Introduction to

Options Pricing Theory

Simone Calogero

Chalmers University of Technology

 Preface

This text presents a self-contained introduction to the binomial model and the Black-Scholes
model in options pricing theory. It is the main literature for the course “Options and
Mathematics” at Chalmers, which provides the students with a first rudimentary knowledge
in mathematical finance (in particular, without using stochastic calculus). The pre-requisites
to follow this text are the standard basic courses in mathematics, such as calculus and linear
algebra. No previous knowledge on probability theory and finance are required. Each chapter
is complemented with a number of exercises and Matlab codes. The exercises marked with
the symbol (?) aim to critical thinking and do not necessarily have a well-defined unique
solution. The solution of the exercises marked with the symbol (•) and the answer to those
marked with the symbol (?) can be found in appendixes B and C at the end of the text.
Further exercises are found in Appendix D. Finally the proof of the theorems marked with
the symbol (∗) can be skipped on first reading.

Remark: The Matlab codes presented in this text are not optimized. Moreover the powerful
vectorization tools of Matlab are not employed, in order to make the codes easily adaptable
to other computer softwares and languages. The task to improve the codes presented in this
text is left to the interested reader.

Front cover picture: 10 years historical price of the Lehman Brothers stock

1
Contents

1 Warm-up 4
1.1 Basic financial concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Qualitative properties of option prices . . . . . . . . . . . . . . . . . . . . . . 17

2 Binomial markets 25
2.1 The binomial stock price . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.2 Binomial markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3 Arbitrage portfolio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.4 Computation of the binomial stock price with Matlab . . . . . . . . . . . . . 36

3 European derivatives 40
3.1 The binomial price of European derivatives . . . . . . . . . . . . . . . . . . . 41
3.1.1 Example: A standard European derivative . . . . . . . . . . . . . . . 46
3.1.2 Example: A non-standard European derivative . . . . . . . . . . . . . 48
3.2 Hedging portfolio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.3 Computation of the binomial price of standard European derivatives with
Matlab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4 American derivatives 59
4.1 The binomial price of American derivatives . . . . . . . . . . . . . . . . . . . 60
4.2 Optimal exercise time of American put options . . . . . . . . . . . . . . . . . 63
4.3 Example of American put option . . . . . . . . . . . . . . . . . . . . . . . . 64
4.4 Hedging portfolio processes of American derivatives . . . . . . . . . . . . . . 67
4.5 Computation of the fair price of American derivatives with Matlab . . . . . . 72

5 Introduction to Probability Theory 75

2
 5.1 Finite Probability Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.2 Random Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.2.1 Expectation and Variance . . . . . . . . . . . . . . . . . . . . . . . . 82
5.2.2 Independence and Correlation . . . . . . . . . . . . . . . . . . . . . . 84
5.2.3 Conditional expectation . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.3 Stochastic processes. Martingales . . . . . . . . . . . . . . . . . . . . . . . . 89
5.4 Applications to the binomial model . . . . . . . . . . . . . . . . . . . . . . . 92
5.4.1 Binomial distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.4.2 General discrete options pricing models . . . . . . . . . . . . . . . . . 98
5.4.3 Quantitative vs fundamental analysis of a stock . . . . . . . . . . . . 99
5.5 Infinite Probability Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.5.1 Joint distribution. Independence . . . . . . . . . . . . . . . . . . . . 103
5.5.2 Central limit theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.5.3 Brownian motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

6 Black-Scholes options pricing theory 109

6.1 Black-Scholes markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.2 Black-Scholes price of standard European derivatives . . . . . . . . . . . . . 113
6.3 Black-Scholes price of European call and put options . . . . . . . . . . . . . 118
6.4 The Black-Scholes price of other standard European derivatives . . . . . . . 122
6.5 Implied volatility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.6 Standard European derivatives on a dividend-paying stock . . . . . . . . . . 127
6.7 Optimal exercise time of American calls on dividend-paying stocks . . . . . . 130

A The Markowitz portfolio theory 133

B Solutions to selected exercises 138

C Answer to selected exercises 167

D Additional exercises 168

3
Chapter 1

Warm-up

The purpose of this chapter is threefold: (1) introduce a few basic financial concepts, (2)
formulate and discuss the main assumptions behind the standard theory of options pricing,
(3) derive some fundamental qualitative properties of option prices.

1.1 Basic financial concepts

For a more detailed discussion on the concepts introduced in this section, see [2].

Financial assets
The term asset may be used to identify any economic resource capable of producing value
and which, under specific legal terms, can be bought and sold (i.e., converted into cash).
Assets may be tangible (e.g., lands, buildings, commodities, etc.) or intangible (e.g., patents,
copyrights, stocks, etc.). Assets are also divided into real assets, i.e, assets whose value is
derived by an intrinsic property (e.g., tangible assets), and financial assets, such as stocks,
options, bonds, etc., whose value is instead derived from a contractual claim on the income
generated by another (possibly real) asset. For example, upon holding shares of the Volvo
stock (a financial asset), we can make a profit from the production and sale of cars even if
we do not own an auto plant (a real asset). As we consider only financial assets in these
notes, the terms “asset” and “financial asset” will be henceforth used interchangeably.

Price
The price of a financial asset is the value, measured in some units of currency (e.g. dollars),
at which the buyer and the seller agree to exchange ownership of the asset. The price is
chosen by the two parties as a result of some kind of “negotiation”. More precisely, the ask

4
price is the minimum price at which the seller is willing to sell the asset, while the bid price
is the maximum price that the buyer is willing to pay for the asset. When the difference
between these two values, called bid-ask spread, becomes zero, the exchange of the asset
takes place at the corresponding price.
A generic financial asset will be denoted by U and its price at time t by ΠU (t). Prices are
generally positive, although some financial assets (e.g., forward contracts) have zero price.
The asset price refers to the price of a share of the asset, where “share” stands for the
minimum amount of an asset which can be traded. In these notes all prices are given in a
fixed currency, which is however left unspecified.

Markets
Financial assets can be traded in official markets or in over the counter (OTC) markets.
In the former case all trades are subject to a common legislation, while in the latter the
exchange conditions are agreed upon by the individual traders. Example of OTC markets
are the currency markets (Forex) and the bond markets, while stock markets, option markets
and futures markets are all examples of official markets. Buyers and sellers of assets in a
market will be called investors or agents.

Long and short position

Besides the usual operations of “buying” and “selling” the asset, we need to consider an
additional common type of transaction, which is called short-selling. Short-selling an asset
(typically a stock) is the practice of selling the asset without actually owning it. Concretely,
an investor is short-selling N shares of an asset if the investor borrows the shares from a
third party and then sell them immediately on the market. The reason for short-selling an
asset is the expectation that the price of the asset will decrease in the future. More precisely,
assume that N shares of an asset U are short-sold at time t = 0 for the price ΠU (0) and
let T > 0 be the time at which the shares must be returned to their original owner. If
ΠU (T ) < ΠU (0), then upon re-purchasing the N shares at time T , and returning them to
the lender, the short-seller will make the profit N (ΠU (0) − ΠU (T )).
An investor is said to have a long position on an asset if the investor owns the asset and
will therefore profit from an increase of the price of the asset. Conversely, the investor is
said to have a short position on the asset if the investor will profit from a decrease of its
value, as it happens for instance when the investor is short-selling the asset.
Finally we remark that any transaction in the market is subject to transaction costs (e.g.,
broker’s commissions and lending fees for short-selling) and transaction delays (trading
in real markets is not instantaneous).

5
Stocks and dividends
The capital stock of a company is the part of the company equity capital that is made
publicly available for trading. Stocks are most commonly traded in official markets (stock
markets). For instance, over 300 company stocks are traded in the Stockholm exchange
market. The price per share at time t > 0 of a generic stock will be denoted by S(t).
A stock may occasionally pay a dividend to its shareholders. This means that a fraction
of the stock price is deposited to the bank account of the shareholders. The day at which
the dividend is paid, as well as its amount in percentage of the opening stock price at this
day, are known in advance. After the dividend has been paid, the price of the stock usually
drops of the amount paid by the dividend.

Market index and ETFs

A market index is a weighted average of the value of a collection of assets traded in one or
more official markets. For example, S&P500 (Standard and Poor 500) measures the average
value of 500 stocks traded at the New York stock exchange (NYSE) and NASDAQ markets.
Market indexes can be regarded themselves as assets. More precisely an ETF (Exchange
Traded Fund) on a market index is a financial asset whose value tracks exactly the value of
a market index (or a given fraction thereof). Hence one share of an ETF on S&P500 will
increase its value of 1% in one day if during that day S&P 500 has gained 1%. An inverse
ETF however will in the same example decrease its value of 1%. Thus ETFs give investors
the possibility to speculate whether the market will gain or loose value in the future.

Portfolio position and portfolio process

Consider an agent that invests on N assets U1 , . . . , UN during the time interval [0, T ]. Assume
that the agent trades on a1 shares of the asset U1 , a2 shares of the asset U2 ,. . . , aN shares
of the asset UN . Here ai ∈ Z, where ai < 0 means that the investor has a short position in
the asset Ui , while ai > 0 means that the investor has a long position in the asset Ui (the
reason for this interpretation will become soon clear). The vector A = (a1 , a2 , . . . , aN ) ∈ ZN
is called a portfolio position, or simply a portfolio. The value of the portfolio at time t is
given by
N
X
VA (t) = ai ΠUi (t), t ∈ [0, T ], (1.1)
i=1

where ΠUi (t) denotes the price of the asset Ui at time t. The value of the portfolio measures
the wealth of the investor: the higher is V (t), the “richer” is the investor at time t. Now we
see that when the price of the asset Ui increases, the value of the portfolio increases if ai > 0
and decreases if ai < 0, which explains why ai > 0 corresponds to a long position on the
asset Ui and ai < 0 to a short position. We also remark that portfolios can be added by using

6
the linear structure on ZN , namely if A, B ∈ ZN , A = (a1 , . . . , aN ), B = (b1 , . . . , bN ) are two
portfolios and α, β ∈ Z, then C = αA + βB is the portfolio C = (αa1 + βb1 , . . . , αaN + βbN ).
In the definition of portfolio position and portfolio value given above, the investor keeps the
same number of shares of each asset during the whole time interval [0, T ]. Suppose now that
the investor changes the position on the assets at some times
0 = t0 < t1 < t2 < · · · < tM = T ;
for simplicity we assume that at each time t1 , . . . tM the change in the portfolio position
occurs instantaneously. Let A0 denote the initial (at time t = t0 = 0) portfolio position of the
investor and Aj denote the portfolio position of the investor in the interval of time (tj−1 , tj ],
j = 1, . . . , M . As positions hold for one instance of time only are clearly meaningless, we
may assume that A0 = A1 , i.e., A1 is the portfolio position in the closed interval [0, t1 ]. The
vector (A1 , . . . , AM ) is called a portfolio process. If we denote by aij the number of shares
of the asset i in the portfolio Aj , then we see that a portfolio process is in fact equivalent to
the N × M matrix A = (aij ), i = 1, . . . , N , j = 1, . . . , M . The value V (t) of the portfolio
process at time t is given by the value of the corresponding portfolio position at time t as
defined by (1.1). Hence for t ∈ (tj−1 , tj ] and j = 1, . . . , M the value of the portfolio process
is given by
N
X
V (t) = VAj (t) = aij ΠUi (t).
i=1
The initial value V (0) = VA0 = VA1 (0) of the portfolio, when it is positive, is called the
initial wealth of the investor.
A portfolio process is said to be self-financing if no cash is ever withdrawn or infused in
the portfolio. Let us look at one example. Suppose that at time t0 = 0 the investor is short
400 shares on the asset U1 , long 200 shares on the asset U2 and long 100 shares on the asset
U3 . This corresponds to the portfolio
A0 = (−400, 200, 100),
whose value is
VA0 = −400 ΠU1 (0) + 200 ΠU2 (0) + 100 ΠU3 (0).
If this value is positive, the investor needs an initial wealth to set up this portfolio position:
the income deriving from short selling the asset U1 does not suffice to open the desired long
position on the other two assets.
As mentioned before, we may assume that the investor keeps the same position in the interval
(0, t1 ], i.e., A1 = A0 . The value of the portfolio process at time t = t1 is
V (t1 ) = VA1 (t1 ) = −400 ΠU1 (t1 ) + 200 ΠU2 (t1 ) + 100 ΠU3 (t1 ).
Now suppose that at time t = t1 the investor buys 500 shares of U1 , sells x shares of U2 , and
sells all the shares of U3 . Then in the interval (t1 , t2 ] the investor has a new portfolio which
is given by
A2 = (100, 200 − x, 0),

7
and so the value of the portfolio process for t ∈ (t1 , t2 ] is given by
V (t) = 100 ΠU1 (t) + (200 − x) ΠU2 (t), t ∈ (t1 , t2 ].
If we now take the limit of this quantity as t → t+ 1 , we get the value of the portfolio
“immediately after” the position has been changed at time t1 . Calling V (t1 +) this limit we
have (assuming that the prices are continuous)
V (t1 +) = 100 ΠU1 (t1 ) + (200 − x) ΠU2 (t1 ).
The difference between the value of the two portfolios immediately after and immediately
before the transaction is then
V (t1 +) − V (t1 ) = 100 ΠU1 (t1 ) + (200 − x) ΠU2 (t1 )
− (−400 ΠU1 (t1 ) + 200 ΠU2 (t1 ) + 100 ΠU3 (t1 ))
= 500 ΠU1 (t1 ) − x ΠU2 (t1 ) − 100 ΠU3 (t1 ).
If this difference is positive, then the new portfolio cannot be created from the old one
without extra cash. Conversely, if this difference is negative, then the new portfolio is
less valuable than the old one, the difference being equivalent to cash withdrawn from the
portfolio. Hence for self-financing portfolio processes we must have V (t1 +) − V (t1 ) = 0 (and
similarly V (tj +) − V (tj ) = 0, for all j = 1, . . . M − 1). This implies in particular that the
number x of shares of the asset U2 to be sold at time t1 in a self-financing portfolio must be
500ΠU1 (t1 ) − 100ΠU3 (t1 )
x= .
ΠU2 (t1 )
Of course, x will be an integer only in exceptional cases, which means that perfect self-
financing strategies in real markets are almost impossible.
The return of a self-financing portfolio process in the interval [0, T ] is given by
R(T ) = V (T ) − V (0), (1.2)
where V (t) denotes the value of the portfolio at time t. If the return is positive, the investor
makes a profit in the interval [0, T ], if it is negative the investor incurs in a loss. When
V (0) > 0 we may also compute the relative return of the portfolio, which is given by
V (T ) − V (0)
R∗ (T ) = . (1.3)
V (0)
A portfolio process is non-self-financing if some cash is withdrawn or added to the portfolio
at some time t ∈ (0, T ), as it happens for instance when one or more assets in the portfolio
pays a dividend in the period (0, T ). This cash flow must be included in the computation of
the return of the portfolio. Assume for instance that the investor adds the cash C1 into the
portfolio at time t1 ∈ (0, T ) and withdraws the cash C2 at time t2 ∈ (0, T ). Then the return
of the portfolio in the interval [0, T ] is V (T ) − V (0) + C2 − C1 .
Finally we remark that investment returns are commonly “annualized” by dividing the return
R(T ) by the time T expressed in fraction of years (e.g., T = 1 week = 1/52 years).

8
Historical volatility
The historical volatility of an asset measures the amplitude of the time fluctuations of the
asset price, thereby giving information on its level of uncertainty. It is computed as the
standard deviation of the log-returns of the asset based on historical data. More precisely,
let [t0 , t] be some interval of time in the past, with t denoting possibly the present time, and
let T = t − t0 > 0 be the length of this interval. Let us divide [t0 , t] into n equally long
periods, say
t0 < t1 < t2 < . . . tn = t, ti − ti−1 = h, for all i = 1, . . . n.
The set of points {t0 , t1 , . . . tn } is called a partition of the interval [t0 , t]. Assume for instance
that the asset is a stock. The log-return of the stock price in the interval [ti−1 , ti ] is given
by1  
S(ti )
Ri = log S(ti ) − log S(ti−1 ) = log , i = 1, . . . n. (1.4)
S(ti−1 )
The (corrected) sample variance of the log-returns is then
n
1 X
∆(t) = (Ri − R̄)2 ,
n − 1 i=1
where n  
1X 1 S(t)
R̄ = Ri = log (1.5)
n i=1 n S(t0 )
is the sample mean of log-returns. To obtain the T-historical variance of the asset we
divide ∆(t) by h measured in fraction of years, that is
n
1 1 X
bT2 (t)
σ = (Ri − R̄)2 (T -historical variance). (1.6)
h n − 1 i=1
The square root of the T −historical variance is the T-historical volatility:
v
u n
1 u 1 X
bT (t) = √ t
σ (Ri − R̄)2 (T -historical volatility). (1.7)
h n − 1 i=1

Note carefully that the historical volatility depends on the partition being used to compute
it.
Suppose for example that t − t0 = T = 20 days, which is quite common in the applications,
and let t1 , . . . t20 be the market closing times at these days. Let h = 1 day = 1/365 years.
Then v
n
√ u1 X
u
b20 (t) = 365
σ t (Ri − R̄)2
19 i=1
1
Throughout these notes, log x stands for the natural logarithm of x > 0 (which is also frequently denoted
by ln x in the literature).

9
 40

80

30

60

20
40
Getinge
10
20

Electrolux
0 0
0 20 40 60 80 0 20 40 60 80

(a) ELUXY-B.ST (b) GETIB.ST

100
20 Seb
80

18

60

40
Scania 16

14
20

0 12
0 20 40 60 80 0 20 40 60 80

(c) SCV-B.ST (d) SEB-C.ST

Figure 1.1: 20-days volatility of 4 stocks in the Stockholm exchange market on May 2nd ,
2014. The caption in each graph shows the ticker of the stock.

is called the 20-days historical volatility. We remark that h = 1/252 is also commonly used
as normalization factor, since there are 252 trading days in one year.
As a way of example, Figure 1.1 shows the 20-days volatility of four stocks in the Stockholm
exchange market from January 1st , 2014 until May 2nd , 2014 (88 trading days). These data
have been obtained with MATHEMATICA by running the following command on May 3rd ,
2014:

FinancialData["ticker", "Volatility20Day", {2014, 1, 1}]

Upon running this command, the software connects to Yahoo Finance and collects the 20-
days volatility data for the stock identified by the ticker symbol “ticker”, starting from the
date {2014, 1, 1} (year, month, day) until the present day. Note that in a few cases the
historical volatility remains approximately constant within periods of about 20 days.

10
Financial derivatives. Options
A financial derivative (or derivative security) is an asset whose value depends on the
performance of one (or more) other asset(s), which is called the underlying asset. There
exist various types of financial derivatives, the most common being options, futures, forwards
and swaps. In this section we discuss option derivatives on a single asset, which could be for
instance a stock2 .
A call option is a contract between two parties: the buyer, or owner, of the call and the
seller, or writer, of the call. The contract gives the owner the right, but not the obligation,
to buy the underlying asset for a given price, which is fixed at the time when the contract is
stipulated, and which is called strike price of the call. If the buyer can exercise this right
only at some given time T in the future then the call option is called European, while if
the option can be exercised at any time earlier than or equal to T , then the option is called
American. The time T is called maturity time, or expiration date of the call. The
writer of the call is obliged to sell the asset to the buyer if the latter decides to exercise the
option. If the option to buy in the definition of a call is replaced by the option to sell, then
the option is called a put option.
In exchange for the option, the buyer must pay a premium to the seller (options are not
free). Suppose that the option is a European option with strike price K and maturity time
T . Assume that the underlying is a stock with price S(t) at time t ≤ T and let Π0 be the
premium paid by the buyer to the seller. In which case is it then convenient for the buyer
to exercise the option at maturity? Let us define the pay-off of the European call as

0 if S(T ) ≤ K
Ycall = (S(T ) − K)+ := max(0, S(T ) − K) = .
S(T ) − K if S(T ) > K

Similarly, we define the pay-off of the European put by


0 if S(T ) ≥ K
Yput = (K − S(T ))+ = .
K − S(T ) if S(T ) < K

Clearly, the buyer should exercise the call option at maturity if and only if Ycall > 0, as in
this case it is cheaper to buy the stock at the strike price rather than at the market price.
Similarly the owner of the put should exercise if and only if Yput > 0, as in this case the
income generated by selling the stock at the strike price is higher then the income generated
by selling it at the market price. Hence the call or put option must be exercised at maturity if
and only if the pay-off is positive, in which case the option is said to expire in the money.
The return for the owner of the option is given by N (Ycall − Π0 ) in the case of the call and by
N (Yput − Π0 ) in the case of the put, where N is the number of option contracts in the buyer
2
Options are available on many different types of assets, including currencies, market indexes, commodi-
ties, etc.

11
 SHtL
600

500 call in the money

400

300

K
200
call out of
100
the money
0
T t
Figure 1.2: The call option with strike K = 200 and maturity T is in the money in the upper
region and out of the money in the lower region. The put option with the same parameters
is in the money in the lower region and out of the money in the upper region.

portfolio3 . Note carefully that the buyer makes a profit only if the pay-off is greater than
the premium. One of the main problems in options pricing theory is to define a reasonable
fair value for the price Π0 of options (and other derivatives).
Let us introduce some further terminology. The European call (resp. put) with strike K is
said to be in the money at time t if S(t) > K (resp. S(t) < K). The call (resp. put)
is said to be out of the money if S(t) < K (resp. S(t) > K). If S(t) = K, the (call
or put) option is said to be at the money at time t. The meaning of this terminology is
self-explanatory, see Figure 1.2.
The pay-off of American calls exercised at time t is Y (t) = (S(t) − K)+ , while for American
puts we have Y (t) = (K − S(t))+ . The quantity Y (t) is also called intrinsic value of the
American option. In particular, the intrinsic value of an out-of-the-money American option
is zero.

