods

The binomial model is to approximate the risk-neutral asset price model dS̃/S̃ = rdt + σdz by a
binomial model, then find the option price by taking expectation of this binomial model. We first
recall the binomial asset model. Then we illustrate how to find option price from this binomial
model.
5.2.1 Binomial method for asset price model

We consider the underlying asset is risk-neutral, i,e.,
dS̃
= rdt + σdz (5.5)
S̃
We shall approximate this continuous model by the following binomial model.
• We choose an N , define time step size ∆t = (T − t)/N . Define

√
∆x = σ ∆t,
u = e∆x , d = e−∆x .
• Define a random walk: {S 0 , ...S N } by
S n+1

u with probability p
= n = 0, ...N − 1
Sn d with probability 1 − p
with initial price

S 0 = S,
where the probability p is determined by
er∆t − d er∆t − e−∆x

p= = ∆x .
u−d e − e−∆x
The reason how u, d and p are chosen in this form will be illustrated later.
This process can be illustrated by a binomial tree. We define
Sj = Sej∆x , −N ≤ j ≤ N.
These are possible prices in every time step up to N . We start from S 0 = S0 = S. The price goes
up to S1 = Su = Se∆x , or down to S−1 = Sd = Se−∆x . In the second step, the possible prices
are
Su2 , Sud, Sdu, Sd2 ,
which are
S−2 , S0 , S2 .
After N steps, the possible values of the asset prices are
{Sj | |j| ≤ N and j + N = even}.
Each S n is a random variable. Let Pjn be the probability of the event that S n = Sj :
Pjn := P (S n = Sj ), |j| ≤ n, j + n = even.
Because the initial position is S0 , we have P00 = 1. After n step random walks, the probability
distribution of S n = Sj is the binomial distribution:
n ` n−` n + j = 2`
n
P (S = Sj ) = Pj =n ` p (1 − p)
0 otherwise.
Here, ` is the number of steps that the price goes up in n steps.

This binomial model depends on two parameters: u and p. ( The down ratio d = 1/u.) They
are determined by the conditions so that the discrete model and the continuous model have the same
mean and variance in one time step ∆t. We recall that these conditions are
pu + (1 − p)d = er∆t
2 )∆t
pu2 + (1 − p)d2 = e(2r+σ .
From these two equations, we get respectively
er∆t − d
p=
u−d
2
e(2r+σ )∆t − d2
p= .
u2 − d2
From these two equations, we get
2
e(2r+σ )∆t − d2
u+d= .
er∆t − d
Since d = u−1 , we get a quadratic equation for u:

2 )∆t
1 + e(2r+σ
u2 − αu + 1 = 0, where α= .
er∆t
Solving this equation, we get
1h p i
u= α + α2 − 4 .
2
From this, we recall that d and p can be expressed in terms of u:
1 er∆t − d
d= , p= .
u u−d
Thus, p, u and d can be expressed in terms of r, σ and ∆t.
By Taylor expansion, a simple calculation gives
√ 1 √
u = 1 + σ ∆t + σ 2 ∆t + O((∆t)3/2 ) ≈ eσ ∆t
2
Thus, up to O((∆t)−3/2 ) error, we can choose

√
u = eσ ∆t
,
This is why we choose √

∆x = σ ∆t
at the beginning. The probability p is thus given by
er∆t − d er∆t − e−∆x

p= = ∆x .
u−d e − e−∆x
A simple Taylor expansion gives
1 1 r∆t 1
p= + − ∆x + O(∆t + (∆x)2 )
2 2 ∆x 4
We may also choose
σ2

1 1 r∆t 1 1 ∆t
p= + − ∆x = + r − .
2 2 ∆x 4 2 2 2∆x
We should require ∆t so that 0 ≤ p ≤ 1. This leads to require
2

r − σ ∆t ≤ 1.

2 ∆x
5.2.2 Binomial method for option

We recall that the option price is the expectation of the payoff Λ(S) under the risk neutral probability
distribution P(S̃(T ), T, S0 , t), then deducted by e−r(T −t) . This transition probability distribution
can be approximated by a binomial distribution Pjn as shown in the last section. In the discrete
model, the risk neutral asset price only takes discrete values Sj , we shall approximate V (Sj , t +
n∆t) by Vjn . We notice that: if the random variable S n = Sj , then

n+1 Sj+1 with probability p
S =
Sj−1 with probability 1 − p
Therefore, the expected value of V at time step n at Sj should satisfy

Vjn = e−r∆t pVj+1 n+1 n+1
+ (1 − p)Vj−1 ,
the expectation of the binomial distribution discounted by e−r∆t . Thus, the binomial method for
option pricing is
(i) Choose N and define ∆t = (T − t)/N . N is chosen such that ∆t ≤ σ 2 /r2 .
√
(ii) Define ∆x = σ ∆t, and for each n = N, ..., 0, define Sj = S0 ej∆x with |j| ≤ N .
(iii) Define VjN = Λ(SjN ), and define Vjn for n = N − 1, N − 2, ..., 0, by

Vjn = e−r∆t pVj+1n+1 n+1
+ (1 − p)Vj−1 , for |j| ≤ n, j + n = even.
(iv) Output: V00 is the option price at time t for asset price S0 .
Example. For put option,
T = 5 months = 0.4167 year

∆t = 1 months = 0.0833 year
r = 0.1, σ = 0.4
S0 = $50,
√ E = $50
σ ∆t
u = e = 1.1224
d = 0.8909
p = 0.5076
er∆t = 1.0084
We begin to generate a binomial tree from S0 = 50 consisting of Sjn = S0 u` dn−` , where n+j = 2`,
−n ≤ j ≤ n. Then we compute Vjn inductively from n = N − 1 to n = 0 by

Vjn = e−r∆t pVj+1
n+1 n+1
+ (1 − p)Vj−1 .
with VjN being the payoff function Λ(SjN ). The value V00 is our answer.
A C code for European option is given below. In computer code, we use the data structure
• S[n][m], V [n][m], m = 0, ..., n, n = 0, ..., N . Before, we use Sjn , where n + j = even and
|j| ≤ n. The relation between j and m is m = (j + n)/2.
• The relation
Vjn = e−r∆t pVj+1
n+1 n+1
+ (1 − p)Vj−1
is replaced by
Vmn = e−r∆t pVm+1
n
+ (1 − p)Vmn .

Binomial_European_option(V,S,S0,E,r,sigma,dt,N)
{
double u,d,p;
double discount, tmp;
int m,n;
discount = exp(-r*dt);
u = exp(sigma*sqrt(dt));
d = 1/u;
p = (1/discount - d)/(u-d);
//Binomial tree for asset price: at level n.

S[0][0]=S0;
for (n = 1; n <= N; ++n)
{
for (m=n; m > 0; --m)
S[n][m] = u*S[n-1][m-1];
S[n][0] = d*S[n-1][0];
}
// Binomial tree for option.

for (m=0; m <=N; ++m)
V[N][m] = payoff(S[N][m],E);
for ( n = N-1; n >= 0; --n)
{
for (m=0; m < n; ++m)
V[n][m] = discount*(p*V[n+1][m+1] + (1-p)*V[n+1][m]);
}
}
5.3 Finite difference methods (for the modified B-S eq.)

In this section, we shall solve the Black-Scholes equation by finite difference methods. Recall that
the Black-Scholes equation is
1 ∂2V ∂V
Vt + σ 2 S 2 2 = r(V − S ).
2 ∂S ∂S
We set up the initial time is 0 and reserve the symbol t for time parameter. Using dimensionless
variables S = Eex , V = Ev, τ = T − t, we have
1 ∂2v 1 ∂v
vτ = σ 2 2 + (r − σ 2 ) − rv, τ ∈ [0, T ].
2 ∂x 2 ∂x
Let v = e−rτ u, then u satisfies
1 ∂2u 1 ∂u
uτ = σ 2 2 + (r − σ 2 ) , τ ∈ [0, T ]. (5.6)
2 ∂x 2 ∂x
The initial condition for u is
max{ex − 1, 0}

for call option
u(x, 0) =
max{1 − ex , 0} for put option
The far field boundary condition for u is
u(−∞, t) = 0, u(x, t) = ex erτ as x → ∞
for a call option, and

u(−∞, t) = 1, u(x, t) = 0 as x → ∞
for a put option.
5.3.1 Discretization methods

To solve (5.6) numerically, we follow the following procedure:
• Discretize space and time. We discretize [0, T ] into N steps, ∆τ = T /N . We choose a

small proper ∆x (to be determined later) and define xj = j∆x, j = −N, ..., N. We shall
approximate u(xj , n∆τ ) by Ujn .
5.3. FINITE DIFFERENCE METHODS (FOR THE MODIFIED B-S EQ.) 73
• Spatial discretization. We shall approximate the spatial derivatives by finite differences:
– ux is approximated by
uj+1 − uj−1
ux ← .
2∆x
– uxx is approximated by
uj+1 − 2uj + uj−1
uxx ← .
(∆x)2
Then the right-hand-side of (5.6) is discretized into
σ 2 Uj+1 − 2Uj + Uj−1 σ 2 Uj+1 − Uj−1

(QU )j ≡ ( ) + (r − )
2 (∆x)2 2 2∆x
• Temporal discretization. For the temporal discretization, we introduce the following three
methods:
– Forward Euler method:

Ujn+1 − Ujn
= (QU n )j .
∆τ
– Backward Euler method:
Ujn+1 − Ujn
= (QU n+1 )j .
∆τ
– Crank-Nicolson method:
Ujn+1 − Ujn 1
= [(QU n+1 )j + (QU n )j ].
∆τ 2
• Boundary conditions: we impose
n n
U−N = 0, UN = eN ∆x ern∆t
for call option.
For forward Euler method, we find Ujn inductively in n for n = 1, ..., N . We start from n = 0,
which is the initial condition. We are given Uj0 , j = −N, ..., N . From n to n + 1, the boundary
n+1 n+1
values U−N and UN are assigned. For −N < j < N , we find Ujn+1 by
Ujn+1 := Ujn + ∆τ Q(U n )j .
We stop this recursive process until all UjN are found.

For other methods, which involves Q(U n+1 ) in the equation of U n+1 . They can be expressed as
a linear system for U n+1 . There are standard method to solve such linear system. We shall discuss
this issue later.
5.3.2 Binomial method is a forward Euler finite difference method

We claim that the forward Euler method for u in backward time τ is indeed the binomial method for
option price V if we choose ∆x properly. Let us use Vjn to approximate the solution V (xj , n∆t),
where ∆t = ∆τ = T /N . We notice that the time variable for V is the forward time t and the
time variable for u and v is the backward time τ . Thus, V (xj , n∆t) = v(xj , (N − n)∆τ ) and
v = e−rτ u. We then have
Vjn = e−r(N −n)∆t UjN −n .
For the forward Euler method, we rewrite it in terms of V and also
UjN −n = UjN −n−1 + ∆τ (QU N −n−1 )j
In terms of Vjn , we have
er∆t Vjn = Vjn+1 + ∆tQ(V n+1 )j

1 ∆t 2 n+1 σ 2 ∆t
= Vjn+1 + σ (Vj+1 − 2Vj
n+1
+ V n+1
j−1 ) + (r − ) (V n+1 − Vj−1
n+1
)
2 (∆x)2 2 2∆x j+1
n+1
= aVj+1 + bVjn+1 + cVj−1
n+1
where
1 ∆t 2 σ 2 ∆t
a= σ + (r − )
2 (∆x)2 2 2∆x
∆t 2 n+1
b = (1 − σ )Vj ,
(∆x)2
1 ∆t 2 σ 2 ∆t
c= σ − (r − ) .
2 (∆x)2 2 2∆x
Notice that
a + b + c = 1.
We should require
a, b, c ≥ 0,
n+1
for stability reason to be explained later. Thus Vjn is the “average” of Vj−1 , Vjn+1 , Vj+1
n+1
with
weights a, b, c, then discounted by e−r∆t .
The stability condition a, b, c ≥ 0 reads
2 ∆t

r − σ ∆t 2
≤ σ ≤1
2 ∆x (∆x)2
These give constraints on ∆x and ∆t:
1 (∆x)2
∆x ≤ r
, ∆t ≤ .