Option markets
Option markets are relatively new compared to stock markets. The first one has been
established in Chicago in 1974 (the Chicago Board Options Exchange, CBOE). In an option
market anyone (after a proper authorization) can be the buyer or the seller of an option.
3
Options are typically sold in multiples of 100 shares, hence the minimum amount of options that one
can buy is 100, which cover 100 shares of the underlying asset.

12
Market options are available on different assets (stocks, debts, indexes, etc.) and with
different strikes and maturities. Most commonly, market options are of American style.
Clearly, the deeper in the money is the option, the higher will be the price of the option
in the market, while the price of an option deeply out of the money is usually quite low
(but never zero!). It is also clear that the buyer of the option is the party holding the long
position on the option, since the buyer owns the option and thus hopes for an increase of its
value, while the writer is the holder of the short position.
One reason why investors buy call options is to protect a short position on the underlying
asset. In fact, suppose that an investor is short-selling 100 shares of a stock at time t = 0 for
the price S(0) and let t0 > 0 be the time at which the shares must be returned to the lender.
At time t = 0 the investor buys 100 shares of an American call option on the stock with
strike K ≈ S(0) and maturity later than t0 . If at time t0 the price of the stock is no lower
than S(0), the investor will exercise the call and thus obtain 100 shares of the stock for the
price K ≈ S(0). So doing the investor will be able to return the shares to the lender with
minimal losses. At the same fashion, investors buy put options to protect a long position on
the underlying asset4 .

Exercise 1.1 (?). Can you think of a reason why investors sell options?

Of course, speculation is also an important factor in option markets. However the stan-
dard theory of options pricing is firmly based on the interpretation of options as derivative
securities and does not take speculation into account.

European, American and Asian derivatives

By far the majority of financial derivatives, including options other then simple calls and
puts, are traded OTC. In this section we discuss a few examples, but first it is convenient to
introduce a precise mathematical definition of European and American derivatives.
Given a function g : (0, ∞) → R, the standard European derivative with pay-off Y =
g(S(T )) and maturity time T > 0 is the contract that pays to its owner the amount Y
at time T > 0. Here S(T ) is the price of the underlying stock at time T , while g is the
pay-off function of the derivative (e.g., g(x) = (x − K)+ for European call options, while
g(x) = (K − x)+ for European put options). Hence, the pay-off of standard European
derivatives depends only on the price of the stock at maturity and not on the earlier history
of the stock price. An example of standard European derivative which is actually traded in
the market (other than call and put options) is the digital option. Denote by H(x) the
Heaviside function, 
1, for x > 0
H(x) = ,
0, for x ≤ 0
4
A short-selling strategy that is not covered by a suitable security is said to be naked.

13
 10 10

8 8

6 6

4 4

SHT L SHT L
2 2
K K
5 10 15 20 5 10 15 20
premium premium
-2 -2

(a) Call option (b) Put option

10 20
premium
8
15
6
L 10
4 premium

SHT L
2 5

SHT L
K K
5 10 15 20
5 10 15 20
-2

(c) Digital option (cash-settled) (d) Digital option (physically-settled)

Figure 1.3: Pay-off function (continuous line) and return (dashed line) of some standard
European derivatives.

and let K, L > 0 be constants expressed in units of some currency (e.g., dollars). The
standard European derivative with pay-off function g(x) = LH(x−K) is called cash-settled
digital call option; this derivative pays the amount L if S(T ) > K, and nothing otherwise.
The physically-settled digital call option has the pay-off function g(x) = xH(x − K),
which means that at maturity the buyer receives either the stock (when S(T ) > K), or
nothing. Digital options are also called binary options. Figure 1.3 shows the pay-off function
of call, put and digital call options with strike K = 10. Drawing the pay-off function of a
derivative helps to get a first insight onto its properties.

Exercise 1.2. Given K, ∆K > 0, consider the standard European derivative with maturity
T and pay-off function

g(x) = (x − K + ∆K)+ − 2(x − K)+ + (x − K − ∆K)+ .

Draw the graph of g and derive the range of S(T ) for which the derivative expires in the
money.

If the pay-off depends on the history of the stock price during the interval [0, T ], and not
just on S(T ), we shall say that the contract is a non-standard European derivative. An

14
example of non-standard European
RT derivative is the so-called Asian call option, the pay-off
1
of which is given by Y = ( T 0 S(t) dt − K)+ .
The value at time t of the European derivative with pay-off Y and expiration date T will be
denoted by ΠY (t) (we do not include the expiration date in our notation).
The term “European” refers to the fact that the contract cannot be exercised before time
T . For a standard American derivative the buyer can exercise the contract at any time
t ∈ (0, T ] and so doing the buyer will receive the amount Y (t) = g(S(t)), where g is the pay-
off function of the American derivative. Non-standard American derivatives can be defined
similarly to the European ones, but with the further option of earlier exercise.

Exercise 1.3. Look for the definition of the following options: Bermuda option, Com-
pound option, Lookback option, Barrier option, Chooser option. Classify them as Ameri-
can/European, standard/non-standard and write down their pay-off function.

Money market
A money market is a (OTC) market in which the objects of trading are short term loans,
i.e., loans with maturity between one day and one year5 . Assets in the money market bear
a very low risk of default and for this reason they are also called risk-free assets, although
this terminology is criticized by many scholars. Examples of risk-free assets in the money
market are commercial papers and repurchase agreements (repo). In contrast to stock and
option markets, money markets are typically accessible only by financial institutions and not
by private investors.
The value at time t of a generic risk-free asset in the money market will be denoted by B(t);
the difference B(t2 ) − B(t1 ) is determined by the interest rate of the asset in the interval
[t1 , t2 ]. We say that a risk-free asset has instantaneous interest rate r(t) in the interval
[t1 , t2 ] if Z t 
B(t) = B(t1 ) exp r(s) ds , for t1 ≤ t ≤ t2 . (1.8)
t1

Under normal market conditions we have B(t2 ) > B(t1 ), i.e., the interest rate of risk-free
assets is usually positive, but exceptions are possible. For instance, short term loans issued
by the Swedish central bank (Riksbanken) have presently (2017) a negative interest rate.
To see how the money market works in practice, suppose that an investor buys a risk-free
asset at time t = 0 which expires at time T > 0. The seller will receive the quantity
B0 = B(0). As part of the agreement, the seller promises to re-purchase the risk-free asset
at time T for B(T ) > B0 . Hence buying an asset in the money market is equivalent to
lend money to the seller, while selling an asset in the money market is equivalent to borrow
money from the buyer. The seller of the risk-free asset is the party holding the short position
on the asset, while the buyer holds the long position.
5
Loan contracts with maturity longer than one year are called bonds and are traded in the bond market.

15
We remark that there exists several ways to define the interest rate of a risk-free asset; in
particular, the interest rate may be compounded discretely in time, rather then continu-
ously as in (1.8). Inasmuch as in these notes we use only (1.8) to compute the value of assets
in the money market, we shall refer to r(t) simply as the interest rate of the risk-free asset,
i.e., the word “instantaneous” will be omitted for brevity. Moreover we shall always assume
that the interest rate is a constant r, so that (1.8) simplifies to

B(t) = B(t1 ) exp(r(t − t1 )), for t1 ≤ t ≤ t2 . (1.9)

The interest rate is measured in yearly percentage. For example, if one share of a risk-free
asset has initial value B(0) = 10 at time t = 0 and interest rate 10% per year, then after 1
month=1/12 years its value is B(1/12) = 10 exp(0.1/12) ≈ 10.0837.
Note that so far we have introduced three strategies that investors can undertake to obtain
cash: short-selling an asset, writing an option or borrowing from the money market.

Frictionless markets
As all mathematical models, also those in options pricing theory are based on a number of
assumptions. Some of these assumptions are introduced only with the purpose of simplifying
the analysis of the models and often correspond to facts that do not occur in reality. Among
these “simplifying” assumptions we impose that

1. There is no bid/ask spread6

2. There are no transaction costs and trades occur instantaneously

3. An investor can trade any fraction of shares

4. No lack of liquidity: there is no limit to the amount of cash that can be borrowed from
the money market

We have seen in the previous sections that real markets do not satisfy exactly these assump-
tions, although in some case they do it with reasonable approximation. For instance, if the
investor is an agent working for a large financial institution, then the above assumptions
reflect reality quite well. However they work very badly for private investors. We summarize
the validity of these assumptions by saying that the market has no friction. The idea is
that when the above assumptions hold, trading proceeds “smoothly without resistance”.
In a frictionless market we may define the portfolio process of an agent who is investing on
N assets during the time interval [0, T ] as a function

A : [0, T ] → RN , A(t) = (a1 (t), . . . , aN (t)),

6
In particular, any offer to buy/sell an asset is matched by an offer to sell/buy the asset.

16
i.e., by assumptions 2 and 3, the number of shares ai (t) of each single asset at time t is
now allowed to be any real number and to change at any arbitrary time in the interval
[0, T ]. Portfolio processes can be added using the linear structure in RN , namely if B =
(b1 (t), . . . , bN (t)), then A + B is the portfolio
A + B = (a1 (t) + b1 (t), . . . , aN (t) + bN (t)).
The value at time t of the portfolio process A is
N
X
VA (t) = a(t)ΠUi (t),
i=1

and clearly
VA (t) + VB (t) = VA+B (t).
Moreover it is clear that, thanks to assumption 3, perfect self-financial portfolio processes in
frictionless markets always exist.
A further simplifying assumption that we make in the rest of these notes is the following:

5. All risk-free assets in the the money market have the same constant interest rate r

Although r is usually positive in the applications, we shall sometimes allow r = 0 or even a

negative interest rate. We shall refer to r as the interest rate of the money market. In the
applications it is customary to choose the value of r to be an interbank offered rate, such
as LIBOR, or EURIBOR, etc., that is the average interest rate at which banks in a given
geographical zone lend money to one another.

1.2 Qualitative properties of option prices

The purpose of this section is to derive some qualitative properties of option prices using only
basic principles, without invoking any specific mathematical model for the market dynamics.
The following notation will be used. S(t) denotes the price at time t > 0 of a given stock,
C(t, S(t), K, T ) denotes the price at time t ∈ [0, T ] of the European call option on the stock
with strike K > 0 and maturity T > 0. The price of the European put option with the same
parameters will be denoted by P (t, S(t), K, T ); finally C(t,
b S(t), K, T ) and Pb(t, S(t), K, T )
denote the values of the corresponding American call and put option.

No-dummy investor principle

Probably the most self-evident of all financial principles is the following, which we call the
no-dummy investor principle7 :
7
More commonly (and respectfully) known as rational investor principle.

17
 Investors prefer more to less and do not undertake
trading strategies which result in a sure loss.

This principle has a number of straightforward consequences. For example, an investor will
never exercise an option which is out of the money, while an option that expires in the money
is always exercised8 . Moreover the price of stocks and options (prior to expire) is always
positive.

Exercise 1.4 (?). Use the no-dummy investor principle to justify the following properties.

(i) The price of a financial derivative tends to its pay-off as maturity is approached. In
particular, for European call/put options,

C(t, S(t), K, T ) → (S(T ) − K)+ , P (t, S(t), K, T ) → (K − S(T ))+ ,

as t → T − and similarly for American options;

(ii) An American derivative is at least as valuable as its European counterpart. In partic-

ular, for call/put options,

b S(t), K, T ) ≥ C(t, S(t), K, T ),

C(t, Pb(t, S(t), K, T ) ≥ P (t, S(t), K, T )

(iii) The price of an American derivative is always larger or equal to its intrinsic value. In
particular, for American call/put options,

b S(t), K, T ) ≥ (S(t) − K)+ ,

C(t, Pb(t, S(t), K, T ) ≥ (K − S(t))+ .

Any reasonable mathematical model for the price of options must be consistent with the
properties (i)-(iii) in the previous exercise. In the rest of this section they are assumed to
hold without any further comment.

Arbitrage-free principle
An arbitrage opportunity is an investment strategy that requires no initial wealth and
which ensures a positive profit without taking any risk. For example, suppose that at time
t = 0 an investor sells one share the American call option with strike K and maturity T1
and buys one share of the American call on the same stock with the same strike but with
maturity T2 > T1 . Suppose that the price of the latter option is lower than the price of
the former, i.e., C
b2 := C(0,
b S(0), K, T2 ) < C(0,
b S(0), K, T1 ) := C b1 . The investor will then
have the cash Cb1 − Cb2 available to buy shares of a risk-free asset in the money market. This
8
Provided of course the owner of the option can afford to exercise. For instance, the buyer of a call option
may not have the cash required to buy the underlying when the call expires in the money.

18
portfolio is clearly riskless: if the buyer of the option with maturity T1 decides to exercise
at some time t ≤ T1 , the investor can pay-off the buyer by exercising his/her own option.
Hence this investment is an example of arbitrage opportunity: it requires no initial wealth,
it entails no risk and it ensures a positive profit. However, why should the investor be able to
find someone willing to pay more for an option that expires earlier? This would be of course
a dummy investment for the buyer. Due to the complexity of modern markets, arbitrage
opportunities do actually exist, but only for a very short time, as they are quickly exploited
and “traded away” by investors.
The previous discussion leads us to assume the validity of the so-called arbitrage-free
principle:

Asset prices in a market are such that no

arbitrage opportunities can be found.

Asset prices that are consistent with this principle are said to be arbitrage free or fair.

Dominance principle
The arbitrage-free principle can be used to derive a number of qualitative properties of option
prices, which are not as obvious as (i)-(iii) in Exercise 1.4. To this purpose we need first to
express the arbitrage-free principle in a more quantitative form. There are several ways to
do this, e.g. by requiring the absence of arbitrage portfolios in the market (see next chapter),
or by imposing the so-called dominance principle [3]:

Dominance Principle: Suppose that t < T is the present time and consider a portfolio
which does not contain dividend-paying assets or short positions on American derivatives. If
the value of the portfolio is non-negative at time T , i.e., V (T ) ≥ 0, then V (t) ≥ 0.

The fact that the dominance principle must hold as a consequence of the arbitrage-free
principle is clear. In fact, if V (t) < 0, then the investor needs no initial wealth to open
the portfolio, while on the other hand the portfolio return is positive, since V (T ) − V (t) ≥
−V (t) > 0. Hence the given portfolio ensures a positive profit without taking any risk, which
violates the arbitrage-free principle.
Let us comment further on the formulation of the dominance principle. First of all, the
requirement that the portfolio does not contain short positions on American derivatives is
necessary, otherwise there is no guarantee that the portfolio exists up to time T (the buyer
may exercise the derivative prior to T ). The reason to require that the assets pay no dividend
is the following. Suppose that a stock pays a dividend of 2 % at time T . Just before that
the investor open a short position on the stock and invest 99% of the income on a risk-free
asset. Hence the value of this portfolio is negative, but it becomes instantaneously positive

19
when the dividend is paid9 .
The following simple theorem will be used for our applications of the dominance principle.
Theorem 1.1. Assume that the dominance principle holds and let A, B be two portfolios
which do not contain dividend pay assets or American derivatives. Then, for t < T :

(a) If VA (T ) = 0, then VA (t) = 0;

(b) If VA (T ) = VB (T ), then VA (t) = VB (t);
(c) If VA (T ) ≥ VB (T ), then VA (t) ≥ VB (t).

Proof. (a) The dominance principle implies VA (t) ≥ 0. As the portfolio A does not contain
American derivatives, the dominance principle applies to the portfolio −A as well. We
obtain V−A (t) = −VA (t) ≥ 0, or VA (t) ≤ 0. Hence VA (t) = 0. (b) follows by part (a) and
the relation VA−B (t) = VA (t) − VB (t). The proof of (c) is similar.

The next theorem collects a number of properties that must be satisfied by option prices as
a consequence of the dominance principle. We denote these properties by (iv)-(vii) in order
to continue the list (i)-(iii) given in Exercise 1.4.
Theorem 1.2. Assume that the dominance principle holds and let r be the interest rate of
the money market. Then, for all t < T ,

(iv) The put-call parity holds

S(t) − C(t, S(t), K, T ) = Ke−r(T −t) − P (t, S(t), K, T ). (1.10)

(v) If r ≥ 0, then C(t, S(t), K, T ) ≥ (S(t) − K)+ ; the strict inequality C(t, S(t), K, T ) >
(S(t) − K)+ holds when r > 0.
(vi) If r ≥ 0, the map T → C(t, S(t), K, T ) is non-decreasing.
(vii) The maps K → C(t, S(t), K, T ) and K → P (t, S(t), K, T ) are convex10 .

Proof. (iv) Consider a constant portfolio A which is long one share of the stock and one
share of the put option, and is short one share of the call and K/B(T ) shares of the
risk-free asset. The value of this portfolio at maturity is
K
VA (T ) = S(T ) + (K − S(T ))+ − (S(T ) − K)+ − B(T ) = 0.
B(T )
9
In practice this is not a feasible strategy as the profit is extremely small and highly surpassed by
transaction costs. Moreover certain markets require to own the stock for a sufficiently long period of time
in order to be entitled to the next dividend.
10
Recall that a real-valued function f on an interval I is convex if f (θx + (1 − θ)y) ≤ θf (x) + (1 − θ)f (y),
for all x, y ∈ I and θ ∈ (0, 1).

20
 Hence, by (a) of Theorem 1.1, VA (t) = 0, for t < T , that is

S(t) + P (t, S(t), K, T ) − C(t, S(t), K, T ) − Ke−r(T −t) = 0,

which is the claim.

(v) We can assume S(t) ≥ K, otherwise the claim is obvious (the price of a call cannot be
negative). By the put-call parity, using that P (t, S(t), K, T ) ≥ 0,

C(t, S(t), K, T ) = S(t) − Ke−r(T −t) + P (t, S(t), K, T ) ≥ S(t) − Ke−r(T −t) ;

the right hand side equals S(t) − K for r = 0 and is strictly greater than this quantity
for r > 0. As S(t) − K = (S(t) − K)+ for S(t) ≥ K, the claim follows.
(vi) Consider a portfolio A which is long one call with maturity T2 and strike K, and short
one call with maturity T1 and strike K, where T2 > T1 ≥ t. By the claim (v) we have

C(T1 , S(T1 ), K, T2 ) ≥ (S(T1 ) − K)+ = C(T1 , S(T1 ), K, T1 ),

i.e., VA (T1 ) ≥ 0, for t < T1 . Hence VA (t) ≥ 0, i.e., C(t, S(t), K, T2 ) ≥ C(t, S(t), K, T1 ),
which is the claim.
(vii) We prove the statement for call options, the argument for put options being the same.
Let K0 , K1 > 0 and 0 < θ < 1 be given. Consider a portfolio A which is short one
share of a call with strike θK1 + (1 − θ)K0 and maturity T , long θ shares of a call with
strike K1 and maturity T , long (1 − θ) shares of a call with strike K0 and maturity T .
The value of this portfolio at maturity is

VA (T ) = −(S(T ) − (θK1 + (1 − θ)K0 ))+ + θ(S(T ) − K1 )+ + (1 − θ)(S(T ) − K0 )+ .

The convexity of the function f (x) = (S(T ) − x)+ gives VA (T ) ≥ 0 and so VA (t) ≥ 0
by the dominance principle. The latter inequality is

C(t, S(t), θK1 + (1 − θ)K0 , T ) ≤ θC(t, S(t), K1 , T ) + (1 − θ)C(t, S(t), K0 , T ),

which is the claim for call options.

Exercise 1.5. Consider the following table of European options prices at time t = 0:

CALL PUT
Maturity Strike Price Maturity Strike Price

1 month 104 20 1 month 96 14

1 month 110 16 1 month 100 16
1 month 116 10 1 month 104 18
3 month 110 15 3 month 110 18

21
Assume that the money market has interest rate r = 0 and that the price of the underlying
asset at time t = 0 is S(0) = 100. Explain why these prices are incompatible with the dom-
inance principle. Find a constant portfolio position which violates the dominance principle.
HINT: Look for violations of the properties (iv)–(vii).

Exercise 1.6. Assume that the dominance principle holds and prove the following.

(viii) If K0 ≤ K1 , then C(t, S(t), K0 , T ) ≥ C(t, S(t), K1 , T ), i.e., the price of European call
options is non-increasing with the strike price. Similarly the price of put options is
non-decreasing with the strike price.

(ix) C(t, S(t), K, T ) ≤ S(t) and P (t, S(t), K, T ) ≤ Ke−r(T −t) .

Exercise 1.7 (•). Assume that the dominance principle holds. Consider the European
derivative U with maturity time T and pay-off Y given by Y = min[(S(T ) − K1 )+ , (K2 −
S(T ))+ ], where K2 > K1 and (x)+ = max(0, x). Draw the pay-off function of the derivative.
Find a constant portfolio consisting of European calls and puts expiring at time T which
replicates the value of U (i.e., whose value at any time t < T equals the value of U).

Exercise 1.8. Suppose K, ∆K > 0. A butterfly spread on call options pays the amount

(S(T ) − K + ∆K)+ − 2(S(T ) − K)+ + (S(T ) − K − ∆K)+

at the maturity T . Draw the pay-off function of the derivative. Show that the value of this
option is non-negative at any point of time.

Exercise 1.9 (•). The price of a contract at time t is N units of currency and it pays at
the maturity date T > t the amount N + αN (S(T ) − K)+ . Show that

1 − e−r(T −t)
α= (1.11)
C(t, S(t), K, T )

if C(t, S(t), K, T ) > 0 and N 6= 0.

Optimal exercise time of American options

Consider now a no-dummy investor owning an American put option. When should the
investor exercise the option? At any time t < T we have, by (iii),

either Pb(t, S(t), K, T ) > (K − S(t))+ or Pb(t, S(t), K, T ) = (K − S(t))+ .