σ2
− 12 σ2
Binomial case: b = 0 When we choose

√
∆x = σ ∆t,
then b = 0. In this case,

er∆t Vjn+1 = pVj+1
n+1 n+1
+ (1 − p)Vj−1
where
1 σ 2 ∆t
p=a= + (r − ) .
2 2 2∆x
The stability condition is satisfied if and only if
0 ≤ p ≤ 1. (5.7)
We see that this finite difference is identical to the binomial method in the previous section.
5.3.3 Stability
Definition 5.4. A finite difference method is called consistent to the corresponding P.D.E. if for any
solution of the corresponding P.D.E., it satisfies
F.D.E (finite difference equation) + (∆x, ∆t)
and → 0 as ∆t, ∆x → 0
Definition 5.5. The truncation error of a finite difference method is defined to be the function
(∆x, ∆t) in the previous definition.
For instance, the truncation error for central difference is
σ2 σ2
Qu = uxx + (r − )ux + O((∆x)2 ).
2 2
And the truncation for various temporal discretizations are
1. Forward Euler:
u(j∆x, (n + 1)∆t) − u(j∆x, n∆t)
− (Qu)(j∆x, n∆t) = O((∆x)2 ) + O(∆t).
∆t
2. Backward Euler:
u(j∆x, (n + 1)∆t) − u(j∆x, n∆t)
− (Qu)(j∆x, (n + 1)∆t) = O((∆x)2 ) + O(∆t).
∆t
3. Crank-Nicolson method
u(j∆x, (n + 1)∆t) − u(j∆x, n∆t) 1
− [(Qu)(j∆x, (n + 1)∆t) + (Qu)(j∆x, n∆t)]
∆t 2
= O((∆x)2 ) + O((∆t)2 ).
The true error Ujn − u(j∆x, n∆t) is usually estimated in terms of the truncation error.
Definition 5.6. A finite difference equation is said to be (L2 −)stable if the norm
X
kU n k2 := |Ujn |2 ∆x
j
is bounded for all n ≥ 0.
Definition 5.7. A finite difference method for a P.D.E. is convergent if its solution Ujn converges to
the solution u(j∆x, n∆t) of the corresponding P.D.E..
Theorem 5.5 (Lax). For linear partial differential equations, a finite difference method is conver-
gent if and only if it is consistent and stable.
This theorem is standard and its proof can be found in most numerical analysis text book. We
therefore omit it here.
Since the consistency is easily to achieve, we shall focus on the stability issue. A standard
method to analyze stability issue is the von Neumann stability analysis. It works for P.D.E. with
constant coefficients. It also works “locally” and serves as a necessary condition for linear P.D.E.
with variable coefficients and nonlinear P.D.E.. We describe his method below.
We take Fourier transform of {Uj }∞ j=−∞ by defining
∞
X
Û (ξ) = Uj e−ijξ
j=−∞
It is a well-known fact that

Z π
X 1
|Uj |2 = √ |Û (ξ)|2 dξ
j
2π −π
≡ kÛ k2
Thus, the boundedness of j |Uj |2 can be estimated by using kÛ k2 . The advantage of using Û is
P
that the finite difference operation becomes a multiplier in terms of Û . Namely,

X Uj+1 − Uj−1
DU (ξ) =
d e−ijξ
2
j
!
X Uj ei(j+1)ξ − Uj ei(j−1)ξ
=
2
j
e − e−iξ X
iξ
= Uj e−ijξ
2
j
= (2i sin ξ)Û (ξ)

For the finite difference operator QU ,m we have

X
\) =
(QU (Qu)j e−ijξ
j
σ2 1 σ2 1
= [ (2cosξ − 2) + (r − ) (2isinξ)]Û
2 (∆x)2 2 ∆x
≡ Q(ξ)
b Û (ξ).
For forward Euler method,

n+1 (ξ) = (1 + ∆tQ(ξ))
U[ b Ucn
= G(ξ)Ucn
= G(ξ)n+1 U
cn
We observe that
Z π Z
cn (ξ)|2 dξ =
|U |G(ξ)|2n U
c0 (ξ)|2 dξ
−pi
Z
2n c0 (ξ)|2 dξ
≤ max |G(ξ)| |U
ξ∈(−π,π)
If |G(ξ)| ≤ 1, i∀ξ ∈ (−π, π), then stability condition holds. On the other hand, if |G(ξ)| > 1 at
some point ξ0 , then by the continuity of G, we have that
|G(ξ)| ≥ 1 +
for some small > 0 and for all ξ with |ξ −ξ0 | ≤ δ for some δ > 0. Let consider an initial condition
such that
Uc0 (ξ) = 1 |ξ − ξ0 | ≤ δ
0 otherwise.
cn will have
Then the corresponding U
Z π Z
cn (ξ)|2 dξ
|U = |G(ξ)|2n U
c0 (ξ)|2 dξ
−pi
→ ∞
as n → ∞. We conclude the above discussion by the following theorem.

Theorem 5.6. For a finite difference equation with constant coefficients, suppose its fourier trans-
form satisfies’
U[n+1 (ξ) = G(ξ)Ucn (ξ)
Then the finite difference equation is stable if and only if
|G(ξ)| ≤ 1 ∀ξ ∈ (−π, π].

Example: Let apply the forward Euler method for the heat equation: ut = uxx . Then
Ujn+1 − Ujn 1 ∆t
= (U n − 2Ujn + Uj−1
n
), =⇒ U n+1 = U n + D2 U n
∆t (∆x)2 j+1 (∆x)2
From von Neumann analysis:
n+1 = [1 +
∆t cn
U[ (2 cos ξ − 2)]U
(∆x)2
∆t ξ cn
= (1 − 4 sin2 )U
(∆x)2 2
≡ G(ξ)Ucn .
Hence
∆t ξ 1 ∆t 1
|G(ξ)| ≤ 1 =⇒ 2
sin2 ≤ =⇒ 2
≤ (stability condition)
(∆x) 2 2 (∆x) 2
If we rewrite the finite difference scheme by
∆t ∆t ∆t
Ujn+1 = n
Uj+1 + (1 − 2 )Ujn + Un
(∆x)2 (∆x) 2 (∆x)2 j−1
n
≡ aUj+1 + bUjn + cUj−1
n
.
Then the stability condition is equivalent to
a, b, c ≥ 0.
Since we have a + b + c = 1 from the definition, thus we see that the finite difference scheme is
nothing but saying Ujn+1 is the average of Uj+1
n , U n and U n with weights a, b, c. In particular, if
j j−1
we choose (∆x)2 = 2 , then b = 0. If we rename a = p, c = 1 − p, then Ujn+1 = pUj+1
∆t 1 n + (1 −
n
p)Uj−1 . This can be related to the random walk as the follows.
Consider a particle move randomly on the grid points j∆x. In one time step, the particle moves
toward right with probability p and left with probability 1 − p. Let Ujn be the probability of the
particle at j∆x at time step n for a random walk.
Ujn+1 = pUj−1
n n
+ (1 − p)Uj+1 .
We can also apply the above stability analysis to backward Euler method and the Crank-Nicolson
method. Let us only demonstrate the analysis for the heat equation. We left the analysis for the
Black-Scholes equations as exercises.
1. For backward Euler method,
Ujn+1 − Ujn 1
= (U n+1 − 2Ujn+1 + Uj+1
n+1
).
∆t (∆x)2 j−1
Then
ξ ∆t [
(1 + (4 sin2 ) )U n+1 = U
cn i =⇒iU[
n+1 = G(ξ)U
cn ,
2 (∆x)2
where
1
G(ξ) = ∆t 2 ξ
.
1+ 4 (∆x)2 (sin 2 )
We find that |G(ξ)| ≤ 1, for all ξ. Hence, the backward Euler method is always stable.
2. For Crank-Nicolson method,
Ujn+1 − Ujn 1 h
n+1 n+1 n+1 n n n
i
= (Uj+1 − 2Uj + Uj−1 ) + (Uj+1 − 2Uj + U j−1 ) .
∆t 2(∆x)2
Its Fourier transform satisfies
n+1 − U
cn
U[ 1 2 ξ [
n+1 2 ξ cn
= −(4 sin )U − (4 sin )U .
∆t 2(∆x)2 2 2
We have
∆t ξ ∆t ξ cn
(1 + 2 2
(sin2 ))U[
n+1 = (1 − 2
2
(sin2 ))U
(∆x) 2 (∆x) 2
and hence
∆t 2 ξ
1 − 2 (∆x)2 sin 2 cn
n+1 =
U[ U .
∆t 2 ξ
1+ 2 (∆x)2 sin 2
∆t 2 ξ 1−α
Let α = 2 (∆x) 2 sin 2 , then G(ξ) = 1+α . We find that for all α ≥ 0, |G(ξ)| ≤ 1, hence
Crank-Nicolson method is always stable for all ∆t, ∆x > 0.
Exercise study the stability criterion for the modified Black-Scholes equation
σ2 σ2
uτ = uxx + (r − )ux ,
2 2
for the forward Euler method, back Euler method and Crank-Nicolson method.
5.3.4 Convergence
Let us study the convergence for finite difference schemes for the modified Black-Scholes equation.
Let us take the forward Euler scheme as our example. The method below can also be applied to
other scheme.
The forward Euler scheme is given by:
Ujn+1 − Ujn
= (QU n )j
∆t
We have known that it has first-order truncation error, namely, suppose unj := u(j∆x, n∆t), where
u is the solution of the modified Black-Scholes equation, then
un+1
j − unj
= (Qun )j + O(∆t) + O((∆x)2 ).
∆t
We subtract the above two equations, and let enj denotes for unj −Ujn and nj denotes for the truncation
error. Then we obtain
en+1
j − enj
= (Qen )j + nj
∆t
Or equivalently,
en+1
j = aenj+1 + benj + cenj−1 + ∆tnj (5.8)
Here, a, b, c ≥ 0 and a + b + c = 1. We can take Fourier transformation ebn of en . It satisfies
n+1 (ξ) = G(ξ)ebn (ξ) + ∆tbn (ξ)

ed
where
G(ξ) = aeiξ + b + ce−iξ .
Recall that the stability |G(ξ)| ≤ 1 is equivalent to a, b, c ≥ 0. Thus, by applying the above
recursive formula, we obtain
kebn k ≤ ked
n−1 k + ∆tkd
n−1 k

n−2 k + ∆t kGd
≤ ked n−2 k + kd
n−1 k

n−2 k + ∆t kd
≤ ked n−2 k + kdn−1 k
n−1
X
≤ keb0 k + ∆t kbk k
k=0
≤ O(∆t) + O((∆x)2 ).
Here, we have used the estimate for the truncation error
kn k = O(∆t) + O((∆x)2 ,
and that n∆t = O(1). We conclude the error analysis as the following theorem.
Theorem 5.7. The error enj := u(j∆x, n∆t) − Ujn for the Euler method has the following conver-
gence rate estimate:
X
( |enj |2 ∆x)1/2 ≤ O(∆t) + O((∆x)2 ), for all n.
j
It is simpler to ontain the maximum norm estimate. Let E(n) := maxj |enj | be the maximum
error. From (5.8), we have
|en+1
j | ≤ a|enj+1 | + b|enj | + c|enj−1 | + ∆t|nj |
≤ aE(n) + bE(n) + cE(n) + ∆t
= E(n) + ∆t
where
:= max |nj | = O(∆t) + O((∆x)2 ).
j,n
Hence,
E(n + 1) ≤ E(n) + ∆t.
Since we take Uj0 = u0j , there is no error initially. Hence, we have
n−1
X
E(n) ≤ ∆t
k=0
≤ n∆t
Since n∆t is a fixed number, as we take the limit n → ∞, we obtain the error is bounded by the
truncation error. We summarize the above discussion as the following theorem.
Theorem 5.8. The error enj := u(j∆x, n∆t) − Ujn for the Euler method has the following conver-
gence rate estimate:
max |enj | ≤ O(∆t) + O((∆x)2 ).
j
Exercise. Prove that the true error of the Crank-Nicolson scheme is O((∆t)2 ) + O((∆x)2 ).
5.3.5 Boundary condition

For the modified Black-Scholes equation, we have
u(−∞, t) = 0, u(x, t) = ex erτ as x → ∞
for a call option, and

u(−∞, t) = 1, u(x, t) = 0 as x → ∞
In computation, we can choose a finite domain (xL , xR ) with xL << −1 and xR >> 1. The
boundary condition at the boundary points are an approximation to the above far field boundary
condition.
In practice, we don’t even use this boundary condition. Indeed, if we want to know u(xk , n∆t),
we can find the numerical domain of this quantity, which is the triangle
{(j∆x, m∆t) | |j − k| ≤ n − m}
We only need to compute u in this domain, which needs no boundary data.

5.4 Converting the B-S equation to finite domain

The transformation x = log(S/E) converts the B-S equation to a heat equation. However, the
domain of x is the whole real line. For numerical computation, it is desirable to have a finite
computation domain. The transformation in this section converts S to ξ with ξ ∈ (0, 1). The
price is that the resulting equation has variable coefficients. But this is not a problem for numerical
computation.
We define the transformation:
S
ξ = (5.9)
S+E
V (S, t)
V = (5.10)
S+E
τ = T − t. (5.11)
Notice that ξ is dimensionless and important values of ξ are near 1/2. With this, the inverse trans-
formation is
Eξ dξ (1 − ξ)2
S= , =
1 − ξ dS E
We plug this transformation to the B-S equation. We allow σ depend on S. Define σ̄(ξ) = σ(Eξ/1−
ξ). Then the resulting equation is
∂V 1 ∂2V ∂V
= σ̄ 2 (ξ)ξ 2 (1 − ξ)2 2 + rξ(1 − ξ) − r(1 − ξ)V , (5.12)
∂τ 2 ∂ξ ∂ξ
for 0 ≤ ξ ≤ 1 and τ > 0. The initial data reads

1−ξ Eξ
V (ξ, 0) = Λ . (5.13)
E 1−ξ
For a call option, the payoff is Λ(S) = max(S − E, 0). The corresponding
V (ξ, 0) = max(S − E, 0)(1 − ξ)/E

= max(2ξ − 1, 0).
Similarly, V (ξ, 0) = max(1 − 2ξ, 0) for a put option.