Exercising the American put at a time t when the strict inequality Pb(t, S(t), K, T ) > (K −
S(t))+ holds is a dummy decision, because the resulting pay-off is lower than the value of

22
the derivative11 . On the other hand, if the equality Pb(t, S(t), K, T ) = (K − S(t))+ holds at
time t, then the optimal strategy for the investor is to exercise the American put, as in this
case the pay-off equals the value of the derivative, i.e., the investor takes full advantage of
the American put. This leads us to introduce the following definition.
Definition 1.1. A time t < T is called an optimal exercise time for the American put
with value Pb(t, S(t), K, T ) if

Pb(t, S(t), K, T ) = (K − S(t))+ .

A similar definition can be justified for American call options, i.e., the optimal exercise time
b S(t), K, T ) = (S(t) − K)+ . However, assuming
of the American call is a time t at which C(t,
that the dominance principle holds (and that the money market has positive interest rate),
b S(t), K, T ) ≥ C(t, S(t), K, T ) > (S(t) − K)+ , for t < T , see (v) in Theorem 1.1.
we have C(t,
It follows that, in an arbitrage-free market, it is never optimal to exercise American call
options prior to maturity when the underlying stock pays no dividend. As opposed to this,
it will be shown in Chapter 6 that when the underlying stock pays a dividend prior to the
expiration date of the American call, it is optimal to exercise the American call just before
the dividend is paid, provided the price of the stock is sufficiently high, see Theorem 6.9.
As in the absence of dividends the optimal strategy is to hold the American call until
maturity, the dominance principle leads us to the following, last property on the fair price
of options:

(x) When the underlying stock pays no dividend (and the money market has positive
interest rate), the fair price of European calls and American call with equal parameters
are the same, i.e.,
C(t,
b S(t), K, T ) = C(t, S(t), K, T ).

Final remarks: The properties (i)-(x) are quite well represented in real markets, thereby
giving indirect support to the arbitrage-free principle. Note also that these properties depend
only on the validity of the arbitrage-free principle (in the form of the dominance principle)
and not on the specific market dynamics. In the following chapters we shall give an alternative
proof of (some of) these properties by using explicit models for the price of stocks and options
in the market.
Exercise 1.10 (Comparison with market data). Call and put options on Nasdaq 100
are of European style and so they can be used to test the properties derived in this section.
The market price of call and put options on Nasdaq 100, for different strikes and maturities,
can be found at the homepage http: // www. marketwatch. com/ investing/ index/ ndx/
options . Compile a table of prices for options “nearly at the money”, for instance using
the first 10 options in and out of the money (if an option appears with zero price it means
11
Of course the buyer might want to close the position on the American put for other reasons. In this case
however it is more profitable to sell the option rather than exercising it.

23
that it has not yet been traded; you can just skip it). Use these data to plot (with Matlab
for instance) the price of call and put options in terms of the strike price and the time of
maturity. Are the properties (vi), (viii), (viii) verified? Next use the put-call parity identity
to compute the value of the interest rate r for each pair of call-put with the same strike and
maturity. What can you conclude? Do the data support the put-call parity?

24
Chapter 2

Binomial markets

In this and the following two chapters we present a time-discrete model for the fair price
of options first proposed in [4] and which is known under the name of binomial options
pricing model. The model is very popular among practitioners due to its implementation
simplicity. The present chapter deals with the dynamics of the underlying asset, which
we assume to be a stock. The following two chapters are concerned with European and
American derivatives on the stock. The time-continuum analogue of the binomial model is
the Black-Scholes model, which will be studied in Chapter 6.

2.1 The binomial stock price

The binomial asset price is a model for the evolution in time of the price of financial
assets. It is often applied to stocks, hence we denote by S(t) the price of the asset at time
t. We are interested in monitoring the stock price in some finite time interval [0, T ], where
T > 0 could be for instance the expiration date of an option on the stock. The price of the
stock at time t = 0 is denoted by S(0) or S0 and is assumed to be known.
The binomial stock price can only change at some given pre-defined times 0 = t0 < t1 <
t2 < · · · < tN = T ; moreover the price at time ti+1 depends only on the price at time ti and
the result of “tossing a coin”. Precisely, letting u, d ∈ R, u > d, and p ∈ (0, 1), we assume

S(ti−1 )eu , with probability p,
S(ti ) =
S(ti−1 )ed , with probability 1 − p,

for all i = 1, . . . , N . Here we may interpret p as the probability to get a head in a coin toss
(p = 1/2 for a fair coin). We restrict to the standard binomial model, which assumes that
the market parameters u, d, p are time-independent and that the stock pays no dividend in
the interval [0, T ]. In the applications one typically chooses u > 0 and d < 0 (e.g., d = −u
is quite common), hence u stands for “up”, since S(ti ) = S(ti−1 )eu > S(ti−1 ), while d stands

25
for “down”, for S(ti ) = S(ti−1 )ed < S(ti−1 ). In the first case we say that the stock price goes
up at time ti , in the second case that it goes down at time ti .
Next we introduce a number of assumptions which simplify the analysis of the model without
compromising its generality1 . Firstly we assume that the times t0 , t1 , . . . tN are equidistant,
that is
ti − ti−1 = h > 0, for all i = 1, . . . , n.
In the applications the value of h must be chosen much smaller than T . Without loss of
generality we can pick h = 1, and so

t1 = 1, t2 = 2, . . . , tN = T = N,

with N >> 1. For instance, if N = 67 (the number of trading days in a period of 3 months),
then h = 1 day and S(t), for t ∈ {1, . . . N }, may refer to the closing price of the stock at
each day. It is convenient to denote

I = {1, . . . , N }.

Hence, from now on, we assume that the binomial stock price is determined by the rule
S(0) = S0 and

S(t − 1)eu , with probability p

(
S(t) = , t ∈ I. (2.1)
S(t − 1)ed , with probability 1 − p

Remark 2.1 (Notation). The notation used in the present notes is the same as in [3],
although it is slightly different from the one used in the standard literature on the binomial
model, see e.g., [6]. In fact the binomial stock price is more commonly written as
(
S(t − 1)u, with probability p
S(t) = ,
S(t − 1)d, with probability 1 − p

with 0 < d < u. All the results in the present text can be translated into the standard
notation by the substitutions eu → u, ed → d. In our notation the log-returns of the stock
take a slightly simpler form, which is useful when passing to the time-continuum limit (see
Section 6.1).

Each possible sequence (S(1), . . . , S(N )) of the future stock prices determined by the bino-
mial model is called a path of the stock price. Clearly, there exists 2N possible paths of the
stock price in a N -period model. Letting2

{u, d}N = {x = (x1 , x2 , . . . , xN ) ∈ RN : xt = u or xt = d, t ∈ I}

1
We come back to the general model in Section 2.4, where it is implemented with Matlab.
2
Note carefully that in the set {u, d}N , the letters u, d mean “up” and “down” and should not be confused
with the numerical parameters u, d in the binomial stock price (2.1).

26
be the space of all possible N -sequences of “ups” and “downs”, we obtain a unique path
of the stock price (S(1), . . . , S(N )) for each x ∈ {u, d}N . For instance, for N = 3 and
x = (u, u, u) the corresponding stock price path is given by

S0 → S(1) = S0 eu → S(2) = S(1)eu = S0 e2u → S(3) = S(2)eu = S0 e3u ,

while for x = (u, d, u) we obtain the path

S0 → S(1) = S0 eu → S(2) = S(1)ed = S0 eu+d → S(3) = S(2)eu = S0 e2u+d .

In general the possible paths of the stock price for the 3-period model can be represented as

S(3)
6
= S0 e3u
u

S(2)
6
= S0 e2u
u d
(
S(1) = S0 eu S(3) = S0 e2u+d
6 6
u d u
(
S(0) = S0 S(2) = S0 eu+d
6
d u d
( (
d
S(1) = S0 e S(3) = S0 eu+2d
6
d u
(
S(2) = S0 e2d
d
(
S(3) = S0 e3d

which is an example of binomial tree. The admissible values for the binomial stock price
S(t) at time t are given by

S(t) ∈ {S0 eku+(t−k)d , k = 0, . . . , t},

for all t ∈ I, where k is the number of times that the price goes up up to the time t included.

Definition 2.1. Given x = (x1 , . . . xN ) ∈ {u, d}N , the binomial stock price S(t, x) at time
t ∈ I corresponding to x is given by

S(t, x) = S0 exp(x1 + x2 + . . . xt ).

27
The vector S x = (S(1, x), . . . S(N, x)) is called the x-path of the binomial stock price. More-
over we define the probability3 of the path S x as

P(S x ) = pNu (x) (1 − p)Nd (x) ,

where Nu (x) is the number of u’s in the sequence x and Nd (x) = N − Nu (x) is the number
of d’s. The probability that the binomial stock price follows one of the two paths S x , S y is
given by P(S x ) + P(S y ) (and similarly for any number of paths).
Theorem 2.1. X
P(S x ) = 1.
x∈{u,d}N

Proof. The sum is

 Nu (x)
X
x
X
Nu (x) Nd (x) N
X p
P(S ) = p (1 − p) = (1 − p) ,
1−p
x∈{u,d}N x∈{u,d}N x∈{u,d}N

where for the last equality we used Nd (x) = N − Nu (x). Now, even though the sum is over
x ∈ {u, d}N , the terms to be summed depend only on Nu (x). Hence, the paths x with the
same value of Nu (x) (i.e., the same number of “ups”) give the same contribution to the sum.
It is left as an exercise to show that given k ∈ {0, 1, . . . , N }, the number of paths x for which
Nu (x) = k is given by the binomial coefficient:
 
N N!
= .
k k!(N − k)!
k

of the term in the sum for which Nu (x) = k equals (p/(1 − p)) , hence, as there are
Each
N
k
of them, we obtain
N   k
X
x
X N p N
P(S ) = (1 − p) .
N k=0
k 1−p
x∈{u,d}

The binomial theorem asserts that (a + b)N = N N

P
k N −k
k=0 k a b . Letting a = p/(1 − p)
and b = 1 we conclude
 N
X
x N p
P(S ) = (1 − p) 1+ = 1.
N
1−p
x∈{u,d}

3
We shall say more about the probabilistic interpretation of the binomial model in Chapter 5.

28
2.2 Binomial markets
A 1+1 dimensional binomial market is a market that consists of a risky asset, say a stock,
and a risk-free asset, such that the price of the risky asset is given by the binomial model (2.1)
and the value of the risk-free asset at time t ∈ I is given by

B(t) = B0 ert , t ∈ I, (2.2)

where B0 = B(0) is the present (at time t = 0) value of the asset. Recall that we assume that
the interest rate r of the money market is constant (not necessarily positive). As we shall
only consider markets with one stock and one risk-free asset, we refer to 1+1 dimensional
binomial markets simply as binomial markets. The constants u, d, r, p are called market
parameters.

Remark 2.2 (Notation). As a follow-up to Remark 2.1, we mention that our notation for
the value of the risk-free asset is also slightly different from the standard one. In fact, most
of the literature on the binomial model, e.g. [6], denotes the value of the risk-free asset at
time t by B(t) = B0 (1 + r)t . To translate our results into the standard notation one just
needs to replace er with (1 + r). As already reported in Remark 2.1, we follow the notation
used in [3].

Remark 2.3 (Discounted price). The discounted price of the stock in a binomial market
is defined by S ∗ (t) = e−rt S(t) and has the following meaning: S ∗ (t) is the amount that
should be invested on the risk-free asset at time t = 0 in order that the value at time t
of this investment replicates the value of the stock at time t. Note that, whenever r > 0,
i.e., as long as buying the risk-free asset ensures a positive return, we have S ∗ (t) < S(t).
The discounted price of the stock measures the loss in the stock value due to the “time-
devaluation” of money expressed by the ratio B0 /B(t) = e−rt . Put it differently, 1 dollar
today (t = 0) has the same (financial) purchasing power of ert dollars at time t in the future.
Hence S ∗ (t) = S(t)/ert measures the future value of the stock relative to the present value
of money.

Definition 2.2. A portfolio process invested in a binomial market is a finite sequence

{(hS (t), hB (t))}t∈I ,

where4 hS (t) ∈ R is the number of shares invested in the stock and hB (t) ∈ R is the number
of shares invested in the risk-free asset during the time interval (t − 1, t], t ∈ I. The value
of the portfolio process at time t is given by

V (t) = hS (t)S(t) + hB (t)B(t). (2.3)

4
Recall that we assume that it is possible to trade any fractional number of shares, hence hS (t), hB (t)
can be any real number. Of course, in the applications they have to be rounded to integer numbers.

29
The initial position of the investor is given by (hS (0), hB (0)). Without loss of generality we
assume that
hS (0) = hS (1), hB (0) = hB (1), (2.4)
i.e, (hS (1), hB (1)) is the investor position on the closed interval [0, 1] (and not just in the semi-
open interval (0, 1]). We recall that hS (t) > 0 means that the investor has a long position
on the stock in the interval (t − 1, t], while hS (t) < 0 corresponds to a short position.
It is clear that the investor will change the position on the stock and the risk-free asset
according to the path followed by the stock price, and so (hS (t), hB (t)) is in general path-
dependent. When we want to emphasize the dependence of a portfolio position on the path
of the stock price we shall write

hS (t) = hS (t, x), hB (t) = hB (t, x).

Similarly, we write V (t) = V (t, x) if we want to emphasize the dependence of the portfolio
value on the path of the stock price. Clearly,

V (t, x) = hS (t, x)S(t, x) + hB (t, x)B(t).

We say that the portfolio process is self-financing if purchasing more shares of one asset
is possible only by selling shares of the other asset for an equivalent value (and not by
infusing new cash into the portfolio), and, conversely, if any cash obtained by selling one
asset is immediately re-invested to buy shares of the other asset (and not withdrawn from the
portfolio). To translate this condition into a mathematical formula, recall that (hS (t), hB (t))
is the investor position on the stock and the risk-free asset during the time interval (t − 1, t].
Assume that the investor sells/buys shares of the assets at time t. Let (hS (t + 1), hB (t + 1))
be the new position on the stock and the risk-free asset in the interval (t, t + 1]. The value
of the portfolio process just before and immediately after changing the position at time t is
given respectively by

V (t) = hS (t)S(t) + hB (t)B(t), V 0 (t) = hS (t + 1)S(t) + hB (t + 1)B(t).

The difference V 0 (t) − V (t), if not zero, corresponds to cash withdrawn or added to the
portfolio as a result of the change in the position on the assets. In a self-financing portfolio,
however, this difference must be zero. We thus must have V 0 (t) = V (t), which leads to the
following definition.
Definition 2.3. A portfolio process {(hS (t), hB (t))}t∈I invested in a binomial market is said
to be self-financing if

hS (t)S(t − 1) + hB (t)B(t − 1) = hS (t − 1)S(t − 1) + hB (t − 1)B(t − 1) (2.5)

holds for all t ∈ I.

Remark 2.4. Constant (i.e., time-independent) portfolio processes are clearly self-financing.
In fact, (2.5) holds if hS (t) = hS (t − 1) and hB (t) = hB (t − 1), for all t ∈ I.

30
Exercise 2.1. Denote ∆f = f (t) − f (t − 1), for any function f of t ∈ I. Show that for any
self-financing portfolio there holds

∆V (t) = hS (t)∆S(t) + hB (t)∆B(t).

Interpret the result.

Our next purpose is to show that the value of a self-financing portfolio at time t = N
determines uniquely the value of the portfolio at any earlier time t = 0, . . . , N − 1. This
result is crucial to justify our definition of binomial fair price of European derivatives in the
next chapter. We begin by fixing some notation. First we define the parameters qu , qd as

er − ed eu − er
qu = , qd = 1 − qu = . (2.6)
eu − ed eu − ed
Note the (qu , qd ) is the unique solution of the linear system

qu + qd = 1, qu eu + qd ed = er . (2.7)

Now, given a self-financing portfolio process, we denote

V u (t) = hS (t)S(t − 1)eu + hB (t)B(t − 1)er ,

which is the value of the portfolio at time t assuming that the stock price goes up at time t,
and
V d (t) = hS (t)S(t − 1)ed + hB (t)B(t − 1)er
which is the value of the portfolio at time t, assuming that the stock price goes down at time
t. We can now prove the main result of this section.

Theorem 2.2. Let {(hS (t), hB (t))}t∈I be a self-financing portfolio process with value V (N )
at time t = N . The portfolio value V (t) at earlier times satisfies the following recurrence
formula:
V (t) = e−r [qu V u (t + 1) + qd V d (t + 1)], for t = 0, . . . , N − 1. (2.8)
Moreover
X
V (t) = e−r(N −t) qxt+1 · · · qxN V (N, x), for t = 0, . . . , N − 1. (2.9)
(xt+1 ,...,xN )∈{u,d}N −t

In particular, at time t = 0 we have

X
V (0) = e−rN (qu )Nu (x) (qd )Nd (x) V (N, x). (2.10)
x∈{u,d}N

31
Proof. Using the definition of V u (t), V d (t), the right hand side of (2.8) is

e−r [qu V u (t + 1) + qd V d (t + 1)] = e−r [qu (hS (t + 1)S(t)eu + hB (t + 1)B(t)er )

+ qd (hS (t + 1)S(t)ed + hB (t + 1)B(t)er )]
= e−r [hS (t + 1)S(t)(qu eu + qd ed ) + hB (t + 1)B(t)er (qu + qd )]
= hS (t + 1)S(t) + hB (t + 1)B(t), (2.11)

where in the last step we used (2.7). By definition of self-financing portfolio, the last member
in (2.11) equals V (t), and so (2.8) is proved.
It remains to establish (2.9); we argue by induction on t = 0, . . . , N − 1.
Step 1. We first prove (2.9) for t = N − 1. In this case the sum in the right hand side
of (2.9) is over two terms, one for which xN = u and one for which xN = d. Hence we have
to prove

V (N − 1) = e−r [qu V (N, (x1 , . . . , xN −1 , u)) + qd V (N, (x1 , . . . , xN −1 , d))]. (2.12)

In the right hand side of (2.12) we recognize

V (N, (x1 , . . . , xN −1 , u)) = hS (N )S(N − 1)eu + hB (N )B(N − 1)er = V u (N )

and

V (N, (x1 , . . . , xN −1 , d)) = hS (N )S(N − 1)ed + hB (N )B(N − 1)er = V d (N ),

and so (2.12) is equivalent to (2.8) at time t = N − 1. As (2.8) has already been established
for all times, the proof of (2.12) is completed.

Step 2. Now assume that (2.9) is true at time t + 1, i.e.,

X
V (t + 1) = e−r(N −t−1) qxt+2 · · · qxN V (N, x).
(xt+2 ,...xN )∈{u,d}N −t−1

Step 3. We now prove (2.9) at time t. By the induction hypothesis of step 2 we have
X
V u (t + 1) = e−r(N −t−1) qxt+2 · · · qxN V (N, x1 , . . . , xt , u, xt+2 , . . . , xN ),
(xt+2 ,...,xN )∈{u,d}N −t−1

X
V d (t + 1) = e−r(N −t−1) qxt+2 · · · qxN V (N, x1 , . . . , xt , d, xt+2 , . . . , xN ).
(xt+2 ,...,xN )∈{u,d}N −t−1

32
Replacing in (2.8) we obtain

X
V (t) = e−r(N −t) qu qxt+2 · · · qxN V (N, x1 , . . . , xt , u, xt+2 , . . . , xN )
(xt+2 ,...,xN )∈{u,d}N −t−1

X
+ qd qxt+2 · · · qxN V (N, x1 , . . . , xt , d, xt+2 , . . . , xN )
(xt+2 ,...,xN )∈{u,d}N −t−1
X
= e−r(N −t) qxt+1 · · · qxN V (N, x),
(xt+1 ,...xN )∈{u,d}N −t

which completes the proof.

Remark 2.5. As the sum in (2.9) is over (xt+1 , . . . , xN ) and V (N ) depends on the full path
x ∈ {u, d}N , then V (t) depends only on the first t steps of the path of the stock price, i.e.,

V (t) = V (t, x1 , x2 , . . . , xt ).

Example. We conclude this section with an example of application of (2.10). Assume

N = 2 and consider a binomial market with the following parameters:

ed = 2, eu = 3, er = 1.

The parameters (2.6) are given by

qu = −1, qd = 2.

The set of possible paths is

{u, d}2 = {(u, u), (u, d), (d, u), (d, d)}.

Hence the sum in (2.10) is over 4 terms. As e−rN = 1 in this example, we have

V (0) = qu2 V (2, (u, u)) + qu qd V (2, (u, d)) + qu qd V (2, (d, u)) + qd2 V (2, (d, d))
= V (2, (u, u)) − 2V (2, (u, d)) − 2V (2, (d, u)) + 4V (2, (d, d)).

Assuming for instance that the possible portfolio values at time t = 2 are

V (2, (u, u)) = 1, V (2, (u, d)) = 2, V (2, (d, u)) = −1, V (2, (d, d)) = 4,

we obtain that the portfolio value at time t = 0 is V (0) = 15.

Exercise 2.2 (?). Use the recurrence formula (2.8) to compute all possible values of V (t)
for t = 0, 1 in the example above.
Exercise 2.3 (?). Assume S0 = B0 = 1 in the example above and find a self-financing
portfolio process with value V (t).

33
2.3 Arbitrage portfolio
Our next purpose is to prove that the binomial market introduced in the previous section
is arbitrage free, provided its parameters satisfy d < r < u (no condition is required on the
probability p). To achieve this we first need to introduce the precise definition of arbitrage
portfolio.
Definition 2.4. A portfolio process {(hS (t), hB (t)}t∈I invested in a binomial market is called
an arbitrage portfolio process if its value V (t) satisfies

1) V (0) = 0;

2) V (N, x) ≥ 0, for all x ∈ {u, d}N ;

3) There exists y ∈ {u, d}N such that V (N, y) > 0.

We say that the binomial market is arbitrage-free if there exists no self-financing arbitrage
portfolio process invested in the stock and the risk-free asset.

Let us comment on the previous definition. Condition 1) means that no initial wealth is
required to set up the portfolio, i.e., the long and short positions on the two assets are
perfectly balanced. In particular, anyone can (in principle) open such portfolio (no initial
wealth is required). Condition 2) means that the investor is sure not to loose money with
this investment: regardless of the path followed by the stock price, the return of the portfolio
is always non-negative. Condition 3) means that there is a strictly positive probability to
make a profit, since along at least one path of the stock price the return of the portfolio is
strictly positive.
Theorem 2.3. The binomial market is arbitrage free if and only if d < r < u.

Proof. We divide the proof in two steps. In step 1 we prove the claim for the 1-period
model, i.e., N = 1. The generalization to the multiperiod model (N > 1) is carried out
in step 2. Note also the claim of the theorem is logically equivalent to the following: There
exists a self-financing arbitrage portfolio in the binomial market if and only if r ∈
/ (d, u). It
is the latter claim which is actually proved below.

Step 1: the 1-period model. Because of our convention (2.4), we can set

hS (0) = hS (1) = hS , hB (0) = hB (1) = hB ,

i.e., the portfolio position in the 1-period model is constant (and thus self-financing) over
the interval [0, 1]. The value of the portfolio at time t = 0 is

V (0) = hS S0 + hB B0 ,

34
while at time t = 1 it is given by one of the following:

V (1) = V (1, u) = hS S0 eu + hB B0 er ,

if the stock price goes up at time t = 1, or

V (1) = V (1, d) = hS S0 ed + hB B0 er ,

if the stock price goes down at time t = 1. Thus the portfolio is an arbitrage if V (0) = 0,
i.e.,
hS S0 + hB B0 = 0, (2.13)
if V (1) ≥ 0, i.e.,

hS S0 eu + hB B0 er ≥ 0, (2.14)
d r
hS S0 e + hB B0 e ≥ 0, (2.15)

and if at least one of the inequalities in (2.14)-(2.15) is strict. Now assume that (hS , hB )
is an arbitrage portfolio. From (2.13) we have hB B0 = −hS S0 and therefore (2.14)-(2.15)
become

hS S0 (eu − er ) ≥ 0, (2.16)
d r
hS S0 (e − e ) ≥ 0. (2.17)

Since at least one of the inequalities must be strict, then hS 6= 0. If hS > 0, then (2.16) gives
u ≥ r, while (2.17) gives d ≥ r. As u > d, the last two inequalities are equivalent to r ≤ d.
Similarly, for hS < 0 we obtain u ≤ r and d ≤ r which, again due to u > d, are equivalent to
r ≥ u. We conclude that the existence of an arbitrage portfolio implies r ≤ d or r ≥ u, that
is r ∈
/ (d, u). This proves that for r ∈ (d, u) there is no arbitrage portfolio in the one period
model, and thus the “if” part of the theorem is proved for N = 1. To prove the “only if”
part, i.e., the fact that r ∈ (d, u) is necessary for the absence of arbitrages, we construct an
arbitrage portfolio when r ∈ / (d, u). Assume r ≤ d. Let us pick hS = 1 and hB = −S0 /B0 .
Then V (0) = 0. Moreover (2.15) is trivially satisfied and, since u > d,

hS S0 eu + hB B0 er = S0 (eu − er ) > S0 (ed − er ) ≥ 0,

hence the inequality (2.14) is strict. This shows that one can construct an arbitrage portfolio
if r ≤ d and a similar argument can be used to find an arbitrage portfolio when r ≥ u. The
proof of the theorem for the 1-period model is complete.

Step 2: the multiperiod model. Let r ∈ / (d, u). As shown in the previous step, there
exists an arbitrage portfolio (hS , hB ) in the single period model. We can now build a self-
financing arbitrage portfolio process {(hS (t), hB (t))}t∈I for the multiperiod model by invest-
ing at time t = 1 the whole value of the portfolio (hS , hB ) in the risk-free asset. The value of

35
this portfolio process satisfies V (0) = 0 and V (N, x) = V (1, x)er(N −1) ≥ 0 along every path
x ∈ {u, d}N . Moreover, since (hS , hB ) is an arbitrage, then V (1, y) > 0 along some path
y ∈ {u, d}N and thus V (N, y) > 0. Hence the constructed self-financing portfolio process
{(hS (t), hB (t))}t∈I is an arbitrage and the “if” part of the theorem is proved. To prove the
“only if” part for the multiperiod model, we use that, by Theorem 2.2,
X
V (0) = e−rN (qu )Nu (x) (qd )Nd (x) V (N, x),
x∈{u,d}N

where qu , qd are given by (2.6), Nu (x) is the number of “ups” in x and Nd (x) = N − Nu (x)
the number of “downs”. Now, assume that the portfolio is an arbitrage. Then V (0) = 0
and V (N, x) ≥ 0. Of course, the above sum can be restricted to the paths along which
V (N, x) > 0, which exist since the portfolio is an arbitrage. But then the sum can be zero
only if either one of the factors qu , qd is zero, or if they have opposite sign. Since u > d, the
denominator in the expressions (2.6) is positive, hence
qu = 0, resp. qd = 0 ⇒ r = d, resp. u = r,
(qu > 0, qd < 0), resp. (qu < 0, qd > 0) ⇒ u < r, resp. r < d.
We conclude that the existence of a self-financing arbitrage portfolio entails r ∈
/ (d, u), which
completes the proof of the theorem.
Remark 2.6. The condition d < r < u is equivalent to require that the parameters qu , qd
given by (2.6) satisfy qu ∈ (0, 1) and qd = 1 − qu ∈ (0, 1), i.e., that the pair (qu , qd ) defines
a probability, which is called risk-neutral probability or martingale probability; the
reason for this terminology will become clear in Chapter 5.
Exercise 2.4. Find an arbitrage portfolio in the binomial market given in the example at
the end of Section 2.2.

2.4 Computation of the binomial stock price with Mat-

lab
In this section we learn how to compute the binomial stock price with Matlab. More precisely,
our goal is to construct a binomial tree for the stock price in some interval [0, T ], with T > 0
measured in fraction of years. Let us start by dividing the interval [0, T ] into N subintervals
of length h = T /N , i.e.,
0 = t1 < t2 < · · · < tN +1 = T, ti+1 = ti + h, i = 1, . . . , N.
Let S(i) = S(ti ). We define the binomial stock price on the given partition of the interval
[0, T ] as
S(i)eu , with probability p
(
S(i + 1) = , i ∈ I.
S(i)ed , with probability 1 − p

36
Now we assume that the smaller is h, the lower will be the change of the stock price on each
subinterval, i.e., the smaller are the parameters u, d. More precisely, we assume that the
parameters u and d can be written in the form
1 − p√ p √
r r
u = αh + σ h, d = αh − σ h, (2.18)
p 1−p
for some constant parameters σ > 0 and α ∈ R, which we call, respectively, the instanta-
neous mean of log-return and the instantaneous volatility of the binomial stock price
in the interval [0, T ]. The reason to choose the parameters u, d in this form will become clear
in Chapter 6, where we prove that this choice makes the binomial stock price converge (in
distribution) to the geometric Brownian motion in the continuum limit h → 0. Note also
that, by inverting (2.18),
1 u − dp
α= [pu + (1 − p)d], σ= √ p(1 − p). (2.19)
h h
The parameter α measures the average movement of the binomial stock price in the interval
[0, T ], while σ measures the average amplitude of the oscillations of the binomial stock price.
The quantity σ 2 is called instantaneous variance.
The following code defines the Matlab function BinomialStock which generates the binomial
tree for the stock price on the partition Q = {t1 , . . . tN +1 } of the interval [0, T ]:

function [Q,S]=BinomialStock(p,alpha,sigma,s,T,N)
h=T/N;
u=alpha*h+sigma*sqrt(h)*sqrt((1-p)/p);
d=alpha*h-sigma*sqrt(h)*sqrt(p/(1-p));
Q=zeros(N+1,1);
S=zeros(N+1);
Q(1)=0;
S(1,1)=s;
for j=1:N
Q(j+1)=j*h;
S(1,j+1)=S(1,j)*exp(u);
for i=1:j
S(i+1,j+1)=S(i,j)*exp(d);
end
end

The arguments of the function are the parameters p, α, σ, the time T > 0 (expressed in
fraction of years), the initial price of the stock s and the number of steps N in the binomial
model. The function returns a column vector Q containing the times t1 , . . . tN +1 of the
partition, and an upper-triangular (N + 1) × (N + 1) matrix S. The column j of S contains,
in decreasing order along rows, the possible prices of the stock at time tj = (j − 1)h. A path

37
of the stock price is obtained by moving from each column to the next one by either staying
in the same row (which means that the price went up at this step) or going down one row
(which means that the price went down at this step). For example, by running the command

[Q,S]=BinomialStock(0.5,0.01,0.2,10,1/12,5);

we get the output

0
0.0167
0.0333
Q=
0.0500
0.0667
0.0833

10.0000 10.2633 10.5335 10.8108 11.0954 11.3875

0 9.7467 10.0033 10.2667 10.5370 10.8144
0 0 9.4999 9.7500 10.0067 10.2701
S= (2.20)
0 0 0 9.2593 9.5030 9.7532
0 0 0 0 9.0248 9.2624
0 0 0 0 0 8.7962
As T = 1/12 years and N = 5, the matrix S shows the stock prices in a period of 1 month
with intervals of 1/(12 ∗ 5) years, i.e., about 6 days5 . The mean of log-return of the stock is
α = 1%, while its volatility is σ = 20%.

Remark 2.7. Throughout these notes, the computations in Matlab are performed in high
precision arithmetic and the results are truncated to the fourth decimal digit. In the appli-
cations, where the results correspond to prices expressed in some unit of currency, a cruder
truncation may be required.

Exercise 2.5. Define a function RandomPath(S) that generates a random path from the
binomial tree S created with the function BinomialStock and compute its probability. Plot
the price of the stock as a function of time.

Recall that in the applications to real-word problems the number N should be chosen suffi-
ciently large (i.e., h should be small compared to T ), which makes it practically impossible
to generate all possible 2N paths of the stock price. Even for a relatively small number of
steps, like N ≈ 50, the number of admissible paths is too big. Note also that the probability
of each single path is practically zero for N & 20. For instance, if we assume p = 1/2, then
each single path has equal probability P(S x ) = (1/2)N , which is approximately 10−4 % for
N = 20. The following code defines a Matlab function ProbStock(S,t,a,b,p) which computes
5
We do not adjust our calculations to take into account that markets are closed in the week-ends.

38
the (much more interesting) probability that the price S(t) lies within a given interval (a, b),
i.e., P(a < S(t) < b).

function Prob=ProbStock(S,t,a,b,p)
ind=find(S(:,t+1)>a & S(:,t+1)<b);
numups=t+1-ind;
k=length(numups);
probpath=0;
for j= 1:k
probpath=probpath+nchoosek(t,numups(j))*p^numups(j)*(1-p)^(t-numups(j));
end
Prob=probpath;

The function computes the probability P(a < S(t) < b) using Definition 2.1, i.e., by summing
up the probability of each path which leads to a price S(t) ∈ (a, b). As Sij in the binomial
tree S is the price at time j assuming that the price goes down i − 1 times, the number of
ups at time t = j is given by j + 1 − i. The variable numups in the code above contains
therefore the number of “ups” for each price at time t which lies in the interval (a, b).

39
Chapter 3

European derivatives

This chapter deals with the pricing theory of European derivatives on a single stock under
the assumption that the price of the underlying is given by S(0) = S0 at time t = 0 and

S(t − 1)eu with probability p
S(t) = , t ∈ I = {1, . . . , N }.
S(t − 1)ed with probability 1 − p
Here 0 < p < 1 and u > d. S(t) is the binomial stock price at time t ∈ I. It is assumed that
the initial price S0 is known. We say that the price goes up at time t if S(t) = S(t − 1)eu
and that it goes down at time t if S(t) = S(t − 1)ed (although this terminology is strictly
correct only if d < 0 and u > 0, which is most often the case in the applications). Moreover
we assume the existence of a risk-free asset with value
B(t) = B0 ert , t ∈ I,
where B0 = B(0) is the initial value of the risk-free asset and r is the constant interest rate
of the money market. We impose d < r < u. In particular, the market is arbitrage-free and
the risk-neutral probability is well defined:
er − ed eu − er
qu = , qd = , qu , qd ∈ (0, 1), qu + qd = 1.
eu − ed eu − ed
We denote by {(hS (t), hB (t)}t∈I a portfolio process invested in the stock and the risk-free
asset, where (hS (t), hB (t)) is the portfolio position in the interval (t−1, t] and hS (0) = hS (1),
hB (0) = hB (1). The value at time t of the portfolio is V (t) = hS (t)S(t) + hB (t)B(t). Recall
that the portfolio process is said to be self-financing if
V (t − 1) = hS (t)S(t − 1) + hB (t)B(t − 1),
which means that no cash is ever withdrawn or added to the portfolio. The value of a self-
financing portfolio process at time t is uniquely determined by its value at time N . In fact
we have the formula
X
V (t) = e−r(N −t) qxt+1 · · · qxN V (N, x). (3.1)
(xt+1 ,...xN )∈{u,d}N −t

40
which we proved in Theorem 2.2.
Definition 3.1. A portfolio process {(hS (t), hB (t)}t∈I is called predictable if there exist N
functions H1 , . . . HN such that Ht : (0, ∞)t → R2 and
(hS (t), hB (t)) = Ht (S0 , . . . , S(t − 1)), t ∈ I.

In particular, for a predictable portfolio process, the position (hS (t), hB (t)) is a determinis-
tic function of the stock price up to time t − 1 and thus is depends on the path x ∈ {u, d}N
only through the stock prices:
(hS (t, x), hB (t, x)) = Ht (S0 , S(1, x), . . . , S(t − 1, x)).
As S(t, x) = S0 exp(x1 + · · · + xt ) is independent of the steps after time t, we conclude that
the position of a predictable portfolio process in the interval (t − 1, t] is determined by the
information available at time t − 1, which is the reason for calling it “predictable”.

3.1 The binomial price of European derivatives

Consider a European derivative on the stock expiring at time T = N . The derivative is called
standard if its pay-off depends only on the price of the stock at maturity, i.e., Y = g(S(N )),
for some function g : (0, ∞) → R, which is called the pay-off function of the derivative.
The European call, for instance, is a standard European derivative with pay-off function
g(z) = (z − K)+ , where K > 0 is the strike price of the call. In the case of non-standard
European derivatives, the pay-off is a function of the stock price at all times earlier and
including N , i.e., Y = g(S(1), . . . , S(N )). Of course, in both cases the pay-off depends on
the path followed by the stock price, in particular,
Y (x) = g(S(N, x)), for standard European derivatives
and
Y (x) = g(S(1, x), S(2, x), . . . , S(N, x)), for non-standard European derivatives.
Now, assume that a European derivative is sold at time t < T for the price ΠY (t). The first
concern of the seller is to hedge the derivative, i.e., to invest the premium ΠY (t) in such
a way that the seller portfolio value at the expiration date is enough to pay-off the buyer
of the derivative. We assume that the seller invests the premium in the 1+1 dimensional
binomial market consisting of the underlying stock and the risk-free asset (delta-hedging).
Definition 3.2. An hedging portfolio process for the European derivative with pay-off Y
and maturity T = N is a portfolio process {(hS (t), hB (t))}t∈I invested in the underlying stock
and the risk-free asset such that its value V (t) satisfies V (N ) = Y ; the latter equality must
be satisfied for all possible paths of the price of the underlying stock, i.e., V (N, x) = Y (x),
for all x ∈ {u, d}N .

41
It follows by (3.1) that the value V (t) of any self-financing hedging portfolio at time t is
given by X
V (t) = e−r(N −t) qxt+1 · · · qxN Y (x1 , . . . , xN ). (3.2)
(xt+1 ,...xN )∈{u,d}N −t

Definition 3.3. The binomial (fair) price of the European derivative with pay-off Y and
maturity T = N is given by
X
ΠY (t) := e−r(N −t) qxt+1 · · · qxN Y (x1 , . . . , xN ), (3.3)
(xt+1 ,...xN )∈{u,d}N −t

for t = 0, . . . , N − 1, while ΠY (N ) := Y . In other words, the binomial price of the European

derivative equals the value of self-financing portfolios hedging the derivative.

Remark 3.1. The identification of the fair price of the derivative with the value of self-
financing hedging portfolios can be motivated as follows. Firstly the only purpose of the
hedging portfolio is to pay-off the buyer of the derivative; the seller does not try to ensure a
profit from the derivative (as it would be if V (N ) > Y ). The second reason is the absence
of cash flow. In fact, if the writer needs to add cash to the portfolio in order to hedge the
derivative, then the contract is clearly unfair for the writer. Conversely, if the writer could
withdraw cash from the portfolio and still be able to hedge the derivative, then the contract
would be clearly unfair for the buyer1 .

Remark 3.2. Note carefully that we have not proved yet that hedging self-financing port-
folios exist. However we know that, if they exist, their value at time t is given by (3.2). The
existence of self-financing hedging portfolios is proved in Theorem 3.3 below.

Remark 3.3. By definition, hedging portfolios of European derivatives are also replicating
portfolios, i.e., the equality V (t) = ΠY (t) holds at all time (and not only at maturity). We
shall see in Chapter 4 that this is not the case for American derivatives.

By Remark 2.5, the value ΠY (t) depends on the path of the stock price up to time t, i.e.,

ΠY (t) = ΠY (t, x1 , . . . , xt ).

Hence the binomial price at time t of European derivatives can be computed using the
information available up to time t. For example, since by Definition 2.1 we have

S(N, x) = S0 exp(x1 + · · · + xN ), S(t, x) = S0 exp(x1 + · · · + xt )

then
S(N, x) = S(t) exp(xt+1 + · · · + xN ),
1
We take the opportunity to make the obvious remark that our theory treats the buyer and the seller on
equal foot.

42
and therefore the binomial price for the standard European derivative with pay-off Y =
g(S(N )) can be written as
X
ΠY (t) = e−r(N −t) qxt+1 · · · qxN g(S(t) exp(xt+1 + · · · + xN )). (3.4)
(xt+1 ,...xN )∈{u,d}N −t

This shows that the binomial price at time t of standard European derivatives is a deter-
ministic function of S(t), namely

ΠY (t) = fg (t, S(t)), (3.5)

where
X
fg (t, z) = e−r(N −t) qxt+1 · · · qxN g(z exp(xt+1 + · · · + xN )).
(xt+1 ,...xN )∈{u,d}N −t

In the particular case of the European call (resp. put) option with strike K and maturity T =
N , the binomial fair price at time t = 0, . . . , N −1 can be written in the form C(t, S(t), K, N )
(resp. P (t, S(t), K, N )), where
X
C(t, S(t), K, N ) = e−r(N −t) qxt+1 · · · qxN (S(t) exp(xt+1 + · · · + xN ) − K)+ ,
(xt+1 ,...xN )∈{u,d}N −t
(3.6)
X
P (t, S(t), K, N ) = e−r(N −t) qxt+1 · · · qxN (K − S(t) exp(xt+1 + · · · + xN ))+ .
(xt+1 ,...xN )∈{u,d}N −t
(3.7)

We use these formula to give an alternative proof of Theorem 1.2, which does not invoke the
dominance principle.

Theorem 3.1 (∗). The binomial price of European calls and puts satisfy the properties in
Theorem 1.2, that is

1 The put-call parity holds

S(t) − C(t, S(t), K, N ) = Ke−r(N −t) − P (t, S(t), K, N ).

2 If r ≥ 0, then C(t, S(t), K, N ) ≥ (S(t) − K)+ ; the strict inequality C(t, S(t), K, N ) >
(S(t) − K)+ holds when r > 0.

3 If r ≥ 0, the map N → C(t, S(t), K, N ) is non-decreasing.

4 The maps K → C(t, S(t), K, N ) and K → P (t, S(t), K, N ) are convex.

43
Proof. Let k be the number of u in (xt+1 , . . . , xN ) and N − t − k be the number of d. Then
ext+1 +···+xN = eku+(N −t−k)d , qxt+1 · · · qxN = quk qdN −t−k .
Now, the number of paths
(xt+1 , . . . , xN ) for which the number of u is k is given by the
binomial coefficient Nk−t . Hence we can rewrite (3.6)-(3.7) as
N −t  
−r(N −t)
X N − t k N −t−k
C(t, S(t), K, N ) = e qu qd (S(t)eku+(N −t−k)d − K)+ ,
k=0
k
N −t  
−r(N −t)
X N − t k N −t−k
P (t, S(t), K, N ) = e qu qd (K − S(t)eku+(N −t−k)d )+ .
k=0
k
We can now prove the properties 1-4.

1 We have
N −t  
−r(N −t)
X N − t k N −t−k
C(t, S(t), K, N ) − P (t, S(t), K, N ) = e qu qd
k=0
k
× [(S(t)eku+(N −t−k)d − K)+ − (K − S(t)eku+(N −t−k)d )+ ].
Using that (z − K)+ − (K − z)+ = z − K, for all z ∈ R, we obtain
N −t  
−r(N −t)
X N − t k N −t−k
C −P =e q u qd (S(t)eku+(N −t−k)d − K)
k=0
k
N −t  
−r(N −t)
X N − t k N −t−k ku+(N −t−k)d
= S(t)e qu qd e
k=0
k
N −t  
X N − t k N −t−k
−K qu q d = S(t)I1 − KI2 .
k=0
k

Hence the put-call parity follows if we show that I2 = e−r(N −t) and I1 = 1. We have
N −t  
−r(N −t)
X N − t k N −t−k ku+(N −t−k)d
I1 = e qu qd e
k=0
k
N −t 
N − t  qu e u  k
X 
−r(N −t) d N −t
=e (qd e ) d
.
k=0
k q de

Using the binomial theorem (1 + a)N = N N

P
k u d r
k=0 k a and the identity qu e + qd e = e
we obtain
 qu eu N −t
I1 = e−r(N −t) (qd ed )N −t 1 +
qd ed
= e−r(N −t) (qu eu + qd ed )N −t = 1.
The proof that I2 = e−r(N −t) is similar.