On the boundaries ξ = 0 and ξ = 1, the diffusion coefficients are degenerate. If the solution
is smooth up to the boundaries, then on the boundary, the equation is degenerate to the following
ordinary differential equations:
∂V (0, τ )
= −rV (0, τ )
∂τ
∂V (1, τ )
= 0.
∂τ
5.5. FAST ALGORITHMS FOR SOLVING LINEAR SYSTEMS 83
The corresponding solutions are
V (0, τ ) = V (0, 0)erτ (5.14)

V (1, τ ) = V (1, 0). (5.15)
We can discretize equation (5.12) by finite difference method. Let ∆ξ and ∆τ are the spatial
and temporal mesh sizes, respectively. Let ξj = j∆ξ, τ n = n∆τ . The boundaries points are ξ0 and
ξM . We use central difference for ∂ 2 V /∂ξ 2 and ∂V /∂ξ. The resulting finite difference equation
reads
dvj 1 2 2 vj+1 − 2vj + vj−1

= σj ξj (1 − ξj )2
dτ 2 ∆ξ 2
vj+1 − vj−1
+rξj (1 − ξj ) − r(1 − ξj )vj
2∆ξ
We can discretize this equation in the time direction by forward Euler method. The stability con-
straint is
1
rξj (1 − ξj ) − σj ξj (1 − ξj ) ∆τ ≤ σj2 ξj2 (1 − ξj )2 ∆τ ≤ 1.
2 2 2

(5.16)
2 ∆ξ 2∆ξ 2
Remark. Many options have non-smooth payoff functions. This causes low order accuracy for
finite difference scheme. Fortunately, many simple payoff function has exact solution. For instance,
the European call option. For general payoff function, we may subtract its non-smooth part for
which an exact solution is available. The remainder is smooth, and a finite difference scheme can
yield high-order accuracy.
5.5 Fast algorithms for solving linear systems

In the backward Euler method and the Crank-Nicolson method, we need to solve linear systems of
the form
AU = F.
For the backward Euler scheme,
A = diag (−a, 1 + a + c, −c)

 
1+a+c −c 0 ···
 −a 1+a+c −c 0 ··· 
 

 0 −a 1 + a + c −c 0 ··· 

:= 
 . . . . . . . . 
 . . . . 


 · · · 0 −a 1 + a + c −c 0 

 ··· 0 −a 1+a+c −c 
··· 0 −a 1+a+c
and  n+1 
UjnL +1
 
aUjL
 ..   .. 
U =  . , F =  . .
UjnR −1 cUjn+1
R
For the Crank-Nicolson scheme We have
AU n+1 = BU n + bn+1/2
where A = diag (− a2 , 1 + a
2 + 2c , − 2c ) B = diag ( a2 , 1 − a
2 − 2c , 2c ), and
a(Ujn+1 + UjnL )/2

 
L

 0 

bn+1/2 = .
.. .
 
 
 0 
n+1 n
c(UjR + UjR )/2
Now, we concentrate on solving the linear system
Ax = f.
The matrix A is tridiagonal and diagonally dominant. Let us rewrite A = diag (a, b, c). Here, the
constants a, b, c are different from the average weights we had before. We may assume b > 0.
We say that A is diagonally dominant if b > |a| + |c|. More generally, A may takes the form
A = diag (aj , bj , cj ). and |bj | > |a| + |cj |. Without loss of generality, we may normalize the
j − th so that bj = 1.
There are two classes of methods to solve the above linear systems. One is called direct methods,
the other is called iterative methods. For one-dimensional case as we have here, direct method is
usually better. However, for high-dimensional cases, iterative methods are better.
5.5.1 Direct methods

Gaussian elimination
Let us illustrate this method by the simple example: A = diag (a, 1, c). We multiple the first
equation by −a and add it into the second equation to eliminate the term xjL +1 in the second
equation. Then the resulting equation becomes
    
1 c 0 0 ··· 0 xjL +1 bjL +1
0 1 − ac c 0 · · · 0 xjL +2  −abjL +1 + bjL +2 
   

0
 a 1 c ··· 0 xjL +3  = 
    bjL +3 

 ..   ..   .. 
 . 0   .   . 
··· 1 xjR −1 bjR −1
We continue to eliminate the term a in the third equation, and so on. Finally, we arrive
    0 
1 c 0 0 ··· 0 xjL +1 bjL +1
0 1 − ac 0 ···   0
c  xjL +2  bjL +2 
0  

0 0 1 − c/(1 − ac) c ··· 0
0 xjL +3  = bjL +3 
   
 
 ..   ..   . 
.
0 0 0 . 0  .   . 
0 0 0 0 ··· 1 xjR −1 b0jR −1
Then xj can be solved easily. The diagonal dominance condition guarantee that the reduced matrix
is also diagonally dominant. Thus, this scheme is numerical stable.
LU decomposition
We decompose A = LU , where
   
1 0 0 ··· 0 ujL +1 vjL +1 0 ··· 0
 .. ..   .. .. 
`j +2
 L 1 . .
 0
 ujL +2 . .
L= 0
 . .. . .. .. 
, U = 0
 .. .. .. 
. 0 . . . 0
 ..  ..
   
.. ..  .. .. 
 . . .0   . . . vjR −2 
0 · · · 0 `jR −1 1 0 ···0 0 ujR −1
It is easy to find a recursion formula to find the coefficients `, u and v’s. Once these are found, we
can find x by solving
Ly = b, U x = y.
These two equations are easy to solve. One can show that both L and U are diagonally dominant if
A is.
If we watch carefully, LU-decomposition is equivalent to the Gaussian elimination.
Cyclic reduction method

Let us take the case A = diag (a, 1, c) to illustrate this method. Consider three consecutive equa-
tions
ax2j−2 + x2j−1 + cx2j = b2j−1

ax2j−1 + x2j + cx2j+1 = b2j
ax2j + x2j+1 + cx2j+2 = b2j+1
We can eliminate the odd-index terms x2j−1 and x2j+1 . Namely, −a × (2j − 1)−eq +(2j)-eq
−c × (2j + 1)-eq: After normalization, we obtain
a0 x2j−2 + x2j + c0 x2j+2 = b0j

Here,
a2 c2
a0 = − , c0 = − ,
1 − 2ac 1 − 2ac
b0j = (b2j − ab2j−1 − cb2j+1 )/(1 − 2ac).
If we rename x0j = x2j . Then we have A0 x0 = b0 , where A0 = diag (a0 , 1, c0 ). Notice that the
system is reduced to half and with the same form. One can show that the iterative mapping
0
a a
7→ 0
c c
converges to (0, 0)t quadratically fast, provided |a| + |c| < 1 initially. Thus, for few iteration, the
matrix A is almost an identity matrix. We can invert it trivially. Once x2j are found, the odd-index
x + 2j + 1 can be found from the equation:
ax2j + x2j+1 + cx2j+2 = b2j+1 .
A careful reader should find that the cyclic reduction is also a version of the Gaussian elimination
method.
5.5.2 Iterative methods

Most iterative methods can be viewed as a proper decomposition of A, then solve an important and
treat the rest as a perturbation term.
Jacob method
In Jacobi method, wer decompose
A=D+B
where D is the diagonal part and B is the off diagonal part. Since A is diagonally dominant, we
may approximate x by the sequence xn , where xn is defined by the following iteration scheme:
Dxn+1 + Bxn = b.
Let the error en := xn+1 − xn . Then
Den = −Ben−1
Or
en = −D−1 Ben−1 .
Let us define the maximum norm
ken k := max |enj |
j
Then
a n−1 c n−1
|enj | = | − ej−1 − ej+1 |
b b
|a| n−1 |c|
≤ | ke k + ken−1 k
|b| |b|
|a| + |c| n−1
= ke k
|b|
Hence,
ken k ≤ ρken−1 k
where ρ = |a|+|c|
|b| < 1 from the fact that A is diagonally dominant. This yields the convergence of
the sequence xn . The limit x satisfies the equation Ax = b.
Gauss-Seidel method
In Gauss-Seidel method, A is decomposed into A = (D + L) + U , where D is the diagonal part, L,

the lower triangular part, and U , the upper triangular part of A. The approximate solution sequence
is given by
(D + L)xn+1 + U xn = b.
As before, the error en := xn+1 − xn satisfies
en = −(D + L)−1 U en−1
To analyze the decay of en , we use Fourier method. Let

X
ebn (ξ) := enj e−ijξ .
j
Then we have
ebn (ξ) = G(ξ)ed

n−1 ,
ceiξ
G(ξ) = −
b + ae−iξ
It is easy to see that the amplification matrix G satisfies
max |G(ξ)| := ρ < 1, provided |b| > |a| + |c|. (5.17)

ξ
This shows that the Gauss-Seidel method also converges for diagonally dominant matrix.
Exercise. Show the above statement (5.17).
Successive over-relaxation method (SOR)

In the methods of Jacobi and Gauss-Seidel, the approximate sequence xn is usually convergent
monotonely. We therefore have a chance to speed them up by an extrapolation procedure described
below.
y n+1 = −D−1 (L + U )xn + b

xn+1 = xn + ω(y n+1 − xn ).
Here, ω is a parameter. In order to speed up, we require ω > 1. We also need to require ω < 2 for
stability. The optimal ω is chosen to minimize the amplification matrix Gω (ξ).
Exercise. Find the amplification matrix Gω and the optimal ω for the matrix A = diag (a, b, c).
Also, determine the rate
ρ := min max |Gω (ξ)|.
ω ξ
Chapter 6
American Option
6.1 Introduction
An American option has the right to exercise any time during the life of the option. We denote the
American call by C and American put by P . The first important thing we should note is that the
value of an American option is greater than or equal to the payoff function: V (S, t) ≥ Λ(S, t).
Otherwise, there is an arbitrage opportunity because we can buy the American option then exercise
it immediately to gain a net profit V − Λ.
We recall that the value of an American call option is equal to that of a European call option,
see Theorem 3.2. However, for other cases like the the American put option or the American call
option on dividend-paying asset, the American options do cost more. We explain why it is so below.
American put First, we notice that there must be some region in 0 < S < E in the S-P plane
where the European put p(S, t) < E − S. In fact, we know that p(0, t) = Ee−r(T −t) < E − 0,
which is below the line p = E − S in a neighborhood of S = 0, see Figure 6.1. On the other hand,
the corresponding American option P (S, t) must satisfy
P (S, t) ≥ Λ(S) = E − S.
Otherwise, we can buy an asset S and a put P (S, t), then exercise it immediately. We have the
right to sell S on E. This gives a risk-free profit: E − P − S > 0. Thus, in the region where
p(S, t) < E − S ≤ P (S, t), we must have p(S, t) < P (S, t).
American call on a dividend-paying asset For an European call on dividend-paying asset, its
value must lie below the payoff function Λ(S) = S − E in some region S > E, see Figure 6.2. In
fact, c(S, t) ∼ Se−D0 (T −t) for large S, which is below the line c = S − E. Therefore, there must
be a region in S > E in (S, t)-plane where c(S, t) < Λ(S) := max{S − E, 0}. In this region, if
we could exercise the corresponding American call option, then based on V (S, t) ≥ Λ(S, t) (i.e.
C(S, t) ≥ S − E), the corresponding American call option C must also satisfy
C(S, t) ≥ Λ(S) = max{S − E, 0} > c(S, t).
89
90 CHAPTER 6. AMERICAN OPTION
E
x
Ee−r(T−t) x
x
E S
Figure 6.1: The dash line is the price of an European put option. There is a region where p(S, t) <
max{E − S, 0}. This can be seen from the boundary condition P (0, t) = Ee−r(T −t) , which is
below the payoff function p = E − S.
We summarize the discussion as below.
Proposition 6. American option V always satisfies V (S, t) ≥ Λ(S). Furthermore,
• for American put option, P (S, t) > p(S, t) in some region where p(S, t) < Λp (S);
• for American call option on dividend-paying asset, C(S, t) > c(S, t) in some region where
c < Λc (S).
6.2 American options as a free boundary value problem

6.2.1 American put option
Existence of optimal exercise price Sf (t). First, there must be some value of S for which it is
optimal from the holder’s point of view to exercise the American option. Otherwise, we should
hold the option for all possible S. Then this option is identical to a European option. But we have
seen that this is not the case. Hence, for each time t, there is an optimal exercise price Sf (t) such
that, when S < Sf (t) one should exercise the put option, and when S > Sf (t), we should hold the
option. In other words, we should have:
• P (S, t) ≡ max(E − S, 0) for S < Sf (t),
• P (S, t) satisfies the Black-Scholes equation for S > Sf (t).