44
2 The proof follows by the put-call parity as in Theorem 1.2.
3 We want to show that

C(t, S(t), K, N ) ≤ C(t, S(t), K, N + 1).

The following proof is quite technical. In C(N + 1) := C(t, S(t), K, N + 1) we replace

the Pascal identity      
N +1−t N −t N −t
= +
k k−1 k
N


(with the convention −1 = 0) and obtain
( N +1−t 
X N − t
−r(N +1−t)
C(N + 1) =e quk qdN +1−t−k (S(t)eku+(N +1−t−k)d − K)+
k=1
k − 1
NX
+1−t   )
N − t k N +1−t−k
+ qu q d (S(t)eku+(N +1−t−k)d − K)+ .
k=0
k

In the first sum we make the change of index j = k − 1, while for the second sum we
use that it is greater than the sum extended only up to N − t (i.e., we neglect the last
term k = N + 1 − t). So doing we obtain
( N −t 
X N − t
C(N + 1) ≥ e −r(N +1−t)
quj+1 qdN −t−j (S(t)e(j+1)u+(N −t−j)d − K)+
j=0
j
−t
N
)
X N − t

+ quk qdN +1−t−k (S(t)eku+(N +1−t−k)d − K)+
k=0
k
( N −t 
X N − t N −t−j
−r(N +1−t)
=e quj qd qu (S(t)eju+(N −t−j)d eu − K)+
j=0
j
N −t   )
X N − t k N −t−k
+ qu qd qd (S(t)eku+(N −t−k)d ed − K)+
k=0
k
N −t  
−r(N +1−t)
X N − t k N −t−k
=e q u qd
k=0
k
× [(S(t)eku+(N −t−k)d qu eu − Kqu )+ + (S(t)eku+(N −t−k)d qd ed − Kqd )+ ].

Using the simple inequality (y)+ + (z)+ ≥ (y + z)+ , we obtain

N −t  
−r(N +1−t)
X N − t k N −t−k
C(N + 1) ≥ e qu qd
k=0
k
× [(S(t)eku+(N −t−k)d (qu eu + qd ed ) − K(qu + qd ))+ ].

45
 As qu eu + qd ed = er , qu + qd = 1 and r ≥ 0 we find
N −t  
−r(N +1−t)
X N − t k N −t−k
C(N + 1) ≥ e qu qd (S(t)eku+(N −t−k)d er − K)+
k=0
k
N −t
X N − t
=e −r(N −t)
quk qdN −t−k (S(t)eku+(N −t−k)d − Ke−r )+
k=0
k
≥ C(N ).

4 The only dependence on K of the functions C(t, S(t), K, N ), P (t, S(t), K, N ) is through
the terms (z−K)+ , (K −z)+ . As both these functions are convex in K (draw a picture),
the result follows.

Exercise 3.1. Show that the binomial price at time t of non-standard European derivatives
is a deterministic function of S(0), S(1), . . . , S(t).

Now let ΠuY (t) denote the binomial price of the European derivative at time t assuming
that the stock price goes up at time t (i.e., S(t) = S(t − 1)eu , or equivalently, xt = u),
and similarly define ΠdY (t), with “up” replaced by “down”. By the proven formula (2.8) in
Theorem 2.2 we have
Theorem 3.2. The binomial price of European derivatives satisfies the recurrence formula

ΠY (N ) = Y, and ΠY (t) = e−r [qu ΠuY (t+1)+qd ΠdY (t+1)], for t ∈ {0, . . . , N − 1}. (3.8)

The recurrence formula (3.8) is very useful to compute the binomial price of standard Euro-
pean derivatives, as shown in the next example.

3.1.1 Example: A standard European derivative

p
Consider the standard European derivative with pay-off Y = ( S(2) − 1)+ at maturity time
T = 2. Assume that the market parameters are given by

u = log 2, d = 0, r = log(4/3), p = 1/4.

Assume also S0 = 1. Compute the possible paths of the derivative price ΠY (t). Compute
also the probability that the derivative expires in the money.

Solution: The stock price and the risk-free asset satisfy


S(t − 1)eu
S(t) = , B(t) = B0 ert t ∈ {1, 2}
S(t − 1)ed

46
where
eu = 2, ed = 1, er = 4/3
Hence
er − ed 1 2
qu =u d
= , qd = 1 − qu = .
e −e 3 3
Now, let us write the binomial tree of the stock price, including the possible values of the
derivative at the expiration time T = 2 (where we use that ΠY (2) = Y ):
√
S(2)
2
= 4 ⇒ ΠY (2) = ( 4 − 1)+ = 1
u

S(1)
9
=2
u d
, √ √
S0 = 1 S(2) = 2 ⇒ ΠY (2) = ( 2 − 1)+ = 2 − 1
2
d u
%
S(1) = 1
d
, √
S(2) = 1 ⇒ ΠY (2) = ( 1 − 1)+ = 0

Using the recurrence formula

ΠY (t) = e−r (qu ΠuY (t + 1) + qd ΠdY (t + 1))

we have, at time t = 1,
3 1 2 √ 1 √
S(1) = 2 ⇒ ΠY (1) = ( · 1 + ( 2 − 1)) = (2 2 − 1)
4 3 3 4
3 1 √ 2 1 √
S(1) = 1 ⇒ ΠY (1) = ( ( 2 − 1) + · 0) = ( 2 − 1)
4 3 3 4

while at time t = 0 we have

3 1 1 √ 2 1 √ 1 √ 3
ΠY (0) = ( · (2 2 − 1) + · ( 2 − 1)) = ( 2 − ).
4 3 4 3 4 4 4

47
Hence we have found the following diagram for the binomial price of the derivative

ΠY (2) = 1
4
u

√
ΠY (1) = 14 (2 2 − 1)
4
u d

√ * √
ΠY (0) = 14 ( 2 − 34 ) ΠY (2) = 2−1
4
d u
* √
ΠY (1) = 14 ( 2 − 1)
d
*
ΠY (2) = 0

As to the probability that the derivative expires in the money, i.e., P(Y > 0), we see from
the above diagram that this happens along the paths (u, u), (u, d), (d, u), hence
 2
(u,u) (u,d) (d,u) 1 1 3 3 1 7
P(Y > 0) = P(S ) + P(S ) + P(S )= + · + · = ,
4 4 4 4 4 16

which corresponds to 43, 75%.

Remark 3.4. Note carefully that the probability that the derivatives expires in the money
is computed with the physical probability p and not with the risk-neutral probability.

3.1.2 Example: A non-standard European derivative

Consider a 3-period binomial market with the following parameters:
4 2 3
eu = , ed = , p= .
3 3 4
Assuming S0 = 2 and r = 0, compute the initial binomial price of the European derivative
with pay-off
 11 
Y = − min(S0 , S(1), S(2), S(3)) , (z)+ = max(0, z),
9 +

and time of maturity T = 3. This is an example of lookback option. Compute the

probability that the derivative expires in the money and the probability that the return of a
constant portfolio with a long position on this derivative be positive.

48
Solution: We start by writing down the binomial tree of the stock price
128
S(3) =
8 27
u

32
S(2) =
8 9
u d
&
8 64
S(1) = S(3) =
8 3 8 27
u d u
&
16
S0 = 2 S(2) =
8 9
d u d
& &
4 32
S(1) = S(3) =
3 8 27
d u
&
8
S(2) = 9
d
&
16
S(3) = 27

To compute the initial binomial price of non-standard European derivatives it is convenient

to use the formula X
ΠY (0) = e−rN (qu )Nu (x) (qd )Nd (x) Y (x), (3.9)
x∈{u,d}N

where Y (x) denotes the pay-off as a function of the path of the stock price, Nu (x) is the
number of times that the stock price goes up in the path x and Nd (x) = N − Nu (x) is the
number of times that it goes down. In this example we have N = 3, r = 0 and
1
qu = qd = .
2
So, it remains to compute the pay-off for all possible paths of the binomial stock price, where
 
11
Y = − min(S0 , S(1), S(2), S(3)) , (z)+ = max(0, z).
9 +

For instance
   
11 11 7
Y (u, u, u) = − min(2, 8/3, 32/9, 128/27) = −2 = max(0, − ) = 0.
9 + 9 + 9

49
Similarly we find
 
11
Y (u, u, d) = − min(2, 8/3, 32/9, 64/27) =0
9 +
 
11
Y (u, d, u) = − min(2, 8/3, 16/9, 64/27) =0
9 +
 
11
Y (u, d, d) = − min(2, 8/3, 16/9, 32/27) = 1/27
9 +
 
11
Y (d, u, u) = − min(2, 4/3, 16/9, 64/27) =0
9 +
 
11
Y (d, u, d) = − min(2, 4/3, 16/9, 32/27) = 1/27
9 +
 
11
Y (d, d, u) = − min(2, 4/3, 8/9, 32/27) = 1/3
9 +
 
11
Y (d, d, d) = − min(2, 4/3, 8/9, 16/27) = 17/27
9 +

Replacing in (3.9) we obtain

ΠY (0) = qu (qd )2 Y (u, d, d) + (qd )2 qu Y (d, u, d) + (qd )2 qu Y (d, d, u) + (qd )3 Y (d, d, d),
the other terms being zero. Hence
1 1 1 1 17  7
ΠY (0) = + + + = .
8 27 27 3 27 54
The probability that the derivative expires in the money is the probability that Y > 0.
Hence we just sum the probabilities of the paths which lead to a positive pay-off:
P(Y > 0) = P(S (u,d,d) ) + P(S (d,u,d) ) + P(S (d,d,u) ) + P(S (d,d,d) )
= p(1 − p)2 + (1 − p)2 p + (1 − p)2 p + (1 − p)3
 2  3
2 3 1 3 1 5
= 3(1 − p) p + (1 − p) = 3 + = ≈ 15, 6%
4 4 4 32
Next consider a constant portfolio with a long position on the derivative. This means that
we buy the derivative at time t = 0 and we wait (without changing the portfolio) until the
expiration time t = 3. The return will be positive (i.e., the buyer makes a profit) if and only
if ΠY (3) > ΠY (0). But ΠY (3) = Y , which, according to the computations above, is greater
that ΠY (0) = 7/54 only when the binomial stock price follows one of the paths (d, d, u) or
(d, d, d). Hence
1
P[R > 0] = P(S (d,d,u) ) + P(S (d,d,d) ) = (1 − p)2 p + (1 − p)3 = (1 − p)2 = ≈ 6, 2%
16

50
Exercise 3.2 (•). A compound option is an option whose underlying is another option.
For instance, given T2 > T1 > 0 and K1 , K2 > 0, a call on a put with maturity T1 and
strike K1 is a contract that gives to its owner the right to buy at time T1 for the price K1
the put option on the stock with maturity T2 and strike K2 . Let S(t) be the price of the
underlying stock of the put option. Assume that S(t) follows a 2-period binomial model with
parameters
7 1 9 1
eu = , ed = , er = , p = , S(0) = 16.
4 2 8 4
23
Assume further that T2 = 2, T1 = 1, K1 = 9 , K2 = 12. Compute the initial price of the call
on the put. Compute also the probability of positive return for the owner of the call on the
put.
Exercise 3.3 (•). Suppose u > r > 0 ≥ d. A non-standard European derivative with
maturity time N has pay-off Y = S(N ) if S(0) < S(1) < · · · < S(N ) and Y = S(0)
otherwise. Find ΠY (0).
Exercise 3.4 (?). A barrier option is an option that expires worthless as soon as the
stock price exceeds (or fall below) a specified level (the barrier of the option2 ). For example,
consider the binomial market in Section 3.1.2 and the barrier call option with strike K = 2
and barrier B = 3. This means that the derivative expires worthless if S(3) ≤ 2 or if S(t) > 3
at some time t ∈ {1, 2, 3}. Compute the initial price ΠY (0) of this barrier option and the
probability that it expires in the money.
Exercise 3.5 (•). A European derivative with expiration T = N pays the amount Y =
log(S(T )/S(0)). Find ΠY (0). Hint: Use the identity
   
N N −1
k=N .
k k−1
ANSWER: ΠY (0) = T e−rT (qu u + qd d).
Exercise 3.6 (?). Can you think of a reason to buy a call on the put, a barrier option, or
a look-back option?

3.2 Hedging portfolio

Next we treat the important problem of building a self-financing hedging portfolio for Eu-
ropean derivatives.
Theorem 3.3. Consider the European derivative with pay-off Y at the time of maturity
T = N . Then the portfolio process given by
hS (0) = hS (1), hB (0) = hB (1) (3.10a)
2
To be more precise, the definition just given refers to a knock-out barrier option with American barrier.
There exists several variants (google it!)

51
and, for t ∈ I,

1 ΠuY (t) − ΠdY (t)

hS (t) = , (3.10b)
S(t − 1) eu − ed
e−r eu ΠdY (t) − ed ΠuY (t)
hB (t) = (3.10c)
B(t − 1) eu − ed
is a self-financing, predictable, hedging portfolio process.

Proof. We first show that the given portfolio hedges the derivative. We have

S(t) ΠuY (t) − ΠdY (t) e−r B(t) eu ΠdY (t) − ed ΠuY (t)
V (t) = hS (t)S(t) + hB (t)B(t) = + .
S(t − 1) eu − ed B(t − 1) eu − ed

Note that e−r B(t)/B(t − 1) = 1, while S(t)/S(t − 1) is either eu or ed . By straightforward

calculations we obtain V u (t) = ΠuY (t) and V d (t) = ΠdY (t), that is V (t) = ΠY (t), for all t ∈ I.
Hence the portfolio process is replicating, and therefore also hedging, the derivative. As to
the self-financing property, we have

ΠuY (t)(1 − ed−r ) + ΠdY (t)(eu−r − 1)

hS (t)S(t − 1) + hB (t)B(t − 1) =
eu − ed
−r u
= e (qu ΠY (t) + qd ΠdY (t)) = ΠY (t − 1),

where we used the definition of qu , qd , as well as the recurrence formula (3.8). By the already
proven fact that V (t) = ΠY (t), for all t ∈ I, we have

hS (t)S(t − 1) + hB (t)B(t − 1) = V (t − 1),

which proves the self-financing property. Finally we show that the portfolio is predictable.
Assume first that the European derivative is standard, i.e., Y = g(S(N )). Then ΠY (t) =
fg (t, S(t)), see (3.5), and therefore

ΠuY (t) = fg (t, S(t − 1)eu ), ΠdY (t) = fg (t, S(t − 1)ed ),

i.e., ΠuY (t) and ΠdY (t) are deterministic functions of S(t−1). It follows that hS (t), hB (t) given
by (3.10b)-(3.10c) are also deterministic functions of S(t − 1), and so this portfolio process is
predictable. In the case of non-standard derivatives we have similarly, by Exercise 3.1, that
ΠuY (t) and ΠdY (t) are deterministic functions of S(0), . . . , S(t − 1), which again implies that
the portfolio (3.10b)-(3.10c) is predictable.

Example. Let us compute the hedging portfolio (3.10b)-(3.10c) for the

√ standard derivative
u d
in Section 3.1.1. When S(1) = 2 we have ΠY (2) = 1 and ΠY (2) = 2 − 1, hence (3.10b)
gives √ √ √
1 1 − ( 2 − 1) 2− 2 2
hS (2) = = =1− > 0 (long position).
S(1) 2−1 2 2

52
 √
When S(1) = 1 we have ΠuY (2) = 2 − 1 and ΠdY (2) = 0, hence
√
1 ( 2 − 1) − 0 √
hS (2) = = 2 − 1 > 0 (long position).
S(1) 2−1
Recall that hS (2) is the position in the stock in the interval (1, 2]. In the interval [0, 1] we
have
√ √ √
1 41 (2 2 − 1) − 14 ( 2 − 1) 2
hS (0) = hS (1) = = > 0 (long position).
S(0) 2−1 4
Note that the seller, in this example, always keeps a long position on the stock in the hedging
portfolio. The (short) position on the risk-free asset can be computed likewise using (3.10c).
Exercise 3.7 (•). Consider a 3-period binomial market with the following parameters:
5 1 1
eu = , ed = , er = 1 p = .
4 2 2
64
Assume S0 = 25
. Consider the European derivative expiring at time T = 3 and with pay-off

Y = S(3)H(S(3) − 1),

where H is the Heaviside function: H(x) = 0, if x ≥ 0, H(x) = 1 if x > 0 (this is an example

of cash-settled digital option). Compute the possible paths of the derivative price and
for each of them give the number of shares of the underlying asset in the hedging portfolio
process. Compute the probability that the return of a constant portfolio with a short position
in the derivative be positive.

3.3 Computation of the binomial price of standard Eu-

ropean derivatives with Matlab
As in Section 2.4, we consider a partition 0 = t1 < t2 < · · · < tN +1 = T of the interval [0, T ]
and the binomial stock price
S(i)eu , with probability p
(
S(i + 1) = , i ∈ I = {1, . . . , N },
S(i)ed , with probability 1 − p

where S(i) = S(ti ) and

1 − p√ p √
r r
T
u = αh + σ h, d = αh − σ h, h= . (3.11)
p 1−p N

The value of the risk-free asset at time ti is given by B(ti ) = B0 erti = B0 e(rh)i := B(i). Hence
the pair (S(i), B(i)) defines a 1+1 dimensional binomial market, in the sense of Section 2.2,

53
with parameters u, d given by (3.11) and interest rate rh. Note that

p √ 1 − p√
r r
αh − σ h < rh < αh + σ h
1−p p

holds for h small, hence the market is arbitrage free provided we take our time partition to
be sufficiently fine. The recurrence formula (3.8) for the price of the European option with
pay-off Y at maturity T becomes

ΠY (N + 1) = Y, and ΠY (i) = e−rh [qu ΠuY (i + 1) + qd ΠdY (i + 1)], for t ∈ I,

where ΠY (i) = ΠY (ti ) and

√ p
√
erh − eαh−σ 1−p
h
qu = q
1−p
√ √ p
√ , qd = 1 − qu .
αh+σ h αh−σ h
e p
−e 1−p

The recurrence formula for standard European options is implemented by the Matlab func-
tion BinomialEuropean defined by the following code:

function P=BinomialEuropean(Q,S,r,g)
h=Q(2)-Q(1);
syms x;
f = sym(g);
N=length(Q)-1;
expu=S(1,2)/S(1,1);
expd=S(2,2)/S(1,1);
qu=(exp(r*h)-expd)/(expu-expd);
qd=(expu-exp(r*h))/(expu-expd);
if (qu<0 || qd<0)
display(’Error: the market is not arbitrage free.’);
P=0;
return
end
P=zeros(N+1);
P(:,N+1)=subs(f,x,S(:,N+1));
for j=N:-1:1
for i=1:j
P(i,j)=exp(-r*h)*(qu*P(i,j+1)+qd*P(i+1,j+1));
end
end

The arguments are the partition Q = {t1 , . . . , tN +1 } and the binomial tree S for the price of
the underlying stock, computed with the function BinomialStock defined in Section 2.4, the
interest rate r of the money market and the pay-off function g (e.g., ’max(x-10,0)’ for a

54
call with strike K = 10). The function returns an upper-triangular (N + 1) × (N + 1) matrix
which contains the binomial tree for the fair price of the derivative. The column j contains
the possible prices of the derivative at time tj . A path of the derivative price is obtained by
moving from each column to next one by either stays in the same row (which means that
the price of the underlying stock went up at this step) or going down one row (which means
that the price of the underlying went down at this step). Note that the Matlab function also
checks that (qu , qd ) defines a probability, i.e., that the market is arbitrage free. If not, the
function stops.
For example, let S be the binomial tree (2.20) and run the command

P=BinomialEuropean(Q,S,0.01,’max(x-10,0)’)

which computes the binomial price of a European call with strike K = 10. The result is
0.2461 0.3817 0.5705 0.8141 1.0971 1.3875
0 0.1140 0.1978 0.3333 0.5387 0.8144
0 0 0.0325 0.0658 0.1333 0.2701
P = (3.12)
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
Observe that if the price of the stock goes down in the first three steps, then the price of the
call becomes zero and remains zero for all subsequent times, regardless of the future path of
the stock price. This happens because the binomial model predicts that the call has no chance
to expire in the money when the stock price goes down in the first three steps and hence the
call becomes worthless. This of course is in contradiction with reality, as the market price
of calls is never zero prior to expire3 . This paradox of the binomial model becomes less and
less important the higher is the number of steps used for the computation. On the one hand
if N is too low the contribution of the paths where the call price is zero is significant for the
computation of the fair price of the call and will determine an underestimation of this price.
On the other hand as N → +∞ the binomial price of the call converges to the Black-Scholes
price (see Chapter 6), and within the Black-Scholes theory the fair price of a call is always
positive before maturity. Moreover it can be shown that the binomial algorithm to compute
the fair price of European derivatives is stable in the following sense: the larger is the number
of steps, the smaller is the numerical error due to truncations. In conclusion, one can trust
the results given by the binomial model only if N is sufficiently large. As a way of example,
Figure 3.1 shows the initial binomial price of a call as a function of N = 2, . . . , 100. It is
clear that for a small number of steps the result is unreliable, while for sufficiently many
steps, say N & 20, the result becomes stable.
Figures 3.2 and 3.3 show the initial price of the call as a function of the parameters α and
σ. We see that (i) the price of the call is very weakly dependent on the mean of log-return α
3
In fact in real life we can never exclude with certainty that the call will expire in the money, irrespective
of how deeply out of the money is the call today.