6.2. AMERICAN OPTIONS AS A FREE BOUNDARY VALUE PROBLEM 91
x
E S
Figure 6.2: Dash line is the European call option on a dividend-paying asset. Notice that c(S, t) ∼
Se−D0 (T −t) for S >> 1, hence C(S, t) < Λ(S) = max(S − E, 0) for some Sf (t).
The first statement can be viewed as the definition Sf (t). That is,
Sf (t) := max{S|S ≤ E and P (S, t) ≡ max(E − S, 0)}. (6.1)
The optimal exercise price can also be viewed as the minimal exercise price that one should hold
the put option:
Sf (·) = min{b(·)|P (S, t) is a B-S solution in S > b(t) satisfying P (b(t), t) = E − b(t)}.
This means that, given a exercise price curve b(·), the holder holds the option if S > b(t) and
exercise it if S < b(t). When it is exercised, the payoff is E − b(t). The value of P (b(t), t) must
be equal to E − b(t) by the no-arbitrage assumption, otherwise we can buy or short P then exercise
it right away to gain a net profit. Thus, given an exercise price curve b(·), we can find a solution
P (S, t) with boundary condition P (b(t), t) = E − b(t). The optimal exercise price Sf (·) is the
smallest such exercise price curve.
Notice that
Sf (t) ≤ E.
If S (or Sf (t)) goes up beyond E, then the right of a put to sell a higher price S with a lower price
E is meaningless. Thus, such put P (S, t) has no value at all. This means that the optimal exercise
price cannot exceed E. Furthermore, at time T , we should have
Sf (T ) = E.
This follows from the Definition (6.1).

However, we don’t know the optimal exercise price Sf (t), t < T in advance. It is a new
unknown, called free boundary. On this free boundary, there will be two boundary conditions for
P . One boundary condition is used to determine the B-S solution PBS uniquely in the region
S > Sf (t), 0 ≤ t ≤ T . The other boundary condition is used to determine the free boundary itself.
The following proposition gives the proper boundary conditions on Sf (t).
Proposition 7. Based on the no-arbitrage hypothesis, the price of an American put should satisfy
the following boundary conditions on the free boundary Sf (t):
• P (Sf (t), t) = E − Sf (t),

∂P
• ∂S (Sf (t), t) = −1.
Remark. Notice that in the region S < Sf (t), P (S, t) ≡ E − S, where we have ∂P/∂S(S, t) ≡
−1. Therefore, the above boundary conditions are equivalent to
• P (S, t) is continuous across the free boundary S = Sf (t).

∂P
• ∂S (S, t) is continuous across the free boundary S = Sf (t).
Proof. 1. Given the optimal price Sf (t), it means that it is the price to the holder to excise it,
the return is E − Sf (t). The corresponding option price must equal E − Sf (t). Otherwise,
there is an arbitrage opportunity. Thus, we get P (Sf (t), t) = E − Sf (t) for 0 ≤ t ≤ T .
2. The free boundary S = Sf (t) is an unknown. We want to derive a constraint to determine it.
Suppose we are given a boundary S = b(t), 0 ≤ t ≤ T . We solve the B-S equation in the
region S > b(t), 0 ≤ t ≤ T with boundary condition
P (b(t), t) = E − b(t)
and final condition P (S, T ) = max(E − S, 0). Such a solution depends on the path b(·). We
denote it by P (S, t; b(·)). The optimal exercise price Sf (t) is the minimal such curve b(·).
Below Sf (t), then one should exercise the put option. Therefore, a small variation of P with
respective to the change of b(·) is zero at the optimal exercise price Sf (·). This means
δP (Sf (t), t; Sf (·))

= 0.
δb(·)
Here, δP/δb means the variation of P w.r.t. the path b(·). 1
1
The variation of P w.r.t. a path b(·) is defined as below. Consider a small change of the path, say b(·) + r(·). The
function r is called the direction of the variation. Then, just like the definition of gradient in n dimensions, we define the
variation of P w.r.t. b(·) at a path b(·) by
Z T
d δP (S, t; b(s))
|=0 P (S, t; b + r) = r(s) ds.
d 0 δb(·)
6.2. AMERICAN OPTIONS AS A FREE BOUNDARY VALUE PROBLEM 93
3. On the boundary S = b(t), P satisfies the boundary condition

P (b(t), t; b(·)) = E − b(t).
We can take variation of this formula w.r.t. b(·), then evaluate at the optimal exercise price
Sf (·). This gives
∂P δb(·) δP (Sf (t), t; Sf (·))
(Sf (t), t; Sf (·)) b(·)=Sf (·) + = −1.
∂S δb(·) δb(·)
Since the second term on the left-hand side is zero. We get
∂P
(Sf (t), t; Sf (·)) = −1.
∂S
Thus, we treat the American put option as the following free boundary value problem. There exists
an optimal exercise price Sf (t) such that
1. for S < Sf (t), early exercise is optimal, and P (S, t) = E − S;
2. for S > Sf (t), one should hold the put option and P satisfies the Black-Scholes equation:
∂P 1 ∂2P ∂P
+ σ 2 S 2 2 + rS − rP = 0;
∂t 2 ∂S ∂S
∂P
3. across the free boundary (Sf (t), t), both P and ∂S are continuous.
We will solve this free boundary value problem in the next section.
6.2.2 American call option on a dividend-paying asset

As we have seen in the introduction of this chapter that an American call option C(S, t) on a
dividend-paying asset has asymptotic value C(S, t) ∼ Se−D0 (T −t) for large S. This value is below
the payoff function Λ ≡ max(S − E, 0). Therefore, there must an optimal Sf (t) such that we
should exercise this call option when S > Sf (t) and hold it when S < Sf (t). On the free boundary
S = Sf (t), based on the no-arbitragy hypothesis, we should require that
• both C and ∂C/∂S are continuous across the free boundary (Sf (t), t) for 0 < t < T .
We summarize this by the following equations
σ2 2 ∂ 2C ∂C
Ct + S 2
+ (r − D0 )S − rC = 0, 0 < S < Sf (t)
2 ∂S ∂S
C(S, t) ≡ Λ(S) ≡ max{S − E, 0}, S > Sf (t).
On the free boundary S = Sf (t), the boundary condition is required
C(Sf (t), t) = Sf (t) − E,
∂C
(Sf (t), t) = 1, 0 ≤ t ≤ T.
∂S
6.3 American option as a linear complementary problem

The American option can also be formulated as a linear complementary problem, where the free
boundary is treated implicitly and numerical schemes can be easily designed. To illustrate this linear
complementary problem, first we notice that an American option V should satisfy the following
conditions:
(i) V (S, t) ≥ Λ(S),
2
(ii) Vt + 21 σ 2 S 2 ∂∂SV2 ≤ r(V − S ∂V
∂S ),
2
(iii) either V (S, t) = Λ(S) or Vt + 21 σ 2 S 2 ∂∂SV2 = r(V − S ∂V
∂S ) should hold,
∂V
(iv) both V and ∂S are continuous.
Here, Λ(S) = max(E − S, 0) is the corresponding payoff function.
We have seen the reasons for (i), and (iv). We explain the reasons of (ii) below. Let us consider
the portfolio Π = V − ∆S. As we have seen that the Delta hedge eliminate the randomness of Π
and yields
1 ∂2V
dΠ = Vt + σ 2 S 2 2 .
2 ∂S
When it is optimal to hold the option, then
∂V
dΠ = r(V − S).
∂S
Otherwise, we should have
∂V
dΠ ≤ r(V − S),
∂S
based on no arbitrage opportunities. Thus, the Black-Scholes is replaced by the Black-Scholes
inequality.
To show (iii), we know that if we exercise the option, then V = Λ, otherwise, we hold the option
and its value should satisfy the Black-Scholes equation.
Properties (i)-(iii) can be formulated as the following linear complementary equations:

 V −Λ≥0 2

Vt + 12 σ 2S 2 ∂∂SV2 − r(V − S ∂V
∂S ) ≤ 0, (6.2)
 (V − Λ) V + 1 σ 2 S 2 ∂ 2 V − r(V − S ∂V ) = 0

t 2 ∂S 2 ∂S
We look for C 1 solution (i.e. both V and ∂V

∂S are continuous) in the domain S ∈ (0, ∞) and t ≥ 0
with final condition
P (S, T ) = Λ(S).
Such a problem is called a linear complementary problem. The advantage of this formulation is that
the free boundary is treated implicitly, which can be solved by some penalty method or projection
method to be explained below.
6.4. *PENALTY METHOD FOR LINEAR COMPLEMENTARY PROBLEM 95
Remark For American put, in the region S < Sf (t), P (S, t) = E − S. We plug it into the
Black-Schole equation
σ 2 2 p2 P

∂P ∂P
+ S +r S − P = −rE < 0.
∂t 2 ∂S 2 ∂S
Thus, the B-S inequality holds.
6.4 *Penalty method for linear complementary problem

We can reformulate this problem in terms of x variable. As before, we use the following change of
variables: V = Ev, S = Eex , τ = T − t. The free boundary now in x-variable is xf (t). The free
boundary value problem is formulated as
(i) for −∞ < x < xf (t), v = 1 − ex , and
σ2 σ2
vτ − vxx − (r − )vx + rv ≥ 0.
2 2
(ii) for xf (t) < x < ∞, v > 1 − ex , and
σ2 σ2
vτ − vxx − (r − )vx + rv = 0,
2 2
∂v
(iii) both v and ∂x are continuous.
The linear complementary problem is formulated as:
(i) v − (1 − ex ) ≥ 0,
σ2 σ2
(ii) vτ − 2 vxx − (r − 2 )vx + rv ≥ 0,
σ2 σ2
(iii) (vτ − 2 vxx − (r − 2 )vx + rv)(v − (1 − ex )) = 0,
(iv) v and vx are continuous.
with initial condition v(x, 0) = Λ(x) = 1 − ex .

We may replace v by ue−rτ to eliminate the term rv. Then we have
σ2 σ2

uτ − uxx − (r − )ux (u − g) = 0
2 2
σ2 σ2
uτ − uxx − (r − )ux ≥ 0
2 2
u − g ≥ 0,
u(x, 0) = g(x, 0),
with u, ux being continuous. Here g(x, τ ) = max{erτ (1 − ex ), 0}. The far field boundary condi-
tions are
u(x, τ ) → 0, as x → ∞, u(x, τ ) → erτ , as x → −∞.
A mathematical theory called parabolic variational inequality gives construction, existence, unique-
ness of the solution. (see reference: A. Friedman,Variational Inequality). Let us demonstrate this
theory briefly. The method we shall use is called the penalty method. Let us consider the following
penalty function:
φN (v) = −e−N v .
It has the properties: (i) φN > 0, (ii) φN (v) → 0 whenever v > 0.. We consider the following
penalized P.D.E.:
1 σ2
uτ − σ 2 uxx − (r − )ux + φN (u − g) = 0
2 2
u(x, 0) = g(x, 0),
with u → 0 as x → +∞, u → erτ as x → −∞.

From a standard theory of nonlinear P.D.E. (by monotone method, for instance), one can show
that the solution uN exists for all N > 0. We then need an estimate for uN and ∂u N
∂x . The bounded-
ness of these two gives that uN has a convergent subsequence, say uNi such that
uNi → u,
∂
with uNi , u, ∂x uNi , ux being continuous. Moreover,
|φNi (uNi − g)| ≤ constant
As Ni → ∞, we conclude u − g ≥ 0. Further, on the set {u − g > 0}, φNi (uNi − g) → 0. Hence

we have
1 σ2
uτ − uxx − (r − )ux = 0 on {u − g > 0}.
2 2
6.5 Numerical Methods

6.5.1 Binomial method for American puts
We recall that the solution of the Black-Scholes equation can be discretized by the following bino-
mial method (or the forward Euler method):
√
• Choose a large N , define ∆t = T /N , ∆x = σ ∆t, Sjn = S0 ej∆x , |j| ≤ n, j + n = even.
• Define
n+1 n+1
er∆t Vjn = pVj+1 + qVj−1 ,
where
er∆t − u √
p= , u = eσ ∆ , d = 1/u, q = 1 − p.
u−d
6.5. NUMERICAL METHODS 97
Similarly, the linear complementary problem can be discretized as

n+1 n+1
er∆t Vjn − pVj+1 Vjn − Λnj

− qVj−1 = 0,
n+1 n+1
er∆t Vjn − pVj+1 − qVj−1 ≥ 0,
Vjn − Λnj ≥ 0
Here, Λnj := E − Sjn . This recursive procedure finding V n from V n+1 for n = N − 1, N − 2, ..., 0
with initial condition
VjN = ΛN j .
This discretized linear complementary problem can be solved by the following projected forward
Euler method:
Vjn = max{e−r∆t (pVj+1
n+1 n+1
+ qVj−1 ), Λnj }. (6.3)
Here, Λnj := Λ(Sjn ). This is called projected binomial method.

Theorem 6.9. The projected binomial method for American option is stable and converges. Namely,
Vjn → V (S, t) with Sj := S0 ej∆x → S and n∆t → t as n → ∞. Furthermore, V (S, t) satisfies
the linear complementary equations (6.2).
Sketch the idea of the proof.

1. One can show that this method is stable and converges in Cb1 (R+ ), the space of C 1 -continuous
and bounded functions, by using the monotone operator theory. (Ref. Majda & Crandell,
Math. Comp. ) An operator Φ : Cb (R) → Cb (R) is called monotone operator if f ≥ g
implies Φ(f ) ≥ Φ(g). For continuous case, the solution operator Φ(t) of the Black-Scholes
equation is a monotone operator. Their discrete version, the one-step solution operator which
maps V n+1 7→ V n is also a monotone operator. The projection operator is a monotone
operator, which can be proven easily by the fact that max(a, c) ≤ max(b, c) if a ≤ b.
n+1 n+1
2. Furthermore, from the projection Vjn = max{e−r∆t (pVj+1 + qVj−1 ), Λnj }, we obtain
Vjn ≥ Λnj

Vjn ≥ e−r∆t pVj+1
n+1 n+1
+ qVj−1 .
Thus the limiting function V (S, t) also satisfies V (S, t) ≥ Λ(S, t) and the Black-Scholes
inequality.
n+1 n+1
3. From construction, the projection: Vjn = max{e−r∆t (pVj+1 + qVj−1 ), Λnj } satisfies either
n+1 n+1
Vjn − Λnj = 0, or Vjn − e−r∆t (pVj+1 + qVj−1 ) = 0. Thus, it satisfies the complementary
condition:
n+1 n+1
er∆t Vjn − pVj+1 Vjn − Λnj = 0.

− qVj−1
From convergence result, their limiting function also satisfies the continuous complementary
condition.
Below is a C code for binomial method for American put.

Binomial_American_option(V,S,S0,E,r,sigma,dt,N)
{
double u,d,p;
double discount, tmp;
int m,n;
discount = exp(-r*dt);
u = exp(sigma*sqrt(dt));
d = 1/u;
p = (1/discount - d)/(u-d);
//Binomial tree for asset price: at level n.

S[0][0]=S0;
for (n = 1; n <= N; ++n)
{
for (m=n; m > 0; --m)
S[n][m] = u*S[n-1][m-1];
S[n][0] = d*S[n-1][0];
}
// Binomial tree for option.

for (m=0; m <=N; ++m)
V[N][m] = payoff_put(S[N][m],E);
for ( n = N-1; n >= 0; --n)
{
for (m=0; m < n; ++m)
{
tmp = p*V[n+1][m+1] + (1-p)*V[n+1][m];
tmp *= discount;
V[n][m] = max(tmp,payoff_put(S[n][m],E));
}
}
}
double payoff_put(S,E)
{
return max(E-S,0);
}
6.5.2 Binomial method for American call on dividend-paying asset

The linear complementary problem for this American call option is
(i) V (S, t) ≥ Λ(S) ≡ max{S − E, 0}

2 2
(ii) Vt + σ2 S 2 ∂∂SV2 + (r − D0 )S ∂V
∂S − rV ≤ 0,

σ2 2 ∂ 2 V ∂V
(iii) Vt + 2 S ∂S 2 + (r − D0 )S ∂S − rV (V − Λ) = 0.
(iv) V and VS are continuous.
The binomial approximation for the B-S equation is

n+1 n+1
er∆t Vjn = pVj+1 + qVj−1 ,
6.5. NUMERICAL METHODS 99
where
σ 2 ∆t σ 2 ∆t σ 2 ∆t σ 2 ∆t
p= + (r − D 0 − ) , q= − (r − D 0 − ) ,
2 (∆x)2 2 2∆x 2 (∆x)2 2 2∆x
and p + q = 1. We choose ∆t and ∆x so that p > 0 and q > 0. For American option, V has to be
greater than Λ(S, t), the payoff function at time t. Hence, we should require

n+1 n+1
+ qVj−1 ), Λnj }.
The above is the projective forward Euler method.

For the corresponding binomial model, first we determine the up/down ratios u and d for the
riskless asset price by
2 )∆t
pu + (1 − p)d = e(r−D0 )∆t , pu2 + (1 − p)d2 = e(2(r−D0 )+σ ,
or equivalently,
p
u = A + A2 − 1, d = 1/u,
1 −(r−D0 )∆t 2
A = (e + e(r−D0 +σ )∆t ),
2
e(r−D0 )∆t − d
p = , iq = 1 − p.
u−d
Then the binomial model is given by
n−1
n uSj−1 , with probability p
Sj = n−1
dSj+1 , with probability q
S00 = S.
and
n+1 n+1
+ qVj−1 ), Sjn − E}.
6.5.3 *Implicit method

For implicit method like backward Euler or Crank-Nicolson method, we need to add the constraints
un+1 ≥ g n+1 for American option. It is important to know that if an iterative method is used, then
we should require this condition hold in each iteration steps. For instance, in the SOR iteration
method,
V n,(k+1) = max{V n,(k) + ω(y n,(k+1) − V n,(k) ), Λn },

y n,(k+1) = (D + L)−1 (−U V n,(k) + f n+1 ), k = 0, · · · K
V n ≡ V n,(K)
This guarantees that V n+1 ≥ Λn+1 .

6.6 Converting American option to a fixed domain problem

6.6.1 American call option with dividend paying asset
We consider the American call option on a dividend paying asset:
σ2 2 ∂ 2V ∂V
Vt + S + (r − D0 )S − rV = 0,
2 ∂S 2 ∂S
V (S, t) ≥ Λ(S) ≡ max{S − E, 0},
0 ≤ S ≤ Sf (t), 0 ≤ t ≤ T.
where Sf (t) is the free boundary. On this free boundary, the boundary condition is required
V (Sf (t), t) = Sf (t) − E,

∂V
(Sf (t), t) = 1, 0 ≤ t ≤ T.
∂S
We also need a condition for Sf at final time
Sf (T ) = max(E, rE/D0 )
Before converting the problem, we first remove the singularity of the final data (i.e. non-smoothness
of the payoff function) as the follows. We may substract V by an European call option c with the
same payoff data. Notice that c(S, t) has exact solution. The new variable V − c satisfies the same
equation, yet it has smooth final data.
To convert the free boundary problem to a fixed domain problem, we introduce the following
change-of-variables: 

 ξ = S/Sf (t)
τ = T −t


 u(ξ, τ ) = (V (S, t) − c(S, t))/E
sf (τ ) = Sf (t)/E

The new equations for these new variables are

σ2 2 ∂ 2 u 1 dsf

∂u
 ∂τ
 = 2 ξ ∂ξ 2 + (r − D0 ) + sf dτ ξ ∂u
∂ξ − ru, 0 ≤ ξ ≤ 1,

0≤τ ≤ T,





u(ξ, 0) = 0, 0≤ξ ≤ 1, (6.4)

 u(1, τ ) = g(sf (τ ), τ ), 0≤τ ≤T
∂u

(1, τ ) = h(sf (τ ), τ ), 0≤τ ≤T


 ∂ξ


sf (0) = max(1, r/D0 ),
where
g(sf (τ ), τ ) = sf (τ ) − 1 − c(Esf (τ ), T − τ ),

∂c
h(sf (τ ), τ ) = sf (τ ) 1 − (Esf (τ ), T − τ ) .
∂S
6.6. CONVERTING AMERICAN OPTION TO A FIXED DOMAIN PROBLEM 101
At the boundary ξ = 0, the Black-Sholes equation is degenerate to
∂u
= −ru.
∂τ
With the trivial initial condition yields
u(0, τ ) = 0, 0 ≤ τ ≤ T.
In practice, we can solve the modified Black-Sholes equation (6.4) with the boundary conditions

u(0, τ ) = 0 0≤τ ≤T
∂u (6.5)
∂ξ (1, τ ) = h(s f (τ ), τ ), 0≤τ ≤T
We can differentiate the other boundary condition in τ and yield an ODE for the free boundary:
∂u ∂g dsf
(1, τ ) = .
∂τ ∂ξ dτ
with sf (0) = max(1, r/D0 ).
6.6.2 American put option

For American put option P , the B-S equation is on the infinite domain S > Sf (t), 0 ≤ t ≤ T .
Through the change-of-variable ( 2
η = ES
u(η, t) = EP S(S,t)
the infinity domain problem is converted to a finite domain problem:
b 2 2 ∂2u

∂u ∂u
∂t + f σ η ∂η 2 − rη ∂η = 0, 0 ≤ η ≤ ηf (t),



0≤t≤T




u(η, T ) = max(η − E, 0), 0 ≤ η ≤ ηf (T ),


 u(η f (t), t) = η f (t) − E, 0 ≤ t ≤ T,
∂u
0 ≤ t ≤ T,

 (ηf (t), t) = 1,
 ∂η



ηf (T ) = max(E, 0)
We can further convert it to a fixed domain problem as that in the last section.
Chapter 7
Exotic Options
Option with more complicated payoff then the standard European or American calls and puts are
called exotic options. They are usually traded over the counter. Their prices are usually not quoted
on an exchange. We list some common exotic options below.
1. Binary options
2. compound options
3. chooser options
4. barrier options
5. Asian options
6. Lookback options
In the last two, the payoff depends on the history of the asset prices, for instance, the averages,
the maximum, etc., we shall call these kinds of options, the path-dependent options, and will be
discussed in the next Chapter.
7.1 Binaries
The payoff function Λ(S) is an arbitrary function. One particular binary option is the cash-or-
nothing call, whose payoff is
Λ(S) = BH(S − E).
This option can be interpreted as a simple bet on an asset price: if S > E at expiry the payoff is B,
otherwise zero. We have seen its value is
Z ∞
−r(T −t)
V =e P(S 0 , T, S, t)Λ(S 0 )dS 0 = e−r(T −t) BN (d2 ),
0
103
104 CHAPTER 7. EXOTIC OPTIONS
where 0 2
log( S )−(r− σ2 )(T −t)
0 1 − S
2
P(S , T, S, t) = p e 2σ (T −t)
2πσ 2 (T − t)
is the transition probability density for asset price in risk-neutral world.
7.2 Compounds
A compound option may be described as an option on an option. We consider the case where
the underlying option is a vanilla put or call and the compound option is vanilla put or call on
the underlying option. The extension to more complicated option on more complicated option is
relatively straightforward. There are four different classes of basic compound options:
1. call-on-call,
2. call-on-put,
3. put-on-call,
4. put-on-put.
Let us investigate the case call-on-call. Other cases can be treated similarly. The underlying option
is
Expiry : T2 , Strike price : E2 .
The compound option on this option
Expiry : T1 < T2 , Strike price : E1 .
The underlying option has value C(S, t, T2 , E2 ). At time T1 , its value C(S, T1 , T2 , E2 ). The payoff
for the compound call option is max{C(S, T1 , T2 , E2 ) − E1 , 0}. Because the compound options
value is governed only by the randomness of S, according to the Black-Scholes analysis, it also must
satisfy the same Black-Scholes equation. We then solve the Black-Scholes equation with payoff
max{C(S, T1 , T2 , E2 ) − E1 , 0}.
7.3 Chooser options

A regular chooser option gives its owner the right to purchase, for an amount E1 at time T1 , either
a call or a put with exercise price E2 at time T2 . Thus, it is a “call on a call or put”. Certainly, we
have T1 < T2 . The payoff at T1 for this call-on-a-call-or-put” is
Λ = max{C(S, T1 ) − E1 , P (S, T1 ) − E1 , 0}.
The compound option also satisfies the Black-Scholes equation for the same reason as above. From
this and payoff function at T1 , we can value V at t. The contract can be made more general by
having the underlying call and put with different exercise prices and expiry dates, or by allowing
the right to sell the vanilla put or call. By using the Black-Scholes formula for vanilla option, there
is no difficulty to value these complex chooser options.
7.4. BARRIER OPTION 105
7.4 Barrier option

Barrier options differ from vanilla options in that part of the option contract is triggered if the asset
price hits some barrier, S = X, say at some time prior to T . As well as being either calls or puts,
barrier options are categorized as follows.
1. up-and-in: the option expires worthless unless S reaches X from below before expiry.
2. down-and-in: the option expires worthless unless S reaches X from above before expiry.
3. up-and-out: the option expires worthless if S reaches X from below before expiry.
4. down-and-out: the option expires worthless if S reaches X from above before expiry.
7.4.1 down-and-out call(knockout)

A European option whose value becomes zero if S ever goes as low as S = X. Sometimes in the
knockout options, one can have boundary of time, or one can have rebate if the barrier is crossed.
In the latter case, the option holder receives a specific amount Z for compensation.
Let us consider the case of a European style down-and-out option without relate. We assume
X < E. The boundary conditions are
V (X, t) = 0, (boundary condition), V (S, t) ∼ S as S → ∞.
The final condition, V (S, T ) = max{S − E, 0}. For S > X, the option becomes a vanilla call, it
satisfies the Black-Scholes equation.
Let us find its explicit solution. Let S = Eex , t = T − (σ2τ/2) , V = Ev. The Black-Scholes
equation is transformed into
vτ = vxx + (k − 1)vx − kv,
r
where k = σ 2 /2
. We make another change of variable :
v = eαx+βτ u.
We choose α, β to eliminate the lower order terms in the derivatives of x:
βeαx+βτ u + eαx+βτ uτ = α2 eαx+βτ u + 2αeαx+βτ ux + eαx+βτ uxx

+(k − 1)(αeαx+βτ u + eαx+βτ ux ) − keαx+βτ u.
This implies that α = − 21 (k − 1) and β = − 14 (k + 1)2 and equation becomes uτ = uxx .