55
 2

1.8

1.6

1.4

1.2
C(0)

0.8

0.6

0.4

0.2

0
0 10 20 30 40 50 60 70 80 90 100
N

Figure 3.1: Initial price C(0) = P (1, 1) of the call computed for increasing values of N
(S0 = 10, K = 10.5, T = 1, α = 0.1, σ = 0.5, p = 1/2). The price stabilizes around 1.8 for
N & 20.

of the stock and (ii) the price of the call increases with the volatility of the stock. Hence the
price of the call is not sensitive to the average movement of the stock price, but rather only
to how uncertain is the stock price, measured by its volatility. In other words, according
to the binomial model, the seller should not charge more for a call option if the stock price
is expected to increase in the future, while rather the premium increases if the price it
expected to be highly volatile. The analytical proof of these results is more easily carried
out in the Black-Scholes model, which is the time-continuum limit of the binomial model
(i.e., N → ∞, h → 0 such that N h = T ); see Chapter 6.
Exercise 3.8 (?). Can you give an intuitive explanation for the found dependence on the
parameter α, σ of the call option binomial price?

Exercise 3.9. Let N = 100 and consider a European call.

1. Plot the initial binomial price of the call as a function of S0

2. Plot the initial binomial price of the call as a function of the strike price
3. Plot the initial binomial price of the call as a function of the interest rate r
4. Verify the validity of the put-call parity
Exercise 3.10. Write a Matlab function which computes the hedging portfolio of standard
European derivatives.

56
 0.24

0.239

0.238

0.237

0.236
C(0)

0.235

0.234

0.233

0.232

0.231

0.23
−0.1 −0.05 0 0.05 0.1 0.15
α

Figure 3.2: Initial price C(0) of the call computed for increasing values of α ∈ [−0.1, 0.1]
(S0 = 10, K = 10.5, T = 1, σ = 0.1, N = 1000, p = 1/2). Clearly, the dependence on α of the
call price is extremely weak.

11

10

6
C(0)

0
0 2 4 6 8 10 12
σ

Figure 3.3: Initial price C(0) of the call computed for increasing values of σ > 0 (S0 =
10, K = 10.5, T = 1, α = 0, N = 1000, p = 1/2). The call price increases with the volatility
and approaches the stock price S0 = 10 for σ large (note however that only values 0 < σ . 2
are realistic).

57
Exercise 3.11. Look for the market price of call options on S&P500 which expire in the
third Friday from now (the expiration date is always a Friday). Select 20 prices, 10 for the
first options in the money and 10 for the first options out of the money. Let S0 be the current
value of S&P500 and σ20 its current 20-days volatility. Compile a table with the following
information: the first column contains the strike price K, the second column the market
price, the third column the binomial price. Use σ = σ20 and r = 0 to compute the binomial
price, but check that the result does not change significantly for, say, 0 ≤ r ≤ 0.05 (which is
a quite large value for the interest rate). Explain why. Plot the difference between the market
price and the theoretical price as a function of K.

The difference between the theoretical fair price and the market price of a call option is
commonly expressed in terms of the implied volatility σimp of the call, which is defined as
the value of the parameter σ to be used in the calculation of the binomial price in order for
the latter to be equal to the market price4 .

Exercise 3.12. Compute numerically the implied volatility of the options analysed in Exer-
cise 3.11. Compare your value of σimp with the one quoted in the market. Plot the implied
volatility as a function of the strike price and discuss your findings.

We shall discuss again the important concept of implied volatility in Chapter 6.

4
Actually, the computation of the implied volatility is typically done using the Black-Scholes model.
However, as already mentioned, the binomial price and the Black-Scholes price are practically the same for
N sufficiently large.

58
Chapter 4

American derivatives

This chapter is concerned with American derivatives on a stock. We assume that the stock
price follows the binomial model
(
S(t − 1)eu with probability p
S(t) = , t ∈ I = {1, . . . , N },
S(t − 1)ed with probability 1 − p
where 0 < p < 1, u > d. It is assumed that S(0) = S0 is known. Moreover we assume that
the interest rate r of the money market is constant, so that the value of the risk-free asset
at time t is
B(t) = B0 ert .
We impose d < r < u. In particular, the binomial market is arbitrage-free and the risk-
neutral probability is well defined:
er − ed eu − er
qu = , qd = , qu , qd ∈ (0, 1), qu + qd = 1.
eu − ed eu − ed
We denote by {(hS (t), hB (t)}t∈I a portfolio process invested in the stock and the risk-free
asset, where (hS (t), hB (t)) is the portfolio position in the interval (t−1, t] and hS (0) = hS (1),
hB (0) = hB (1). The value at time t of the portfolio is V (t) = hS (t)S(t) + hB (t)B(t). Recall
that the portfolio process is said to be self-financing if
V (t − 1) = hS (t)S(t − 1) + hB (t)B(t − 1), t ∈ I,
which means that no cash is ever withdrawn or added to the portfolio. Recall also that
the portfolio process is called predictable if there exist N functions H1 , . . . HN such that
Ht : (0, ∞)t → R2 and
(hS (t), hB (t)) = Ht (S(0), . . . , S(t − 1)).
Another way to say this is that the portfolio position at time t is a deterministic function of
the stock price up to time t − 1. In particular the portfolio position in the interval (t − 1, t]
is determined by the information available up to and included the time t − 1.

59
We shall need the proven recurrence formula for the binomial fair price ΠY (t) of European
derivatives. Namely, denoting ΠuY (t) the value of the European derivative at time t assuming
that the stock price goes up at time t (i.e., S(t) = S(t − 1)eu , or equivalently, xt = u), and
similarly for ΠdY (t), with “up” replaced by “down”, we have seen in Theorem 3.2, that ΠY (t)
satisfies

ΠY (N ) = Y, and ΠY (t) = e−r [qu ΠuY (t + 1) + qd ΠdY (t + 1)], for t ∈ {0, N − 1}. (4.1)

4.1 The binomial price of American derivatives

In contrast to European derivatives, American derivatives can be exercised at any time prior
or including the expiration date. Let Y (t) be the pay-off of an American derivative exercised
at time t. We assume that t ∈ I = {1, . . . , N }, while t = 0 is the present time. Let S(t) be
the binomial stock price of the underlying stock at time t. We restrict ourselves to standard
American derivatives, which means that

Y (t) = g(S(t)), t ∈ {0, 1, . . . , N },

where g : (0, ∞) → [0, ∞) is the pay-off function of the derivative1 . For example, g(z) =
(z − K)+ for American call options and g(z) = (K − z)+ for American put options, where
(z)+ = max(0, z) and K is the strike price of the option. Y (t) is also called intrinsic value
of the derivative. If Y (t) = 0 then the derivative is out of the money at time t. The initial
pay-off Y (0) is also denoted by Y0 ; as it depends only on the initial price S0 = S(0) of the
stock, the value of Y0 is known. Moreover Y (t) is clearly path-dependent, i.e., Y (t) = Y (t, x).
As usual, Y u (t) denotes the intrinsic value at time t assuming that the stock price goes up
at time t and similarly for Y d (t), with “up” replaced by “down”.
Next we want to introduce a reasonable definition for the binomial fair price of the American
derivative with intrinsic value Y (t) and maturity T = N . We denote this price by Π b Y (t),
while ΠY (t) denotes the binomial price of the corresponding European derivative with pay-
off Y (N ) = g(S(N )) at maturity T = N . As already discussed in Chapter 1, any meaningful
definition of fair price for American derivatives must satisfy the following: (a) Π b Y (N ) =
b Y (t) ≥ Y (t) and (c) Π
Y (N ), (b) Π b Y (t) ≥ ΠY (t). Property (a) fixes the price of the American
derivative at time N . Due to (b) and (c), a reasonable definition for the fair price of the
American derivative at time t = N − 1 is
b Y (N − 1) = max(Y (N − 1), ΠY (N − 1)).
Π (4.2)

Using the recurrence formula (3.8), we have

ΠY (N − 1) = e−r (qu ΠuY (N ) + qd ΠdY (N )) = e−r (qu Π

b u (N ) + qd Π
Y
b d (N )),
Y
1
For non-standard American derivatives, the pay-off has the form Y (t) = gt (S(1), . . . , S(t)), where
gt : (0, ∞)t → [0, ∞).

60
where for the second equality we used that Π
b Y (N ) = ΠY (N ). Hence (4.2) becomes

b Y (N − 1) = max[Y (N − 1), e−r (qu Π

Π b dY (N ))].
b uY (N ) + qd Π

This suggests to introduce the following definition:

Definition 4.1. The binomial (fair) price Π b Y (t) of the standard American derivative with
intrinsic value Y (t) = g(S(t)) at time t ∈ {0, 1, . . . , N } is defined by the recurrence formula

Π
b Y (N ) = Y (N ) (4.3)
b Y (t) = max[Y (t), e−r (qu Π
Π b u (t + 1) + qd Π
b d (t + 1))], t ∈ {0, . . . , N − 1}. (4.4)
Y Y

Exercise 4.1 (?). Suppose an investor buys the American derivative at time N −1. Since the
derivative can only be exercised at time t = N , why is the price of the American derivative not
the same as the corresponding European derivative? Namely, why Π b Y (N − 1) = max(Y (N −
1), ΠY (N − 1)) and not Π b Y (N − 1) = ΠY (N − 1)?

Exercise 4.2 (?). Why did we not define the binomial fair price of the American derivative
as Π
b Y (t) = max(Y (t), ΠY (t))?

In Section 4.4 we shall give further support to the statement that Π

b Y (t) does actually define
a fair price for American derivatives. For the moment we limit ourselves to show that
Definition 4.1 captures the properties of American derivatives discussed in Chapter 1.
Theorem 4.1 (∗). Let Π b Y (t), t ∈ I, denote the binomial price of the standard American
derivative with intrinsic value Y (t) = g(S(t)) and maturity T = N . Let ΠY (t) denote the
binomial price of the corresponding European derivative with pay-off Y (N ) = g(S(N )) at
maturity N . The following holds:

(a) Π
b Y (t) is a deterministic function of S(t).

b Y (t) ≥ ΠY (t), for all t ∈ I.

(b) Π
(c) Assuming r ≥ 0, the American call and the European call with the same strike and
maturity have the same binomial price.

Proof. (a) We have to show that Π b Y (t) = Ht (S(t)) for some functions Ht : (0, ∞) → R. We
argue by induction. First this is true at time N , because Π b Y (N ) = Y (N ) = g(S(N )), hence
HN = g. Now assume that the claim is true at time t + 1, i.e., there exists Ht+1 : (0, ∞) → R
such that Π
b Y (t + 1) = Ht+1 (S(t + 1)). Hence

b u (t + 1) = Ht+1 (S(t)eu ),
Π b d (t + 1) = Ht+1 (S(t)ed ).
Π
Y Y

Thus, using Y (t) = g(S(t)), we have

b Y (t) = max(g(S(t)), e−r (qu Ht+1 (S(t)eu ) + qd Ht+1 (S(t)ed )) = Ht (S(t)),
Π

61
where Ht (z) = max(g(z), e−r (qu Ht+1 (zeu ) + qd Ht+1 (zed )).
(b) Also in this case we can argue by induction. The claim is true at maturity N , since
at this time the value of both derivatives equals the pay-off by definition. Now assume the
claim is true at time t + 1, i.e.,
b Y (t + 1) ≥ ΠY (t + 1).
Π
Hence
e−r (qu Π Y
b d (t + 1)) ≥ e−r (qu Πu (t + 1) + qd Πd (t + 1)),
b u (t + 1) + qd Π
Y Y Y
which implies
b Y (t) = max(Y (t), e−r (qu Π
Π b dY (t + 1)) ≥ e−r (qu Π
b uY (t + 1) + qd Π b uY (t + 1) + qd Π
b dY (t + 1))
≥ e−r (qu ΠuY (t + 1) + qd ΠdY (t + 1)) = ΠY (t),
where we used that the price of European calls satisfies the recurrence relation (4.1)
(c) To prove that American and European calls with equal parameters have the same value,
we first observe that it suffices to prove that Π b call (t) ≤ Πcall (t), since we already proved in (b)
that Πb call (t) ≥ Πcall (t) holds (we prove this for all standard American derivatives!). We argue
again by induction. At maturity Π b call (N ) = Πcall (N ). Assume that Π b call (t+1) ≤ Πcall (t+1).
Then
e−r (qu Π b dcall (t + 1)) ≤ e−r (qu Πucall (t + 1) + qd Πdcall (t + 1)).
b ucall (t + 1) + qd Π (4.5)
Now let Πput (t) be the price of the European put with the same parameters as the European
call. We have shown in Theorem 3.1 that the put-call parity holds:
Πcall (t) = S(t) − Ke−r(T −t) + Πput (t).
As Πput (t) ≥ 0 and r ≥ 0 we have Πcall (t) ≥ S(t) − K. Since Πcall (t) ≥ 0, we have
Πcall (t) ≥ max(S(t) − K, 0) = (S(t) − K)+ = Y (t),
where Y (t) is the intrinsic value of the American call. As
Πcall (t) = e−r (qu Πucall (t + 1) + qd Πdcall (t + 1)),
we have
Y (t) ≤ e−r (qu Πucall (t + 1) + qd Πdcall (t + 1)).
Hence, using that max(a, b) ≤ max(c, b) holds when a ≤ c,
b call (t) = max(Y (t), e−r (qu Π
Π b ucall (t + 1) + qd Π
b d (t + 1))
call

≤ max(e−r (qu Πucall (t + 1) + qd Πdcall (t + 1)), e−r (qu Π

b u (t + 1) + qd Π
call
b d (t + 1))).
call

By (B.5),
max(e−r (qu Πucall (t + 1) + qd Πdcall (t + 1)), e−r (qu Π
b u (t + 1) + qd Π
call
b d (t + 1)))
call
= e−r (qu Πucall (t + 1) + qd Πdcall (t + 1)) = Πcall (t).
b call (t) ≤ Πcall (t), and the proof is complete.
Hence Π

62
4.2 Optimal exercise time of American put options
The last claim in Theorem 4.1 is consistent with what we have seen in Chapter 1, namely
that it is never optimal to exercise an American call option prior to expiration when the
underlying stock pays no dividends. The example treated in Section 4.3 below shows that
this is not true for put options. To this regard we remark that, when the American put is
priced using the binomial model, an optimal exercise time in the sense of Definition 1.1 is a
time t ∈ I such that

(K − S(t))+ = max[(K − S(t))+ , e−r (qu Π

b u (t + 1) + qd Π
Y
b d (t + 1))],
Y

that is to say, a time t ∈ I at which

(K − S(t))+ ≥ e−r (qu Π

b u (t + 1) + qd Π
Y
b d (t + 1)).
Y (4.6)

It can be shown that the inequality (4.6) holds at time t ∈ {1, . . . , N − 1} if and only if
S(t) ≤ S∗ (t), where S∗ (t) < K depends only on the market parameters. The exact expression
for S∗ (t) is unknown, but it can be easily determined numerically, see Exercise 4.9. This
fact is very important in the applications, for it tells us that the American put should be
exercised as soon as the price of the stock falls below the value S∗ (t), irrespective of whether
the market price of the put option equals its intrinsic value (which is extremely unlike to
happen). In the next theorem the exact value of S∗ (t) is derived in the 2-period model.
Theorem 4.2 (∗). In a 2-period binomial model with parameters u > 0, d < 0, 0 < r < u,
the earlier exercise at time t = 1 of the American put with strike K and maturity T = 2 is
optimal if and only if S(1) ≤ S∗ (1), where
1 − e−r qd
S∗ (1) = K . (4.7)
1 − e−r qd ed

Proof. We need to study for which values of S(1) is (4.6) satisfied at time t = 1. Since
b u (2) = (K − S(1)eu )+ and Π
Π b d (2) = (K − S(1)ed )+ , (4.6) becomes
Y Y

(K − S(1))+ ≥ e−r (qu (K − S(1)eu )+ + qd (K − S(1)ed )+ ). (4.8)

We study the inequality (4.8) for separate values of S(1) in the intervals

S(1) ∈ [0, Ke−u ] := I1 , S(1) ∈ [Ke−u , K] := I2 ,

S(1) ∈ [K, Ke−d ] := I3 , S(1) ∈ [Ke−d , +∞) := I4 .

For S(1) ∈ I1 we have,

e−r (qu (K − S(1)eu )+ + qd (K − S(1)ed )+ ) = e−r (qu eu (Ke−u − S(1)) + qd ed (Ke−d − S(1))).

Using qu + qd = 1 and qu eu + qd ed = er we obtain

e−r (qu (K − S(1)eu )+ + qd (K − S(1)ed )+ ) = K − S(1) = (K − S(1))+ , for S(1) ∈ I1 .

63
Similarly, for S(1) ∈ I2 we have

−r u d (K − S(1))+ for S(1) ≤ S∗ (1)
e (qu (K − S(1)e )+ + qd (K − S(1)e )+ ) =
e−r qd ed (Ke−d − S(1)) for S(1) > S∗ (1)

where S∗ (1) is given by (4.7). Treating similarly the cases S(1) ∈ I3 and S(1) ∈ I4 we find
that the price of the American put at time t = 1 is given by

 (K − S(1))+ for 0 < S(1) ≤ S∗ (1)
Π(1) =
b e qd e (Ke − S(1)) for S∗ < S(1) ≤ Ke−d
−r d −d

0 for S > Ke−d



We conclude that it is optimal to exercise the American put at time t = 1 if and only if
S(1) ≤ S∗ (1).

4.3 Example of American put option

In this section we consider an example of American put option. We let the strike price
K = 3/4, and so  
3
Y (t) = − S(t) .
4 +

We present an example in N = 2 periods. Consider a market with the following parameters:

7 1 9
eu = , ed = , er = ,
4 2 8
so that qu = qd = 1/2. Assuming S0 = 1, the binomial tree for the stock price is
49
S(2) =
7 16
u

7
S(1) =
9 4
u d
'
7
S0 = 1 S(2) =
9 8

d u
$
1
S(1) =
2
d
#
1
S(2) =
4

64
When the price of the stock in the paths above is within a box, the put option is in the
money. In fact, the binomial tree for the intrinsic value Y (t) of the American put is

Y (2) = 0
8
u

Y (1) = 0
9
u d
&
Y0 = 0 Y8 (2) = 0
d u
%
1
Y (1) = 4
d
&
1
Y (2) = 2

Let us first compute the price of the corresponding European put with pay off
 
3
− S(2) = Y (2) = ΠY (2)
4 +

at time of maturity T = 2. Using the recurrence formula (4.1) we have

    
4 u d 4 3 7 3 1
ΠY (1) = [ΠY (2) + ΠY (2)] = − S(1) + − S(1) .
9 9 4 4 + 4 2 +

Hence    
43 49 3 7
ΠuY (1) = − + − = 0,
94 16 + 4 8 +
    
d 4 3 7 3 1 2
ΠY (1) = − + − = .
9 4 8 + 4 4 + 9
Therefore, again by (4.1), we have
4 8
ΠY (0) = [ΠuY (1) + ΠdY (1)] = .
9 81

65
In conclusion, we have obtained the following paths for the binomial price of the European
derivative
Π7Y (2) = 0
u

ΠY (1) = 0
7
u d
'
8
ΠY (0) = ΠY (2) = 0
81 7
d u
'
2
ΠY (1) = 9
d
'
1
ΠY (2) = 2

Exercise 4.3 (?). Let {(hS (t), hB (t)), t = 0, 1, 2} be a hedging, self-financing portfolio process
for this European put. Can you guess whether hS (0) will be positive or negative without
computing the portfolio?

Now we compute the prices of the American put option. Of course, at time of maturity it is
the same as its European counterpart (by definition of hedging portfolio). At time t = 1 we
have, by (4.4),
 
4 bu d
ΠY (1) = max Y (1), (ΠY (2) + ΠY (2))
b b
9
     
4 3 7 3 1
= max Y (1), − S(1) + − S(1) .
9 4 4 + 4 2 +

Since    
u 3 7 d 3 1 1
Y (1) = − = 0, Y (1) = − = ,
4 4 + 4 2 + 4
we find  
b uY (1) = max[0, 0] = 0, d 1 2 1
Π ΠY (1) = max ,
b =
4 9 4
and so  
4 bu d 1
ΠY (0) = max Y (0), (ΠY (1) + ΠY (1)) = .
b b
9 9
Hence the binomial price of the American put corresponding to the different paths of the

66
stock price is as follows:

Π
b Y (2) = 0
7
u

Π
b Y (1) = 0
8
u d

%
b Y (0) = 1
Π Π
b Y (2) = 0
9 :
d u
$
b Y (1) = 1
Π
4
d

%
1
Π
b Y (2) =
2

Note that the binomial price of the American put and of the European put are different in
two instances, which are indicated in the paths above by putting the price of the American
put within a box. In particular, their initial price is different. When the prices are different,
the American put is more expensive than the European put. Moreover, if (and only if)
the stock price goes down at time t = 1, the binomial price of the American put equals its
intrinsic value prior to maturity, hence in this case (and only in this case) the earlier exercise
at time t = 1 is optimal. This result agrees with Theorem 4.2, since the value (4.7) for this
option is given by
3 1 − 98 12
 
15 1 7
S∗ (1) = = ∈ , .
4 1 − 89 12 12 28 2 4

4.4 Hedging portfolio processes of American deriva-

tives
In this section we describe how to obtain hedging portfolio processes for American derivatives.
Recall that for the European derivative with pay-off Y at the expiration date T = N ,
hedging portfolio processes have been defined by imposing that their value V (t) satisfies
V (N ) = Y , see Definition 3.2. Because of the earlier exercise option of American derivatives,
this definition cannot be adopted in the American case. In fact, since the goal of an hedging
portfolio is to secure the writer position (i.e., the short position) and the buyer of the
American derivative has the right to exercise the derivative at any time t ∈ I, then we need
to ensure that the value of the writer portfolio is at any time—and not only at maturity—
sufficient to pay-off the buyer, i.e., V (t) ≥ Y (t). This leads us to define hedging portfolio

67
processes for American derivatives as follows.
Definition 4.2. A portfolio process {hS (t), hB (t)}t∈I is said to be hedging the American
derivative with intrinsic value Y (t) and maturity T = N if

V (N ) = Y (N ), V (t) ≥ Y (t), t = 0, . . . , N − 1,

where V (t) = hS (t)S(t) + hB (t)B(t) is the value of the portfolio process at time t.