Let x0 = log( Ex ), or X = Eex0 . The boundary condition becomes u(x0 , τ ) = 0 and u(x, τ ) ∼
e(1−α)x−βτ as x → ∞. The initial condition becomes
1 1
u(x, 0) = u0 (x) = max{e 2 (k+1)x − e 2 (k−1)x , 0}.
This follows from the payoff function being

S
Λ = max{S − E, 0} = E max{ − 1, 0} = E max{ex − 1, 0},
E
and V = eαx+βτ u(x, T − (σ2τ/2) ), with u(x, 0) = u0 (x) = e−αx max{ex −1, 0} = max{e(−α+1)x −
e−αx , 0}. Notice that because X < E, we have x0 < 0, and

0 for x0 < x < 0,
u0 (x) = (−α+1)x −αx
max{e −e , 0} otherwise.
We use method-of-reflection to solve above heat equation with zero boundary condition. We reflect
the initial condition about x0 as

u0 (x), for x0 < x < ∞
u(x, 0) =
−u0 (2x0 − x), for − ∞ < x < x0 .
The equation and the initial condition are unchanged under the change-of-variable: x → 2x0 − x,
u → −u. From the uniqueness of the solution, the solution has the property:
u(2x0 − x, t) = −u(x, t).
From this, we can obtain that u(x0 , t) = −u(x0 , t) = 0.

Since C = Eeαx+βτ u1 is the vanilla call, where u1 satisfies the heat equation with the initial
condition: 1 (k+1)x 1
e2 − e 2 (k−1)x for x > 0
u1 (x, 0) =
0, for x ≤ 0
Using this and the method of reflection, we may express V in terms of C as the follows. First, we
may write V = Eeαx+βτ (u1 + u2 ), where the initial condition for u2 is the reflected condition from
u1 :
u0 (2x0 − x, 0) for x ≤ 0
u2 (x, 0) =
0 for x > 0
The solution u1 corresponds to C(S, t). The solution u2 is corresponds to
e2α(x−x0 ) C(x2 /S, t). We conclude
S −(k−1)
V = C(S, t) − ( ) C(X 2 /S, t).
X
7.4.2 down-and-in(knock-in) option

An “in” option becomes worthless unless the asset price reaches the barrier before expiry. If S
crosses the line S = X at some time prior to expiry, then the option becomes a vanilla option. It is
common for in-type barrier option to give a rebate, usually a fixed amount, if the barrier is not hit.
This compensates the holder for the loss of the option.
The boundary condition for an “in” option is the follows. The option is worthless as S → ∞,
i.e., V (S, t) → 0 as S → ∞. At T , if S > X, then V (S, T ) = 0. For t < T , V (X, t) = C(X, t).
Since the option immediately turns into a vanilla call and must have the same value of this vanilla
call. For S ≤ X, V (S, t) = C(X, t). We only need to solve V for S > X. V still satisfies the same
Black-Scholes equation for all S, t, because its randomness is fully correlated to the randomness of
S.
7.5. ASIAN OPTIONS AND LOOKBACK OPTIONS 107
We may write V = c − V , where c is the value of a vanilla call. Then the boundary condition
for V is V (S, t) = c − V ∼ S − 0 = S as S → ∞. And
V (X, t) = c(X, t) − V (X, t) = 0, V (S, T ) = c(S, T ) − V (S, T ) = c(S, T ) = Λ(S).
We observe that V is indeed a “down-and-out” barrier option. In other words, 1(down-and-in) plus
1(down-and-out) equal to 1 vanilla call. This is because one and only one of the two barrier options
can be active at expiry and whichever it is, its value is the value of a vanilla call.
7.5 Asian options and lookback options

In Asian options and lookback options, their payoff functions depend on the history of the underly-
ing asset. For example,
1. a European-type average strike option has the following payoff function

Z T
1
max{ST − S(τ )dτ, 0}.
T 0
2. an American-type average strike option,

Z t
1
Λ(S, t) = max{S − S(τ )dτ, 0}.
t 0
3. geometric mean, RT
log S(τ )dτ
Λ(S, T ) = max{S − e 0 , 0}.
4. Lookback call,
Λ(S, T ) = max{S − J, 0}, J = max S(τ )
0≤τ ≤T
In general, the payoff depends on I, which is defined by

Z t
I= f (S(τ ), τ )dτ,
0
where f is a smooth function. The payoff function is Λ(S, I). It is important to notice that I(t) is
independent of S(t). The value of an asian option should depend on S, t s well as I. Indeed, we
shall see in the next chapter that dI = f dt. The only randomness is through S, therefore V can be
valued through a delta hedge.
For the lookback option, it will be treated as a limiting case of an asian option. We shall discuss
this in the next chapter.
Chapter 8
Path-Dependent Options
8.1 Introduction
If the payoff depends on the history of the underlying asset price, such an option is called a path-
dependent option. The Asian options and the Russian options (Lookback options) are the typical
examples. The payoff functions for these options are, for example,
RT
1. average strike call option: Λ = max{S − T1 0 S(τ )dτ, 0},
RT
2. average rate call option: Λ = max{ T1 0 S(τ )dτ − E, 0}
RT RT
1 log(S(τ ))dτ
3. geometric mean: the arithmetic mean T 0 S(τ )dτ above is replaced by e 0 .
4. lookback strike put: Λ = max{max0≤τ ≤T S(τ ) − S, 0}.
5. lookback rate put: Λ = max{E − max0≤τ ≤T S(τ ), 0}.
8.2 General Method

Let f be a smooth function, define
Z t
I(t) = f (S(τ ), τ )dτ.
0
In previous examples, f (S(τ ), τ ) = S(τ ) for arithmetic mean and f (S(τ ), τ ) = log S(τ ) for
geometric mean.
Notice that I(t) is a random variable and is independent of S(t). (This is because I(t) is the sum
of increment of functions of S before time t, and each increment of S(τ ), τ < t, is independent
of S(t).) Therefore, we should introduce another independent variable I besides S to value the
derivative V (S, I, t).
109
110 CHAPTER 8. PATH-DEPENDENT OPTIONS
The stochastic differential equations governed by S and I are

dS
= µdt + σdz,
S
dI(t) = f (S(t), t)dt.
Notice that there is no noize term in dI. The only randomness is from dS. Therefore, we can use
delta hedge to eliminate this randomness. Namely, we consider the portfolio
Π = V − ∆S,
as before. We have
1 ∂2S
dΠ = dV − ∆dS = (Vt + σ 2 S 2 )dt + VI dI + VS dS − ∆dS.
2 ∂V 2
∂V
We choose ∆ = ∂S to eliminate the randomness in dΠ. From the arbitrage assumption, we arrive
1 ∂2V ∂V ∂V
Vt + σ 2 S 2 2 + f = r(V − S ).
2 ∂S ∂I ∂S
8.3 Average strike options

8.3.1 European calls
Let us consider an average strike call option with European exercise feature. Its payoff function is
defined by
1 T
Z
max{S − S(τ )dτ, 0).
T 0
Or, in terms of I, Λ(S, I, T ) = max{S − TI , 0}.
We notice that the modified Black-Scholes equation
∂V σ2 ∂ 2V ∂V
Vt + S + S 2 2 + rS − rV = 0,
∂I 2 ∂S ∂S
and the initial data for the average strike options are invariant under the transformation: (S, I) →
λ(S, I). Therefore, we expect that its solution is a function of the scale-invariant variable R = I/S.
Notice that this is also reflected in that
dR = (1 + (σ 2 − µ)R)dt − σRdz,
depends on R only. Since the initial data can be expressed as Λ = S max{1 − R/T, 0}, we may
also expect that V = SH(R, t). This reduces one independent variable.
Plug V = SH into the above modified Black-Scholes equation:
1 ∂H ∂2H ∂H ∂
SHt + σ 2 S 2 (2 +S )+S·S + rS (SH) − r(SH) = 0.
2 ∂S ∂S 2 ∂I ∂S
8.3. AVERAGE STRIKE OPTIONS 111
From
∂ ∂R ∂ R ∂
= , =−
∂S ∂S ∂R S ∂R
∂2 2I ∂ I 2 ∂2 1 ∂ 2 ∂
2
= + = (2R + R ),
∂S 2 S 3 ∂R S 4 ∂R2 S2 ∂R ∂R2
we obtain
1 R ∂H 1 ∂H ∂2H
SHt + σ 2 S 2 (2(− ) + S · 2 (2R + R2 ))
2 S ∂R S ∂R ∂R2
1 R
+S 2 HR + rSH + rS 2 (− HR − rSH) = 0
S S
Finally, we arrive
σ2 2 ∂ 2H ∂H
Ht + R + (1 − rR) = 0.
2 ∂R2 ∂R
The payoff function
R
Λ(R, T ) = max{1 − , 0} = H(R, T ), (final condition) .
T
we should require the boundary conditions.
• H(∞, t) = 0. Since as R → ∞ implies S → 0, then V → 0, then H → 0.
• H(0, t) is finite. When R = 0, we have S(τ ) = 0, for all τ with probability 1. That implies
that Λ = 0, and consequently, H is finte at (0, t).
Next, we expect that the solution is smooth up to R =. This implies that HR , HRR are finite at
2
(0, t). We have the following two cases: (i) If R2 ∂∂RH2 = O(1) 6= 0 as R → 0 then H = O(log R)
for R → 0. Or (ii) if R ∂H∂R = O(1) 6= 0 as R → 0, then H = O(log R). Both cases contradict
to H(0, t) being finite. Hence, we have RHR (0, t), R2 HRR (0, t) are zeros as R → 0. Hence the
boundary condition H(0, t) is finite is equivalent to Ht (0, t) + HR (0, t) = 0.
This equation with boundary condition can be solved by using the hypergeometric functions.
However, in practice, we solve it by numerical method.
8.3.2 American call options

We consider the average strike call option with American exercise feature. In this case,
1 ∂2H ∂H
Ht + σ 2 R 2 + (1 − rR) ≤ 0
2 ∂R2 ∂R
H −Λ ≥ 0
1 2
∂ H ∂H
(Ht + σ 2 R2 2
+ (1 − rR) )(H − Λ) = 0,
2 ∂R ∂R
Rt
where Λ(R, t) = max{1 − Rt , 0}, R(t) = I(t)/S(t), I(t) = 0 S(τ )dτ.
8.3.3 Put-call parity for average strike option

We study the put-call parity for average strike options with European exercise feature. Consider a
portfolio is C − P . The corresponding payoff function is
R R R
S max{1 − , 0} − S max{ − 1, 0} = S(1 − ) ≡ S · H(R, T ).
T T T
Since the Black-Scholes equation in linear in R, we only need to solve the equation with final
condition (i) H(R, T ) = 1, (ii) H(R, T ) = − RT . For (i), H(R, t) ≡ 1. For (ii), since the final
condition and the P.D.E. is linear in R, we expect that the solution is also linear in R. Thus, we
consider H is of the following form
H(t, R) = a(t) + b(t)R
Plug this into the equation, we obtain

d d
a + bR + (1 − rR)b = 0.
dt dt
and
d d
a + b = 0, b − rb = 0.
dt dt
with a(T ) = 0, b(T ) = − T1 . This differential equation can be solved easily:
1 1
b(t) = − e−r(T −t) , a(t) = − (1 − e−r(T −t) ).
T rT
Consequently, we obtain the put-call parity:
1 −r(T −t) 1 t
Z
1 −r(T −t)
C −P = S(1 − (1 − e − e S(τ )dτ )
rT T S 0
1 t
Z
S
= S− (1 − e−r(T −t) ) − e−r(T −t) S(τ )dτ
rT T 0
8.4 Lookback Option

A lookback option is a derivate product whose payoff depends on the maximum or minimum of
its underlying asset price. For instance, the payoff function for a lookback option with European
exercise feature is
Λ = max S(τ ) − S(T ).
0≤τ ≤T
Such an option is relatively expansive because it gives the holder an extremely advantageous payoff.
As before, let us introduce J(t) = max0≤τ <t S(τ ). Since S(τ ), τ < t are independent of S(t),
we see that J(t) is independent of S(t). This suggest that we should introduce another independent
variable J to value the lookback option in addition to S and t. We can derive a stochastic differential
8.4. LOOKBACK OPTION 113
equation for J as before. Indeed, it is dJ = 0. However, we shall give a more careful approach. We
shall use the fact that for a continuous function S(·),
Z t n1
n
max |S(τ )| = lim |S(τ )| dτ .
0≤τ ≤t n→∞ 0
We leave its proof as an exercise.

Remark. For the minimum, we have
Z t n1
n
lim (S(τ )) dτ = min S(τ ).
n→−∞ 0 0≤τ ≤t
Let us introduce
Z t 1
In = (S(τ ))n dτ, Jn = Inn .
0
The s.d.e. for Jn ,

Z t+dt 1
Z t 1
dJn = ( (S(τ ))n dτ ) n − ( (S(τ ))n dτ ) n
0 0
1 S(t)n
= dt.
n Jnn−1
Now, as before we consider the delta hedge:
Π = P − ∆S.
From the arbitrage assumption, we can derive the equation for P (S, Jn , t):
1 S n ∂P 1 2 2 ∂2P
dΠ = Pt dt + dt + σ S dt
n Jnn−1 ∂Jn 2 ∂S 2
∂P
= r(P − S)dt
∂S
S
Taking n → ∞, using the facts that Jn → J and J ≤ 1, we arrive
1 ∂2P ∂P
Pt + σ 2 S 2 2 + rS − rP = 0.
2 ∂S ∂S
This is the usual Black-Scholes equation. The role of J here is only a parameter. This is consistent
to the fact that
dJ = 0.
8.4.1 A lookback put with European exercise feature

The range for S is 0 ≤ S ≤ J. This is because S ≤ J, for 0 ≤ t ≤ T . We claim that
P (0, J, t) = Je−r(T −t) .
Firstly, we have that Λ(0, J, T ) = max{J − S, 0} = J. Secondly, if S(t) = 0, then S(τ ) = 0
for t ≤ τ ≤ T . The asset price process becomes deterministic. Therefore, the value of P is the
discounted payoff: P (0, J, t) = Je−r(T −t) .
Next, we claim that
∂P
(J, J, t) = 0.
∂J
From µ > 0, the current maximum cannot be the final maximum with probability 1. The value of
P must be insensitive to a small change of J.
We can use method of image to solve this problem. Its solution is given by
S
P = S(−1 + N (d7 )(1 + k −1 )) + Je−r(T −t) N (d5 ) − k −1 ( )1−k N (d6 ),
J
where
J σ2 √
d5 = [ln( ) − (r − )(T − t)]/σ T − t
S 2
S σ2 √
d6 = [ln( ) − (r − )(T − t)]/σ T − t
J 2
J σ2 √
d7 = [ln( ) − (r + )(T − t)]/σ T − t
S 2
r
k =
σ 2 /2
8.4.2 Lookback put option with American exercise feature

We have the following linear complementary equation,
LBS P ≤ 0, P − Λ ≥ 0, (LBS P )(P − Λ) = 0,
where
∂ σ2S 2 ∂ 2 ∂
LBS = + 2
+ rS − r.
∂t 2 ∂S ∂S
The final condition
P (S, J, T ) = Λ(S, J, T ) = J − S.
The boundary condition
∂P
(J, J, t) = 0.
∂J
We require P , ∂P ∂P
∂S ∂J are continuous.
For lookback call option, we simply replace max0≤τ ≤t S(τ ) by min0≤τ ≤t S(τ ). For instance,
its payoff is Λ = S(T ) − min0≤τ ≤T S(τ ).
Chapter 9
Bonds and Interest Rate Derivatives
9.1 Bond Models

A bond is a long-term contract under which the issuer promises to pay the bondholder coupon
payment (usually periodically) and principal (at the maturity dates). If there is no coupon payment,
the bond is called a zero-coupon bond. The principal of a bond is called its face value.
9.1.1 Deterministic bond model

The value of a bond certainly depends on the interest rate. Let us first assume that the interest rate is
deterministic temporarily, say r(τ ), t ≤ τ ≤ T , is known. Let B(t, T ) be the bond value at t with
maturity date T , k(t) be its coupon rate. This means that in a small dt, the holder receives coupon
payment k(t) dt. From the no-arbitrage argument,
dB + k(t) dt = r(t)Bdt.
together with the final condition:
B(T, T ) = F ( face value),
the bond value can be solved and has the following expression:
Z T
− tT r(τ ) dτ
R RT
r(s) ds
B(t, T ) = e F+ k(τ )e τ dτ .
t
Thus, the bond value is the sum of the present face value and the coupon stream.
However, the life span of a bond is long (usually 10 years or longer), it is unrealistic to assume
that the interest rate is deterministic. In the next subsection, we shall provide a stochastic model.
9.1.2 Stochastic bond model

Let us assume that the interest rate satisfies the following s.d.e.:
dr = u(r, t)dt + w(r, t)dz,
115
116 CHAPTER 9. BONDS AND INTEREST RATE DERIVATIVES
where dz is the standard Wiener process. The drift u and the variance w2 are proposed by many
researchers. We shall discuss these issues later. To find the equation for B with stochastic property
of r, we consider a portfolio containing bonds with different maturity dates:
Π = V (t, r, T1 ) − ∆V (t, r, T2 ) ≡ V1 − ∆V2 .
The change dΠ in a small time step dt is
dΠ = dV1 − ∆dV2 ,
where
dVi
= µi dt + σi dz,
Vi
1
µi = Vi,t + uVi,r + fbw2 Vi,rr
Vi
1
σi = wVi,r .
Vi
We we choose ∆ = V1,r /V2,r , then the random term is canceled in dΠ. From the no-arbitrage
argument, dΠ = rΠdt. We obtain
µ1 V1 dt − ∆µ2 V2 dt = r(V1 − ∆V2 ) dt.
This yields
(µ1 − r)V1 /V1,r = (µ2 − r)V2 /V2,r ,
or equivalently
µ1 − r µ2 − r
= .
σ1 σ2
Since the left-hand side is a function of T1 , while the right-hand side is a function of T2 . Therefore,
it is independent of T . Let us express it as a known function λ(r, t):
µ−r
= λ(r, t).
σ
Plug µi and σi back to this equation, and drop the index i, we obtain
1
Vt + w2 Vrr + (u − λw)Vr − rV = 0.
2
The function λ(r, t) = µ−rσ is called the market price, since it gives the extra increase in expected
instantaneous rate of return on a bond per an additional unit of risk.
This stochastic bond model depends on three parameter functions u(r, t), w(r, t) and λ(r, t). In
the next section, we shall provide some model to determine them.
9.2. INTEREST MODELS 117
9.2 Interest models

There are many interest rate models. We list some of them below.
• Merton (1973): dr = αdt + σdz.
• Vasicek (1977): dr = β(α − r)dt + σdz.
• Dothan (1978): dr = σrdz.
• Marsh-Rosenfeld (1983): dr = (αrδ−1 + βr)dt + σrδ/2 dz.
• Cox-Ingersoll-Ross (1985) dr = β(α − r)dt + σrf dz.

b
• Ho-Lee (1986): dr = α(t)dt + σdz.
• Black-Karasinski (1991): d ln r = (a(t) − b(t) ln r) dt + σdz.
The main requirements for an interest rate model are
• positivity: r(t) ≥ 0 almost surely,
• mean reversion: r should tends to increase (or to decrease) and toward a mean.
The C-I-R and B-K models have these properties.

Below, we shall illustrate a unified approach proposed by Luo, Yen and Zhang.
9.2.1 A functional approach for interest rate model

The idea is to design r to be a function of x(t) and t, (i.e. r = r(x(t), t)) with x(t) governed by
a simple stochastic process. We notice that the Ornstein-Uhlenbeck process dx = −ηxdt + σdz
has the property to tend to its mean (which is 0) time asymptotically. While the Bessel process
dx = /xdt + σdz has positive property. We then design the underlying basic process is the sum of
these two processes:
dx = −ηx + dt + σdz.
x
In general, we allow η, , and σ are given functions of t. With this simple process, we can choose
r = r(x(t), t). Then all interest models mentioned above correspond to different choices of r(x, t),
η, and .
• Merton’s model: we choose = η = 0, r = x + αt.

2
• Dothan’s model: = η = 0, r = ex−σ /2t .
Rt
• Ho-Lee: = η = 0, r = x + 0 α(s) ds.
• Vasicek: = 0, η = β, r = x + α.
• C-I-R: β = 2η, α = (σ 2 + 2)/(8η), r = 41 x2 .

Rt
• Black-Karasinski: = 0, r = exp(g(t, x)), where g = x + 0 a(s) ds, η = b(t).
With the interest rate model, the zero-coupon bond price V is given by
RT
V (x, t) = E e t r(s,x(s)) ds | xt = x , t < T.
From the Feymann-Kac formula, V satisfies

1 ∂2

∂ ∂
+ + −ηx + − r V = 0.
∂t σ 2 ∂x2 x ∂x
This model depends three parameter functions (t), η(t), σ(t), and r(x, t). There is no unified
theory available yet with this approach and the approach of the previous subsection.
9.3 Convertible Bonds

A convertible bond is a bond plus a call option under which the bond holder has the right to convert
the bond into a common shares. Thus, it is a function of r, S, t and T . Let the stochastic processes
governed by S and r are
dS
= µdt + σdzS ,
S
dr = udt + wdzr .
Suppose the correlation between dzS and dzR is
dzS dzr = ρ(S, r, t)dt.
The final value of the convertible bond V (r, S, T ) = F , the face value of the bond. Suppose the
bond can be converted to nS at any time priori to T . Then we have
V (r, S, t) ≥ nS.
We also have the boundary conditions:
lim V (r, S, t) = nS
S→∞
lim V (r, S, t) = 0.
r→∞
At S = 0 or r = 0, we should require V (r, 0, t) or V (0, S, t) to be finite.

The Black-Scholes analysis for a convertible bond is similar to the analysis for a bond. Let
Vi = V (r, S, t, Ti ), i = 1, 2. Consider a portfolio
Π = ∆1 V1 + ∆2 V2 + ∆S S.
In a small time step dt, the change of dΠ is
dΠ = ∆1 dV1 + ∆2 dV2 + ∆S dS,

9.3. CONVERTIBLE BONDS 119
where
dVi
= µi dt + σr,i dzr + σS,i dzS ,
Vi
σ2 2 w2

1
µi = Vi,t + S Vi,SS + ρSwVi,Sr + Vi,rr + µSVi,S + uVi,r ,
Vi 2 2
1
σS,i = σSVi,S
Vi
1
σr,i = wVi,r .
Vi
This implies
dΠ = (∆1 µ1 V1 + ∆2 µ2 V2 + ∆µS) dt
+ (∆1 σS,1 V1 + ∆2 σS,2 V2 + ∆σS) dzS
+ (∆1 σr,1 V1 + ∆2 σr,2 V2 ) dzr .
We choose ∆1 , ∆2 and ∆S to cancel the randomness terms dzr and dzS . This means that
(∆1 σS,1 V1 + ∆2 σS,2 V2 + ∆σS) dzS = 0

(∆1 σr,1 V1 + ∆2 σr,2 V2 ) dzr = 0.
And it yields
dΠ = (∆1 µ1 V1 + ∆2 µ2 V2 + ∆µS) dt
= r(∆1 V1 + ∆2 V2 + ∆S) dt.
Or equivalently,
∆1 (µ1 − r)V1 + ∆2 (µ2 − r)V2 + ∆(µ − r)S = 0.
This equality together with the previous two give that there exist λr and λS such that
(µ1 − r) = λS σS,1 + λr σr,1

(µ2 − r) = λS σS,2 + λr σr,2
(µ − r) = λS σ
The functions λr and λS are called the market prices of risk with respect to r and S, respectively.
Plug the formulae for µi , σr,i and σS,i , we obtain the Black-Scholes equation for a convertible bond:
σ2 2 ∂ 2 ∂2 w2 ∂ 2

∂ ∂ ∂
+ S + ρSw + + rS + (u − λr w) − r V = 0.
∂t 2 ∂S 2 ∂S∂r 2 ∂r2 ∂S ∂r
Appendix A
Basic theory of stochastic calculus
A.1 From random walk to Brownian motion

This section gives a more rigorous approach to construct Brownian motion from random walk.
Without loss of generality, we shall construct motion B(t) for t ∈ [0, 1]. Let us consider a hierar-
chical grids: for ` ≥ 1 integer, define
t`j := j2−` , j = 1, ..., 2` ,
A hierarchical independent random variables ∆`j are i.i.d.: ∆`j ∼ ∆,

1 prob. 1/2
∆ :=
−1 prob. 1/2.
This corresponds to a binomial trial with sample space
Ω` := {(a`1 , ..., a`2` )|a`j = 1 or − 1}

` `
The σ-algebra F ` is defined to be F ` := 2Ω . There are 2(2 ) elements in Ω` . The Probability P ` at
`
each sample in Ω` is defined to be 2−2 .
We define the random variables
j
1 X `
x`j := √ ∆k , j = 1, ..., 2` ,
2` k=1
and define a path x` which is piecewise linear and its values at grid points are defined to be
x` (t`j ) := x`j .
Each sample ω ∈ Ω` corresponds to a unique piecewise linear path. We may also imagine the
sample space Ω` is the collection of such paths.
`
The grids G` := {t`j }2j=0 , the path x` , the sample space Ω` and the σ-algebra F ` have the nested
structure:
121
122 APPENDIX A. BASIC THEORY OF STOCHASTIC CALCULUS
(1) G`−1 ⊂ G` with t`−1

j = t`2j ;
(2) Ω`−1 ⊂ Ω` with a`2j−1 = a`2j = a`−1

j ;
(3) F `−1 ⊂ F ` .
(4) The probability P ` on F ` , as restricted to F `−1 is identical to P `−1
With this nested structure, we can define the Brownian motion to be

[ [
Ω= Ω` , F = F `,
` `
with probability P := P ` on F ` .
The Brownian motion B(t) has the following properties:
1. B(0) = 0 and B(t) is continuous. Indeed, we should expect that B(·) is Hölder continuous
continuous of order 1/2;
2. If t1 < t2 ≤ t3 < t4 , then B(t4 ) − B(t3 ) and B(t2 ) − B(t1 ) are independent;
3. B(s + t) − B(s) ∼ N (0, t). This is mainly due to the central limit theorem.
A.2 Brownian motion