Now, the most important hedging portfolios are those which replicate the American deriva-
tive, i.e., the portfolio processes whose value V (t) satisfies V (t) = Π b Y (t), for all t ∈ [0, T ].
Note that replicating portfolio processes are hedging portfolios, because Π b Y (t) ≥ Y (t). In
the European case any self-financing hedging portfolio process is (trivially) replicating, be-
cause ΠY (t) has been defined as the common value of any such portfolio. However, in the
American case, the value at time t of an hedging portfolio process could be strictly greater
than the binomial price Π b Y (t) of the American derivative given in Definition 4.1. If this hap-
pens, then the writer should withdraw cash from the portfolio in order to replicate the value
of the American derivative. This leads us to introduce the important concept of portfolio
processes generating a cash flow. We argue as we did for self-financing portfolios. Recall
that (hS (t), hB (t)) is the investor position on the stock and the risk-free asset during the
time interval (t − 1, t]. Let V (t) = hS (t)S(t) + hB (t)B(t) be the value of this portfolio. At
the time t, the investor sells/buys shares of the two assets. Let (hS (t + 1), hB (t + 1)) be
the new position on the stock and the risk-free asset in the interval (t, t + 1]. Then the
value of the portfolio process immediately after changing the position at time t is given by
V 0 (t) = hS (t + 1)S(t) + hB (t + 1)B(t). The cash flow C(t) is defined as V 0 (t) − V (t) = −C(t)
and corresponds to cash withdrawn (if C(t) > 0) or added (if C(t) < 0) to the portfolio as
a result of the change in the position on the assets (for a self-financing portfolio we have of
course C(t) = 0, for all t ∈ {0, . . . , N }). This leads to the following definition.
Definition 4.3. A portfolio process {(hS (t), hB (t))}t∈I is said to generate the cash flow
C(t − 1), t ∈ I, if

hS (t)S(t − 1) + hB (t)B(t − 1) = hS (t − 1)S(t − 1) + hB (t − 1)B(t − 1) − C(t − 1), t ∈ I,

or, equivalently,

V (t) − V (t − 1) = hS (t)(S(t) − S(t − 1)) + hB (t)(B(t) − B(t − 1)) − C(t − 1).

In particular, if C(t − 1) > 0, then the cash is withdrawn from the portfolio, causing a
decrease of its value, while if C(t − 1) < 0, then the cash is added to the portfolio, causing
an increasing of its value.
Remark 4.1. As we assume hS (0) = hS (1) and hB (0) = hB (1), then C(0) = 0. Therefore
the first time at which the investor can add/remove cash from the portfolio is after chang-
ing the position (instantaneously) at time t = 1, i.e., when passing from (hS (1), hB (1)) to
(hS (2), hB (2)), generating the cash flow C(1).

68
Example: Consider a constant portfolio process that consists of only one share of the stock,
that is hS (t) ≡ 1 and hB (t) ≡ 0. The value at time t of this portfolio is V (t) = hS (t)S(t) =
S(t). Suppose that in the interval of time (t − 1, t) the stock pays a dividend of 1%. This
means that the fraction S(t − 1)/100 is deposited into the account of the investor, while the
price of the stock decreases of the same amount. Hence the value of the portfolio at time t
is
V (t) = S(t) − S(t − 1)/100.
Therefore

V (t) − V (t − 1) = S(t) − S(t − 1)/100 − S(t − 1) = hS (t)(S(t) − S(t − 1)) − C(t − 1),

where hS (t) = 1 and C(t) = S(t − 1)/100.

We now show how to build a predictable replicating portfolio for American derivatives.
Theorem 4.3. Consider the standard American derivative with intrinsic value Y (t) and
maturity T = N . Let Π b Y (t) be its binomial price as given by (4.3)-(4.4). Define the portfolio
process {(b
hS (t), b
hB (t))}t∈I and the cash flow process C(t) recursively as follows:

C(0) = 0, b Y (t − 1) − e−r [qu Π

C(t − 1) = Π b u (t) + qd Π
b d (t)], t ∈ {2, . . . , N }, (4.9)
Y Y

hS (1) = b
b hS (0), hB (0) = b
b hB (1) (4.10a)
and, for t = 1, . . . , N ,
1 Π b d (t)
b u (t) − Π
Y Y
hS (t) =
b , (4.10b)
S(t − 1) eu − ed
e−r eu Π
b d (t) − ed Π
Y
b u (t)
Y
hB (t) =
b . (4.10c)
B(t − 1) eu − ed
The portfolio process (4.10) is predictable, replicates the American derivative and generates
the cash-flow (4.9).

Proof. We have shown in Theorem 4.1 that Π b Y (t) is a deterministic function of S(t), i.e.,
b Y (t) = ft (S(t)), for some function ft : (0, ∞) → (0, ∞). Hence
Π
b u (t) = ft (S(t − 1)eu ),
Π b d (t) = ft (S(t − 1)ed )
Π
Y Y

are deterministic functions of S(t−1), by which it follows immediately that the portfolio pro-
cess (4.10) is predictable. To show that the portfolio replicates the derivative and generates
the cash flow (4.9) we have to show that

V (t) = Π
b Y (t), for all t ∈ {0, . . . N } (4.11)

and
V (t − 1) = b
hS (t)S(t − 1) + b
hB (t)B(t − 1) + C(t − 1), for all t ∈ I. (4.12)

69
The proof is straightforward: just replace (4.9) and (4.10) into (4.11)-(4.12). For instance,
assuming that the price goes up at time t, we compute

V u (t) = b
hS (t)S(t − 1)eu + b hB (t)B(t − 1)er
! !
b u (t) − Π
Π b d (t) e u bd
Π (t) − ed bu
Π (t)
Y Y Y Y
= eu +
eu − ed eu − ed
b u (t),
=Π Y

and at the same fashion one proves that V d (t) = Π b d (t). Hence (4.11) holds. In a similar
Y
fashion, replacing (4.9) and (4.10) into the right hand side of (4.12) we find that the latter
b Y (t − 1), which we already proved to be equal to V (t − 1). Hence (4.12) holds
is equal to Π
as well.

The previous theorem is telling us that the writer can hedge the derivative and still be able
to withdraw cash from the portfolio. While this might seem unfair from the buyer point of
view, we remark that whether the writer is allowed or not to withdraw a positive amount
of cash from the portfolio (i.e., C(t) > 0) depends on the “smartness” of the buyer. In fact,
using (4.4) in (4.9), we have, for t ∈ {2, . . . , N },

C(t − 1) = max(Y (t − 1), e−r (qu Π b dY (t)) − e−r (qu Π

b uY (t)) + qd Π b uY (t)) + qd Π
b dY (t)).

This quantity is positive at time t − 1 if and only if only Y (t − 1) > e−r (qu Π
b u (t)) + qd Π
Y
b d (t)),
Y
which implies that t − 1 is an optimal exercise time. Hence the writer of the American
put can withdraw cash from the portfolio only if the buyer fails to exercise the derivative
optimally. If however the buyer exercises the derivative optimally, then the seller needs the
full value of the portfolio to pay-off the buyer and thus no cash can be withdrawn.

Example. Computing the cash flow at t = 1 for the American put considered in Section 4.3,
we find     
b Y (1) − 4 3 7 3 1
C(1) = Π − S(1) + − S(1) .
9 4 4 + 4 2 +

If the stock price goes up at time t = 1 we obtain C u (1) = 0; if the stock price goes down at
time t = 1 we obtain
    
d 1 4 3 7 3 1 1
C (1) = − − + − = .
4 9 4 8 + 4 4 + 36

Hence if at time t = 1 the price of the stock goes down and the buyer does not exercise
1
the American put, then the writer can withdraw the cash 36 from the portfolio. The value
1 1 2
remaining in the portfolio is 4 − 36 = 9 . The American put can be hedged with this value
by short selling 45 shares of the stock and lending ( 29 + 45 · 12 ) = 45
28
in the money market. So
28 9 7
doing the writer will receive the amount 45 · 8 = 10 at time t = 2; if the stock price goes up

70
at time t = 2 the American put expires out of the money and the writer will use the amount
7
10
to buy 45 shares of the stock at the price S(2) = 78 and close the short position on the
stock without losses. If the price of the stock goes down at time t = 2, the writer will give
the pay-off 12 to the buyer of the American put and use the remaining value 10 7
− 12 = 51 in
the portfolio to buy 45 shares of the stock at the price S(2) = 41 and again close the short
position on the stock without losses.
Exercise 4.4 (•). Consider the American derivative with intrinsic value
Y (t) = min(S(t), (24 − S(t))+ )
and expiring at time T = 3. The initial price of the underlying stock is S(0) = 27, while at
future times it follows the binomial model

4S(t)/3 with probability 1/2
S(t + 1) =
2S(t)/3 with probability 1/2
for t = 0, 1, 2. Assume also that the interest rate of the money market is zero. Compute
the possible paths of the binomial price of the derivative. In which case it is optimal for the
buyer to exercise the derivative prior to expiration? What is the amount of cash that the
seller can withdraw from the portfolio if the buyer does not exercise the derivative optimally?
Exercise 4.5 (•). Consider a 3-period binomial model with the following parameters:
5 1
eu = , ed = , er = 1, p ∈ (0, 1).
4 2
64
Let S(0) = 25 be the initial price of the stock. Consider an American style derivative on the
stock with maturity T = 3 and intrinsic value


Y (t) = 3 − S(t) H S(t) − 7/5 ,
where H(x) is the Heaviside function and |x| is the absolute value of x (recall that H(x) = 1
if x > 0, H(x) = 0 if x ≤ 0). Compute the binomial price of the derivative at each
time t ∈ {0, 1, 2, 3} (max. 1 point) and the initial position on the stock in the hedging
portfolio. Compute the cash that the seller can withdraw from the portfolio if the buyer does
not exercise the derivative at optimal times. Compute the probability that the derivative is
in the money at time t and the probability that the return for the buyer is positive at time t,
where t ∈ {0, 1, 2, 3}.
Exercise 4.6 (•). Consider a 2-period binomial model with the same parameters as in Ex-
ercise 4.5. Compute the price at time t ∈ {0, 1, 2} of the American put on the stock with
maturity T = 2 and strike price K2 = 11 5
and identify the possible optimal exercise times
prior to maturity. Next consider the compound option which gives to its owner the right
8
to buy the American put at time t = 1 for the price K1 = 25 . Compute the price of the
compound option at time t = 0 and the hedging portfolio for the compound option (assume
B(0) = 1). Compute the maximum expected return in the interval t ∈ [0, 2] for the owner of
the compound option as a function of p ∈ (0, 1).
Exercise 4.7 (?). Compute the replicating portfolio for the American put given in Section 4.3
(assume B0 = 1).

71
4.5 Computation of the fair price of American deriva-
tives with Matlab
We work under the same set-up as in Section 3.3. Namely we consider a partition 0 = t1 <
t2 < · · · < tN +1 = T of the interval [0, T ] and the binomial stock price

S(i)eu , with probability p

(
S(i + 1) = , i ∈ I = {1, . . . , N },
S(i)ed , with probability 1 − p

where S(i) = S(ti ) and

1 − p√ p √
r r
u = αh + σ h, d = αh − σ h. (4.13)
p 1−p

The value of the risk-free asset at time ti is given by B(ti ) = B0 erti = B0 e(rh)i := B(i). Hence
the pair (S(i), B(i)) defines a 1+1 dimensional binomial market with parameters u, d given
by (4.13) and interest rate rh. The definition (4.1) of binomial price of American derivative
becomes

Π b Y (i) = max(Y (i), e−rh (qu Π

b Y (N + 1) = Y (N + 1) Π b u (i + 1)) + qd Π
b d (i + 1)), t ∈ I,
Y Y

where Π
b Y (i) = Π
b Y (ti ) and
√ p
√
erh − eαh−σ 1−p
h
qu = q
1−p
√ √ p
√ , qd = 1 − qu .
αh+σ h αh−σ h
e p
−e 1−p

Moreover Y (i) = g(S(i)), i = 1, . . . , N + 1, is the intrinsic value of the American derivative.

The following code defines a function BinomialAmerican which computes the binomial price
and the cash flow of standard American derivatives.

function [P,C]=BinomialAmerican(Q,S,r,g)
h=Q(2)-Q(1);
syms x;
f = sym(g);
N=length(Q)-1;
expu=S(1,2)/S(1,1);
expd=S(2,2)/S(1,1);
qu=(exp(r*h)-expd)/(expu-expd);
qd=(expu-exp(r*h))/(expu-expd);
if (qu<0 || qd<0)
display(’Error: the market is not arbitrage free.’);
P=0;
return

72
end
P=zeros(N+1);
P(:,N+1)=double(subs(f,x,S(:,N+1)));
C(:,N+1)=0;
Y=double(subs(f,x,S));
for j=N:-1:1
for i=1:j
P(i,j)=max(Y(i,j),exp(-r*h)*(qu*P(i,j+1)+qd*P(i+1,j+1)));
C(i,j)=P(i,j)-exp(-r*h)*(qu*P(i,j+1)+qd*P(i+1,j+1));
end
end

The binomial price of the derivative is stored in the upper-triangular (N + 1) × (N + 1)

matrix P , while the cash flow is stored in the upper-triangular (N + 1) × (N + 1) matrix C.
We use the convention that at time of maturity the cash flow is equal to zero.
For example, using as inputs the binomial stock price (2.20), the interest rate r = 0.01 and
the pay-off function of the put with strike 10, i.e., g(x) = (10 − x)+, we obtain the output
0.2385 0.1120 0.0320 0 0 0
0 0.3619 0.1899 0.0633 0 0
0 0 0.5297 0.3133 0.1250 0
P =
0 0 0 0.7407 0.4970 0.2468
0 0 0 0 0.9752 0.7376
0 0 0 0 0 1.2038
0 0 0 0 0 0
0 0 0 0 0 0
C= 0 0 0 0 0 0
0 0 0 0.0017 0.0017 0
0 0 0 0 0.0017 0
Remark 4.2. Using the pay-off of the call, g(x) = (x − 10)+ , we obtain that the price of
the American call is exactly the same as the corresponding European call, see (3.12), while
C ≡ 0. This of course is consistent with the proven fact that, in the absence of dividends, it
is never optimal to exercise an American call prior to expire.
Exercise 4.8. Repeat Exercise 3.9 for American put options.
Exercise 4.9. Plot the pairs (ti , S(ti )) in the plane, for each optimal exercise time ti of
the American put. Experiment for different values of K, T . Use a large number of steps
(N ≈ 100). What happens when T → +∞?

In the time-continuum limit, N → +∞, h → 0 such that N h = T , the points (ti , S(ti ))
for each optimal exercise time form the so-called the optimal exercise curve (or optimal
exercise boundary) of the American put option.

73
Exercise 4.10. Show numerically that the initial binomial price of the American put with
strike K and maturity T converges, as T → +∞, to the value v(S0 ), where v is the function
(
K −x 0≤x≤L
v(x) =
2r
x − σ2
(K − L) L x>L

and
2r
L= K.
2r + σ 2
How is this result related to your findings in Exercise 4.9?

74
Chapter 5

Introduction to Probability Theory

The main goal of this chapter is to reformulate the binomial options pricing model in the
language of probability theory. To this purpose we shall first review some basic concept in
probability; a more systematic presentation of this theory can be found e.g. in [5]. For an
application of probability theory to portfolio optimization, see Appendix A.

5.1 Finite Probability Spaces

Let Ω be a set containing a finite number of elements ω1 , ω2 , . . . , ωM . We denote Ω as
Ω = {ω1 , . . . , ωM }, or Ω = {ωi }i=1,...,M (5.1)
and call it a sample space. The elements ωi ∈ Ω, i = 1, . . . , M , are called sample points.
The sample points identify the possible outcomes of an experiment. For instance, if the
experiment consists in “throwing a die”, then
Ω = {1, 2, 3, 4, 5, 6} (M = 6),
while for the experiment “tossing a coin once”, we have
Ω = Ω1 := {H, T } (M = 2),
where H stands for “Head” and T for “Tail”; the subscript 1 in Ω1 indicates that the coin
is tossed only once. In the experiment “tossing a coin twice” we have
Ω = Ω2 := {(H, H), (H, T ), (T, H), (T, T )} = Ω1 × Ω1 (M = 22 = 4)
and in the experiment “tossing a coin N times” we have
Ω = ΩN := {ω = (γ1 , γ2 , . . . , γN ); γj = H or T, j = 1, . . . , N }
= Ω1 × Ω1 × · · · × Ω1 = {H, T }N (M = 2N ).
| {z }
N times

75
We denote by 2Ω the power set of Ω, i.e., the set of all subsets of Ω. It consists of the
empty set ∅, the subsets containing one element, i.e., {ω1 }, {ω2 }, . . . , {ωM }, which are called
atomic sets, the subsets containing two elements, i.e.,
{ω1 , ω2 }, . . . , {ω1 , ωM }, {ω2 , ω3 }, . . . , {ω2 , ωM }, . . . , {ωM −1 , ωM },
the subsets containing 3 elements and so on, and the set Ω = {ω1 , . . . , ωM } itself. Thus 2Ω
contains 2M elements. For instance
2Ω1 = {∅, {H}, {T }, {H, T } = Ω1 }.
Exercise 5.1. Write down 2Ω2 .

The elements of 2Ω (i.e., the subsets of Ω) are called events. They identify possible events
that occur in the experiment. For example
{2, 4, 6} ≡ [the result of throwing a die is an even number],
{(H, H), (T, T )} ≡ [tossing a coin twice gives the same outcome in both tosses].
Exercise 5.2. Write down the following events A, B, C ∈ 2Ω4 :
A = [number heads=number tails], B = [successive tosses are different],
C = [there exist at least three identical tosses].

If A, B ∈ 2Ω are events, then A ∪ B is the event that A or B happens, while A ∩ B is the

event that both A and B happen. If the sets A, B ⊂ Ω are disjoint, i.e., A ∩ B = ∅, the
events A and B cannot occur simultaneously. For instance, {1, 3, 5} ∩ {2, 4, 6} = ∅ means
that the outcome of a die roll cannot be both even and odd.
The atomic set {ωi } identifies the event that the outcome of the experiment is exactly ωi .
We want to assign a probability P to such special events. To this purpose we introduce M
real numbers p1 , p2 , . . . pM such that
M
X
0 < pi < 1, for all i = 1, . . . , M , and pi = 1. (5.2)
i=1

We define pi to be the probability of the event {ωi }, that is

P({ωi }) = pi , i = 1, . . . , M.
Any generic event A ∈ 2Ω can be written as the disjoint union of atomic events, e.g.,
{ω1 , ω3 , ω6 } = {ω1 } ∪ {ω3 } ∪ {ω6 }.
This leads to define the probability of a generic event A ∈ 2Ω as
X X
P(A) = P({ωi }) = pi . (5.3)
i:ωi ∈A i:ωi ∈A

76
For instance P({ω1 , ω3 , ω6 }) = p1 + p3 + p6 . We shall also write the definition of P(A) as
X
P(A) = P({ω}).
ω∈A

In particular
X M
X
P(Ω) = P({ω}) = pi = 1.
ω∈Ω i=1
We also set
P(∅) = 0, (5.4)
which means that it is impossible that the experiment gives no outcome. Clearly ∅ is the
only event with zero probability: any other such event is excluded a priori by the sample
space. At this point every event has been assigned a probability.
Definition 5.1. Given p = (p1 , . . . , pM ) satisfying (5.2) and a set Ω = {ω1 , . . . , ωM }, the
function P : 2Ω → [0, 1] defined by (5.3)-(5.4) is called a probability measure. The pair
(Ω, P), is called a finite probability space.
Remark 5.1. If we want to emphasize the dependence of the probability measure on the
parameters p1 , . . . , pm , we shall denote it by Pp .

Examples

• In the experiment “throwing a die”, let p1 = p2 = p3 = p4 = p5 = p6 = 1/6. We say

that the die is fair, as any number between 1 and 6 has the same probability to appear
as a result of the roll. Then P({2, 4, 6}) = 1/6 + 1/6 + 1/6 = 1/2, i.e., we have 50%
chances to get an even number.
• In Ω1 we let P({H}) = pH , P({T }) = pT , where pT = 1 − pH , pH ∈ (0, 1). The coin is
said to be fair if pH = pT = 1/2.
• In Ω2 we assign a probability to the atomic events by setting
P({(H, H)}) = pH · pH = p2H ,
P({(T, H)}) = P({(H, T )}) = pT pH , P({(T, T )}) = p2T .
Notice that P({(H, H)}) + P({(T, H)}) + P({(H, T )}) + P({(T, T )}) = (pH + pT )2 = 1,
as it should be. This way of assigning probabilities is equivalent to the assumption
that the two tosses are independent, i.e., the result of the first toss does not influence
the result of the second toss (see Definition 5.4 below). As an example of computing
the probability of non-atomic events, consider the event
A = [the outcome of the two tosses is the same].
Then
P(A) = P({(H, H), (T, T )}) = P({(H, H)}) + P({(T, T )}) = p2H + p2T .