Let (Ω, F, P ) be a probability space. A process is a function X : [0, ∞) × (Ω, F) 7→ (E, B), such
that for each t ≥ 0, X(t) is a random variable. Here, E = Rd , B is the set of all Borel sets in E.
Let
Ft = σ{X(s), s ≤ t}
F t = σ{X(s), s ≥ t}
A process is called Markov if
P (A | Ft ) = P (A | X(t)), ∀A ∈ F t .
This is equivalent to
P {X(r) ∈ B|Ft } = P {X(r) ∈ B|X(t)} ∀r > t,
A Markov process is characterized by its transition probability:
P (t, x, s, B) := P {X(s) ∈ B | X(t) = x}
with initial distribution

P {X(0) ∈ B} = ν(B).
A.2. BROWNIAN MOTION 123
Theorem 1.10. If X is a Markov process, then the corresponding transition probability P satisfies
Chapman-Kolmogorov equation:
Z
P (t0 , x0 , t1 , dx1 )P (t1 , x1 , t2 , B) dx1 = P (t0 , x0 , t2 , B).
Conversely, if P is a function satisfies Chapman-Kolmogorov equation, then there is a Markov

process whose transition probability is P .
Two standard Markov processes are the Wiener process and the Poisson process. The Wiener
process has the transition probability density function
1 |x − y|2
p(t, x, s, y) = exp(− ).
(2π(t − s))d/2 2(s − t)
Such a distribution is called a normal distribution with mean x and variance s − t.
Definition 1.8. Brownian motion: A process is called a Brownian motion(or a Wiener process) if
1. Bt − Bs has normal distribution with mean 0 and variance t − s,
2. it has independent increments: that is Bt − Bs is independent of Bu for all u ≤ s < t,
3. Bt is continuous in t.
It is easy to see that Bt is markovian and its transition probability is
1 |x − y|2
p(t, x, s, y) = exp(− ).
(2π(t − s))d/2 s−t
Definition 1.9. A process {Xt | t ≥ 0} is called martingale if
(i) EXt < ∞,
(ii) E(Xu | Xs , 0 ≤ s ≤ t} = Xt .
This means that if we know the value of the process up to time t and Xt = x, then the future
expectation of Xu is x.
Theorem 1.11. 1. Bt is a martingale.
2. Bt2 − t is a martingale.
Proof. 1. E(Bt ) = 0. From the fact that Bt+s − Bt is independent of Bt , for all s > 0, we
obtain E(Bt+s − Bt | Bt ) = 0, for all s > 0. Hence,
E(Bt+s | Bt ) = E((Bt+s − Bt ) + Bt | Bt )
= E((Bt+s − Bt ) | Bt ) + E(Bt | Bt )
= Bt
2. E(Bt2 ) = t < ∞.
3. Use
2
Bt+s = ((Bt+s − Bt ) + Bt )2
= (Bt+s − Bt )2 + 2Bt (Bt+s − Bt ) + Bt2
and the fact that Bt+s − Bt is independent of Bt , we obtain

2
E(Bt+s | Bt ) = s + Bt2 .
Hence,
2
E(Bt+s − (t + s) | Bt ) = Bt2 − t.
We can show that the Brownian motion has infinite total variation in any interval. This means
that X
lim |B(ti ) − B(ti−1 )| = ∞.
However, its quadratic variation, defined by
X
[B, B](a, b) := lim |B(ti ) − B(ti−1 )|2
is finite. Here, a = t0 < t1 < · · · < tn = b is a partition of (a, b), and the limit is taken to be
max(ti − ti−1 ) → 0.
Theorem 1.12. We have [B, B](0, t) = t almost surely.

Pn
Proof. Let us partition (0, t) evenly into 2n subintervals. Let Tn = 2i=1 |B(ti ) − B(ti−1 )|2 . We
see that X X
ETn = E(|B(ti ) − B(ti−1 )|2 )) = |ti − ti−1 | = t.
X
V ar(Tn ) = V ar( |B(ti ) − B(ti−1 )|2 )
X
= V ar((B(ti ) − B(ti−1 ))2 )
X
= 2 (ti − ti−1 )2
= 2t2 2−n
P∞
Hence, n=1 V ar(Tn ) < ∞. From Fubini theorem,
∞
!
X
2
E (Tn − ETn ) < ∞.
n=1
This implies E((Tn − ETn )2 ) → 0 and hence, Tn → ETn almost surely.

A.3. STOCHASTIC INTEGRAL 125
A.3 Stochastic integral

We shall define the integral Z t
f (s)dB(s).
0
The method can be applied with B replaced by a martingale, or a martingale plus a function with
finite total variation. We shall require that f ∈ H 2 , which means:
(i) f (t) depends only on the history Ft of Bs , for s ≤ t,
Rt
(ii) 0 E|f |2 < ∞.
For this kind of functions, we can define its Itô’s integral as the follows.
1. We define Itô’s integral for f ∈ H 2 and being step functions. Its Itô’s integral is define by
Z t Xn
f (s) dB(s) = f (ti−1 )(B(ti ) − B(ti−1 )).
0 i=0
2. We use above step functions fn to approximate general function f ∈ H 2 . Using the fact that,
for step functions g ∈ H 2 ,
Z t 2 ! Z t
E g(s) dB(s) = E|g(s)|2 ds,
0 0
one can show that Z t

lim fn (s) dB(s)
n→∞ 0
almost surely.
An typical example is Z t
1 t
B(s) dB(s) = B(t)2 − .
0 2 2
From the definition, the integral can be approximated by
n
X
In = B(ti−1 )(B(ti ) − B(ti−1 ).
i=1
We have
n
1 X 2
B (ti ) − B 2 (ti−1 ) − (B(ti ) − B(ti−1 ))2

In =
2
i=1
n
1 2 1X
= B (t) − (B(ti ) − B(ti−1 ))2
2 2
i=1
We have seen that the second on the right-hand side tends to 12 t almost surely.
A.4 Stochastic differential equation

A stochastic differential equation has the form
dXt = a(Xt , t)dt + b(Xt , t)dB(t) (1.1)
A Markov process X is said to be a strong solution of this s.d.e. if it satisfies

Z t Z t
Xt − X0 = a(Xs , s) ds + b(Xs , s)dB(s).
0 0
Theorem 1.13 (Itô’s formula). If X satisfies the s.d.e. dX = adt + bdB, then
1
df (X(t)) = (af 0 (X(t)) + b2 f 00 (X(t)))dt + f 0 (X(t))bdB(t).
2
We shall give a brief idea of the proof. In a small time step (ti−1 , ti ), let ∆B = B(ti )−B(ti−1 ),
∆t = ti − ti−1 . We have
1
f (X(ti−1 ) + ∆X) − f (X(ti−1 )) = f 0 ∆X + f 00 (∆X)2 + o((∆X)2 ).
2
We notice that
(∆X)2 = (a∆t + b∆B)2

= a2 (∆t)2 + 2ab∆t + b2 (∆B)2
≈ b2 ∆t + o(∆t).
Plug ∆X and (∆X)2 into the Taylor expansion formula for f . This yields the Itô’s formula.
A.5 Diffusion process

For a s.d.e.(1.1), we define the operator Tt on functions f ∈ L2 (R) by
Tt f = Ex,t (f (X(s)), s > t.
The operator Tt maps L2 (R) to itself. It is a semigroup, that is
Tt1 ◦ Tt2 = Tt1 +t2 .
This can be proven from the Markovian property of X(·). Indeed, in terms of the transition proba-
bility density function p(t, x, s, y),
Z
Tt f = p(t, x, s, y)f (y) dy.
A.5. DIFFUSION PROCESS 127
For a semigroup Tt , we define its generator as

Tt f − f
Lf := lim .
t→0 t
From Itô’s formula,
1
Lf := af 0 + b2 f 00 .
2
Rt
This follows from Itô’s formula and E( 0 g(s) dB(s) = 0).
With a fixed s > t and f , we define
u(x, t) := (Tt f )(x, t) = Ex,t (f (X(s)).
Theorem 1.14. If X satisfies the s.d.e. (1.1), then the associate u(x, t) := Ex,t (f (X(s)), s > t,
satisfies the backward diffusion equation:
ut + Lu = 0,
and has final condition u(s, x) = f (x).
Proof. First, we notice that
u(x, t − h) = Ex,t−h (f (X(s))
= Ex,t−h EX(t),t (f (X(s))
= Ex,t−h (u(X(t), t))
This shifts final time from s to t. Now, we consider
u(x, t − h) − u(x, t) 1
= Ex,t−h (u(X(t), t) − u(X(t − h), t))
h h
1 t
Z
= Lu(X(s), t) ds
h t−h
→ Lu(x, t)
as h → 0+. Next, u(x, s − h) = Ex,s−h f (X(s)), we have X(s) → x as h → 0+ almost surely.
Hence u(x, s − h) → f (x) as h → 0+.
Since u can be represent as

Z
u(x, t) = p(x, t, s, y)f (y) dy,
we obtain that p satisfies

pt + Lx p = 0,
and
p(s, x, s, y) = δ(x − y).
For diffusion equation with source term, its solution can be represented by the following Feymann-
Kac formula.
Theorem 1.15 (Feymann-Kac). If X satisfies the s.d.e. (1.1), then the associate
Z s
u(x, t) := Ex,t f (X(s)) exp g(X(τ ), τ ) dτ , s > t (1.2)
t
solves the backward diffusion equation:
ut + Lu + gu = 0,
with final condition u(s, x) = f (x).
Proof. As before, we shift final time from s to t:

Z s
u(x, t − h) = Ex,t−h f (X(s)) exp g(X(τ ), τ ) dτ
t−h
Z s
= Ex,t−h EX(t),t f (X(s)) exp g(X(τ ), τ ) dτ
t−h
Z s Z t
= Ex,t−h EX(t),t f (X(s)) exp g(X(τ ), τ ) dτ · EX(t),t exp g(X(τ ), τ ) dτ
t t−h
Z t
= Ex,t−h u(X(t), t) exp g(X(τ ), τ ) dτ .
t−h
Here, we have used independence of X in the regions (t − h, t) and (t, s). Now, from Itô’s formula,
we have
Z t
u(x, t − h) − u(x, t) = Ex,t−h u(X(t), t) exp g(X(τ ), τ ) dτ − u(X(t − h), t)
t−h
Z t Z s
= Ex,t−h d u(X(s), t) exp g(X(τ ), τ ) dτ
t−h t−h
Z t
= Ex,t−h (Lu(X(s), t) + u(X(s), t)g(X(s), s)) ds.
t−h
From this, it is easy to see that
u(x, t − h) − u(x, t)
lim = Lu(x, t) + g(x, t)u(x, t).
h→0+ h

Financial Mathematics: I-Liang Chern

Uploaded by

Copyright:

Available Formats

Financial Mathematics: I-Liang Chern

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Financial Mathematics: I-Liang Chern

Uploaded by

Copyright:

Available Formats

FINANCIAL MATHEMATICS

December 14, 2016

2 Asset Price Model 9

3.2.2 European options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

4 Variations on Black-Scholes models 55

7 Exotic Options 103

8 Path-Dependent Options 109

9 Bonds and Interest Rate Derivatives 115

A Basic theory of stochastic calculus 121

1.2 Financial Derivatives

• the certain day—maturity or expiration date,

• the certain price—delivery price,

• the person who holds the contract is called in long position.

1.2.2 Futures (futures contracts)

Forward Contract Future Contract

• European options : Options can only be exercised at the maturity date.

• S current asset price

• ST asset price at time T

• c premium, the price of call option

• r bank interest rate

1.3 Payoff functions

future (long) future (short)

(a) left (b) right

Payoff of call options

• Long call: Λ = max{ST − E, 0}

• Short call: Λ = min{E − ST , 0}

call option (long):max{ST−K,0} call option (short):−max{ST−K,0}=min{K−ST,0}

(a) left (b) right

Payoff of put options

• Long put: Λ = max{E − ST , 0}

• Short put: Λ = min{ST − E, 0}

1.3.1 Premium of an option

the current time t = Feb 3

E 2650 2700 2750 2800 2850 2900 2950 3000

put option (long):max{K−ST,0} put option (short):−max{K−ST,0}=min{ST−K,0}

(a) left (b) right

1.4 Other kinds of options

1.5 Types of traders

2. Hedgers (to make the outcomes more certain)

1.6 Basic assumption

Asset Price Model

2.1 Efficient market hypothesis

2. Market responds immediately to any new information about an asset.

2.2 The asset price model

2.2.1 The discrete asset price model

2.2.2 The continuous asset price model

• Deterministic part: This can be modeled by

The overall asset price model is then given by

2.3 The solution of the discrete asset price model

of the asset price to be:

2.4 Binomial distribution

explicitly and also find its properties.

P (Zn = m∆x|Z0 = 0) = P (Yn = `)

with ` = (n + m)/2, we get

To compute moments, for instance

< m >= 2 < ` > −n = 0.

To compute the second moment < m2 >, from m = 2` − n, we have

< m2 >= 4 < `2 > −4n < ` > +n2 = n.