77
 • In ΩN = {H, T }N , we assign a probability to each of the 2N atomic events by letting
P({ω}) = (pH )NH (ω) (pT )NT (ω) , for all ω ∈ ΩN ,
where NH (ω) and NT (ω) = N − NH (ω) are respectively the number of heads and tails
in the N -toss corresponding to ω. Again, this definition of probability makes the N -
tosses independent. Since for all k = 0, . . . , N the number of N -tosses ω ∈ ΩN having
NH (ω) = k is given by the binomial coefficient
 
N N!
= ,
k k!(N − k)!
then
X X X  pH NH (ω)
NH (ω) NT (ω) N
P({ω}) = (pH ) (pT ) = (pT )
ω∈ΩN ω∈ΩN ω∈Ω
pT
N
N   k
X N pH
= (pT )N .
k=0
k pT
PN

k N
By the binomial theorem, (1 + a)N = a , hence
k=0 k
 N
X
N pH
P(Ω) = P({ω}) = (pT ) 1+ = (pT + pH )N = 1.
ω∈Ω
pT
N

The probability of any other event A ∈ 2ΩN is the sum of the probabilities of the atomic
events whose (disjoint) union forms the set A.

The last example deserves to be given a separate definition.

Definition 5.2. Given 0 < p < 1, the pair (ΩN , Pp ) given by ΩN = {H, T }N and
X
Pp (A) = pNH (ω) (1 − p)NT (ω) , for all A ∈ 2ΩN , (5.5)
ω∈A

is called a N -coin toss probability space. Here NH (ω) is the number of H in the sample
ω and NT (ω) = N − NH (ω) is the number of T .

Example. Given k ∈ {0, . . . , N }, let Ak = {ω : NH (ω) = k} ⊂ ΩN be the event that the

number of heads in the N -toss is k. Then
X X X
P(Ak ) = pNH (ω) (1−p)NT (ω) = pk (1−p)N −k = pk (1−p)N −k 1 = pk (1−p)N −k Mk ,
ω∈Ak ω∈Ak ω∈Ak

N
P

where Mk = ω∈Ak 1 is the number of N -tosses with k heads. As Mk = k
we get
 
N k
P(Ak ) = p (1 − p)N −k .
k

78
Exercise 5.3. Compute P(A), where A = {ω : NH (ω) = NT (ω)} ⊂ ΩN .

It is possible that the occurrence of an event A affects the probability that a second event
B occurred. For instance, for a fair coin we have P({H, H}) = 1/4, but if we know that
the first toss is a tail, then P({H, H}) = 0. This simple remark leads to the definition of
conditional probability.

Definition 5.3. Given two events A, B such that P(B) > 0, the conditional probability
of A given B is defined as
P(A ∩ B)
P(A|B) = .
P(B)
Similarly, if B1 , B2 , . . . , Bn are events such that P(B1 ∩ · · · ∩ Bn ) > 0, the conditional prob-
ability of A given1 B1 , . . . , Bn is

P(A ∩ B1 ∩ · · · ∩ Bn )
P(A|B1 , . . . , Bn ) = .
P(B1 ∩ · · · ∩ Bn )

If the occurrence of B does not affect the occurrence of A, i.e., if P(A|B) = P(A), we say
that the two events are independent. By the previous definition, the independence property
is equivalent to the following.

Definition 5.4. Two events A, B are said to be independent if

P(A ∩ B) = P(A)P(B).

Similarly, n events A1 , . . . , An are said to be independent if

P(Ak1 ∩ · · · ∩ Akm ) = P(Ak1 ) · . . . · P(Akm ),

for all 1 ≤ k1 < k2 < · · · < km ≤ n.

Exercise 5.4. Write down all the identities that must be verified in order for three events
A, B, C to be independent.

Exercise 5.5. Compute the probability of the events defined in Exercise 5.2 and verify that
there is no pair of independent events among them. Compute

P(A|B), P(B|A), P(A|C), P(B|C).

Give an example of two non-disjoint independent events defined on Ω4 .

1
i.e., given the simultaneous occurrence of the events B1 , . . . , Bn .

79
5.2 Random Variables
In general the purpose of an experiment is to determine the value of quantities which depend
on the outcome of the experiment (e.g., the velocity of a particle, which is determined by
successive measurements of its position). We call such quantities random variables.

Definition 5.5. Let (Ω, P) be a finite probability space. A random variable is a function
X : Ω → R. If g : R → R, then the random variable Y = g(X) is said to be X-measurable.

Note that the property of Y being X-measurable means that the value attained by Y can
be inferred by the value attained by X, i.e., Y (ω) = g(X(ω)).
Since Ω = {ω1 , . . . , ωM }, then X can attain only a finite number of values, which we denote
x1 , . . . , xM , namely
X(ωi ) = xi , i = 1, . . . , M.
The values x1 , . . . , xM need not be distinct. If X(ωi ) = c, for all i = 1, . . . , M , we say that X
is a non-random, or deterministic, constant (the value of X is independent of the outcome
of the experiment).
The image of X is the finite set defined as

Im(X) = {x ∈ R such that X(ω) = x, for some ω ∈ Ω},

that is the set of possible values attainable by X. Given a ∈ R, we denote

{X = a} = {ω ∈ Ω : X(ω) = a},

which is the event that X attains the value a. Of course, {X = a} = ∅ if a ∈

/ Im(X). In
general, given I ⊆ R, we denote

{X ∈ I} = {ω ∈ Ω : X(ω) ∈ I},

which is the event that the value attained by X lies in the set I. Moreover we denote

{X = a, Y = b} = {X = a} ∩ {Y = b}, {X ∈ I1 , Y ∈ I2 } = {X ∈ I1 } ∩ {Y ∈ I2 }.

The probability that X takes value a is given by

X
P(X = a) = P({X = a}) = pi .
i:X(ωi )=a

If a ∈
/ Im(X), then P(X = a) = P(∅) = 0. More generally, given any open subset I of R, we
write X
P(X ∈ I) = P({X ∈ I}) = pi ,
i:X(ωi )∈I

80
which is the probability that the value of X belongs to I. For example, in the probability
space of a fair die consider the random variable

X(ω) = (−1)ω , ω ∈ {1, 2, 3, 4, 5, 6}. (5.6)

Then X(ω) = 1 if ω is even and X(ω) = −1 is ω is odd. Moreover

P(X = 1) = P({2, 4, 6}) = 1/2, P(X = −1) = P({1, 3, 5}) = 1/2,

whereas
P(X 6= ±1) = P(∅) = 0.
The set A = {2, 4, 6} is said to be resolved by X, because the occurrence of the event A (i.e.,
the fact that the outcome of the throw is an even number) is equivalent to X taking value 1.
In general, given a random variable X : Ω → R, the events resolved by X are the sets of the
form {X ∈ I}, for some I ⊆ R. These events comprise the so called information carried
by X. The idea is that even if the outcome of an experiment is unknown, measuring the
value attained by a random variable gives some information on the result of the experiment.
Definition 5.6. Given a random variable X : Ω → R, the function fX : R → [0, 1] defined
by
fX (x) = P(X = x)
is called the distribution of X, while FX : R → [0, 1] given by

FX (x) = P(X ≤ x), x∈R

is called the cumulative distribution of X.

Note that fX (x) is non-zero if only if x ∈ Im(X), and that FX is a non-decreasing function.
For example, for the random variable (5.6) defined on the probability space of a fair die we
have 
 0, x < −1,
FX (x) = 1/2, x ∈ [−1, 1),
1, x ≥ 1.


The probability that a random variable X takes value in the interval [a, b] can be written in
terms of the distribution of X as
X X
P(a ≤ X ≤ b) = P(X = xi ) = fX (xi ). (5.7)
i:X(ωi )=xi ∈[a,b] i:a≤xi ≤b

In a similar fashion, if g : R → R then

X X
P(a ≤ g(X) ≤ b) = P(X = xi ) = fX (xi ). (5.8)
i:g(X(ωi ))=g(xi )∈[a,b] i:a≤g(xi )≤b

We shall use these formulas later on.

81
5.2.1 Expectation and Variance
Next we define the expectation and variance of random variables. We may think of the
expectation of X as an estimate on the average value of X and the variance of X as a
measure of how far is this estimate from to the precise value of X.
Definition 5.7. Given a finite probability space (Ω, P), the expectation (or expected
value) of X : Ω → R is defined by
M
X
E[X] = X(ωi )P(ωi ).
i=1

We shall also write the definition of E[X] as

X
E[X] = X(ω)P({ω}). (5.9)
ω∈Ω

For instance, in the N -coin toss probability space (ΩN , Pp ) we have

X
E[X] = X(ω)pNH (ω) (1 − p)NT (ω) , (5.10)
ω∈ΩN

where NH (ω) is the number of heads and NT (ω) = N − NH (ω) is the number of tails in the
N -toss ω ∈ ΩN , see Definition 5.2.
Exercise 5.6 (?). Let X
: ΩN → R, X(ω) = NH (ω) − NT (ω). Compute E[X]. HINT: Use
−1
N
the identity k k = Nk−1


N.

We can rewrite the definition of expectation as

X
E[X] = x P(X = x),
x∈Im(X)

or equivalently, X
E[X] = xfX (x). (5.11)
x∈Im(X)

The importance of (5.11) is that it allows to compute the expectation of X from its distri-
bution, without any reference to the original probability space. For instance, if we are told
that a random variable X takes the following values:

 1 with probability 1/4
X= 2 with probability 1/4 , (5.12)
−1 with probability 1/2


then we can compute E[X] using (5.11) as

1 1 1 1
E[X] = 1 · +2· −1· = .
4 4 2 4
Some simple properties of the expectation are collected in the following theorem.

82
Theorem 5.1. Let X, Y be random variables on a finite probability space (Ω, P), g : R → R,
a, b ∈ R. The following holds:

1. E[aX + bY ] = aE[X] + bE[Y ] (linearity).

2. If X ≥ 0 and E[X] = 0, then X = 0.

3. If Y = g(X), then X
E[g(X)] = g(x)fX (x). (5.13)
x∈Im(X)

Exercise 5.7. Prove Theorem 5.1.

Definition 5.8. The variance of a random variable X : Ω → R is defined by

Var[X] = E[(E[X] − X)2 ].

Using the linearity of the expectation, we obtain easily the formula

Var[X] = E[X 2 ] − E[X]2 . (5.14)

Moreover Var[aX] = a2 Var[X] holds for all constants a ∈ R. The variance of the sum
of two random variables is in general different from the sum of their variances, unless the
random variables are uncorrelated (see Theorem 5.2 below). Furthermore the variance of
a random variable is always non-negative and it is zero if and only if the random variable
is a deterministic constant. Hence we may also interpret the variance as a measure of the
“randomness” of a random variable.
Using (5.13) with g(x) = x2 , we can rewrite the definition of variance in terms of the
distribution function of X as
 2
X X
Var[X] = x2 fX (x) −  xfX (x) , (5.15)
x∈Im(X) x∈Im(X)

which allows to compute Var[X] without any reference to the original probability space. For
instance for the random variable (5.12) we find
 2
1 1 1 1 27
Var[X] = 1 · + 4 · + 1 · − = .
4 4 2 4 16

Exercise 5.8 (?). Let R be the relative return of a portfolio which is long 1 share of the
derivative in Section 3.1.2. Compute E[R] and Var[R].

Remark 5.2. The variance of the relative return of a portfolio is a measure of the portfolio
risk, see Appendix A.

83
Example: mean of log return and volatility of the binomial stock price

Let 0 = t0 < t1 < · · · < tN = T be a partition of the interval [0, T ] with ti − ti−1 = h, for all
i = 1, . . . N . Given u > d, consider a random variable X such that X = u with probability
p and X = d with probability 1 − p. We may think of X as being defined on Ω1 = {H, T },
with X(H) = u and X(T ) = d. Now, the binomial stock price at time ti can be written as
S(ti ) = S(ti−1 ) exp(X). Hence the log-return of the stock in the interval [ti−1 , ti ] is

S(ti )
R = log S(ti ) − log S(ti−1 ) = log = X.
S(ti−1 )
It follows that the expectation and the variance of the log-return of the stock in the interval
[ti−1 , ti ] is
E[R] = E[X] = (pu + (1 − p)d),
Var[R] = Var[X] = [pu2 + (1 − p)d2 − (pu + (1 − p)d)2 )] = p(1 − p)(u − d)2 .
Thus the parameters α, σ 2 defined in (2.19) can be rewritten as
1 1
α= E[log S(ti ) − log S(ti−1 )], σ2 = Var[log S(ti ) − log S(ti−1 )]. (5.16)
h h
We see now that α is the expected log-return of the binomial stock price in the interval
[ti−1 , ti ] per unit of time (instantaneous log-return), while σ 2 is the instantaneous variance.
The parameter σ itself is called instantaneous volatility of the binomial stock. It is part of
the assumptions in the binomial model that the parameters α and σ are the same for every
interval [ti−1 , ti ] of the partition.

5.2.2 Independence and Correlation

We have seen before that a random variable X carries information. Now, if Y = g(X) for
some (non-constant) function g : R → R, then Y carries no more information than X: any
event resolved by knowing the value of Y is also resolved by knowing the value of X. The
other extreme case is when two random variables carry independent information.

Definition 5.9. Two random variables X1 , X2 : Ω → R are said to be independent if the

events {X1 ∈ I1 }, {X2 ∈ I2 } are independent events, for all regular2 sets I1 , I2 ⊂ R. This
means that
P(X ∈ I1 , X2 ∈ I2 ) = P(X1 ∈ I1 )P(X2 ∈ I2 ).
More generally, n random variables X1 , . . . , Xn : Ω → R are independent if the events
{X1 ∈ I1 }, {X2 ∈ I2 }, . . . , {Xn ∈ In } are independent (in the sense of Definition 5.4 ) for
all regular sets I1 , I2 , . . . In ⊂ R.

2
By regular sets we mean intervals or sets which can be written as the union of countably many intervals.

84
Exercise 5.9 (•). Show that when X, Y are independent random variables, then the only
events which are resolved by both variables are ∅ and Ω. Show that two deterministic constants
are always independent. Finally assume Y = g(X) and show that in this case the two random
variables are independent if and only if Y is a deterministic constant.

Note that the independence property is connected with the probability defined on the sample
space. Thus two random variables may be independent with respect to some probability and
not-independent with respect to another. We shall use later the following important result:

Theorem 5.2. Let X1 , X2 , . . . Xn be independent random variables, k ∈ {1, . . . , n − 1} and

g : Rk → R, f : Rn−k → R. Then the random variables

Y = g(X1 , X2 , . . . , Xk ), Z = f (Xk+1 , · · · , Xn )

are independent.

Exercise 5.10 (•). Prove the theorem for n = 2.

Definition 5.10. The covariance of two random variables X, Y : Ω → R is defined by

Cov(X, Y ) = E[(E[X] − X)(E[Y ] − Y )].

If Cov(X, Y ) = 0, the two random variables are said to be uncorrelated.

Using the linearity of the expectation we can rewrite the definition of covariance as

Cov(X, Y ) = E[XY ] − E[X]E[Y ]. (5.17)

The interpretation is the following: If Cov(X, Y ) > 0, then Y tends to increase (resp.
decrease) when X increases (resp. decrease), while if Cov(X, Y ) < 0, the two variables have
tendency to move in the opposition direction. For instance, assuming Var[X] > 0,

Cov(X, 2X) = 2Var[X] > 0, Cov(X, −2X) = −2Var[X] < 0.

Exercise 5.11 (•). Let (Ω, P) be a finite probability space and X, Y : Ω → R be two random
variables. Prove that X, Y independent ⇒ X, Y uncorrelated. Show with a counterexample
that the opposite implication is not true. Finally show that

Var[X + Y ] = Var[X] + Var[Y ] if and only if X, Y are uncorrelated. (5.18)

Exercise 5.12 (•). Assume that X, Y are not deterministic constants. Prove the inequality
p p
− Var[X]Var[Y ] ≤ Cov(X, Y ) ≤ Var[X]Var[Y ]. (5.19)

Show that the left (resp. right) inequality becomes an equality if and only if there exists a
negative (resp. positive) constant a0 and a real constant b0 such that Y = a0 X + b0 .

85
By inequality (5.19) it is convenient to introduce the following definition.
Definition 5.11. Let X, Y : Ω → R be two random variables such that Var[X] and Var[Y ]
are both positive (i.e., X, Y are not deterministic constants). Then
Cov(X, Y )
Cor(X, Y ) = p ∈ [−1, 1]
Var[X]Var[Y ]
is called the correlation of X, Y .

Hence, the closer is Cor(X, Y ) to 1 (resp. −1) the more Y has the tendency to move in the
same (resp. opposite) direction of X. For instance Cor(X, 2X) = 1, and Cor(X, −2X) = −1.
Exercise 5.13. Compute the correlation of the random variables X, Y : Ω3 → R given by
X(ω) = NT (ω) − NH (ω), Y (ω) = NH (ω).
Definition 5.12. If X, Y are two random variables on a finite probability space, then the
function fX,Y : R2 → [0, 1] given by
fX,Y (x, y) = P(X = x, Y = y)
is called the joint distribution of X and Y .
Exercise 5.14. Let X, Y have the joint distribution fX,Y . Show that the distributions of X
and Y are given by
X X
fX (x) = fX,Y (x, y), fY (y) = fX,Y (x, y)
y∈Im(Y ) x∈Im(X)

Show that X, Y are independent if and only if fX,Y (x, y) = fX (x)fY (y).

If the joint distribution is given, then the covariance of two random variables can easily be
computed without any reference to the original probability space. In fact, since
M
X X X
E[XY ] = X(ωi )Y (ωi )P(ωi ) = x y fX,Y (x, y), (5.20)
i=1 x∈Im(X) y∈Im(Y )

then
X X X X
Cov(X, Y ) = x y fX,Y (x, y) − xfX (x) y fY (y), (5.21)
x∈Im(X) y∈Im(Y ) x∈Im(X) y∈Im(Y )

where the distributions fX , fY are computed from fX,Y as shown in Exercise 5.14. In con-
clusion, if the joint distribution of two random variables X, Y is given, then all relevant
information on X, Y (independence, expectation, variance, correlation, etc.) can be inferred
without any reference to the original probability space.

Example. Consider two random variables X, Y such that Im(X) = {−1, 1, 3, 4}, Im(Y ) =
{−1, 0, 1, 2} and let their joint probability distribution be defined as in the following table

86
 Y ↓, X → -1 1 3 4
-1 1/64 2/64 1/64 4/64
0 5/64 1/64 9/64 6/64
1 6/64 10/64 1/64 12/64
2 2/64 1/64 1/64 2/64

For instance, fX,Y (−1, 1) = 1/64, fX,Y (−1, 0) = 5/64, and so on. Let us compute E[X],
E[Y ] and Cov(X, Y ).
X X X
E[X] = xfX (x) = x fX,Y (x, y)
x∈Im(X) x∈Im(X) y∈Im(Y )
X X X X
=− fX,Y (−1, y) + fX,Y (1, y) + 3 fX,Y (3, y) + 4 fX,Y (4, y)
y∈Im(Y ) y∈Im(Y ) y∈Im(Y ) y∈Im(Y )
1 5 6 2 2 1 10 1
= −( + + + )+( + + + )
64 64 64 64 64 64 64 64
1 9 1 1 4 6 12 2 33
+ 3( + + + ) + 4( + + + )= .
64 64 64 64 64 64 64 64 16
X X X
E[Y ] = yfY (y) = y fX,Y (x, y)
y∈Im(Y ) y∈Im(Y ) x∈Im(X)
X X X X
=− fX,Y (x, −1) + 0 · fX,Y (x, 0) + fX,Y (x, 1) + 2 fX,Y (x, 2)
x∈Im(X) x∈Im(X) x∈Im(X) x∈Im(X)
1 2 1 4 5 1 9 6
= −( + + + )+0·( + + + )
64 64 64 64 64 64 64 64
6 10 1 12 2 1 1 2 33
+( + + + ) + 2( + + + )= .
64 64 64 64 64 64 64 64 64
X X
E[XY ] = x y fX,Y (x, y)
x∈Im(X) y∈Im(Y )
1 2 1
+ 1(−1) + 3(−1)
= (−1)(−1)
64 64 64
55
+ ············ = .
64
Hence Cov(X, Y ) = E[XY ] − E[X]E[Y ] ≈ −0.204.
Exercise 5.15. Compute Cor(X, Y ).

5.2.3 Conditional expectation

If X, Y are independent random variables, knowing the value of Y does not help to estimate
the random variable X. However if X, Y are not independent, then we may use the infor-
mation on the value attained by Y to find an estimate of X which is better than E[X]. This
leads to the important concept of conditional expectation.

87
Definition 5.13. Let (Ω, P) be a finite probability space, X, Y : Ω → R random variables
and y ∈ R. The expectation of X conditional to Y = y (or given the event {Y = y}) is
defined as X
E[X|Y = y] = P(X = x|Y = y) x
x∈Im(X)

where P(X = x|Y = y) is the conditional probability of the event {X = x}, given the event
{Y = y} (see Def. 5.3). The random variable
E[X|Y ] : Ω → R, E[X|Y ](ω) = E[X|Y = Y (ω)]
is called the expectation of X conditional to Y .

In a similar fashion one defines the conditional expectation with respect to several random
variables, i.e., E[X|Y1 = y1 , Y2 = y2 , . . . YN = yN ] and E[X|Y1 , . . . , YN ]. For example, in the
probability space of a fair die, consider
X(ω) = (−1)ω , Y (ω) = (ω − 1)(ω − 2)(ω − 3), ω ∈ {1, 2, 3, 4, 5, 6}. (5.22)
Note that Im(Y ) = {0, 6, 24, 60}. Then we compute
E[X|Y = 0] = P(X = 1|Y = 0) − P(X = −1|Y = 0)
P(X = 1, Y = 0) P(X = −1, Y = 0)
= −
P(Y = 0) P(Y = 0)
P({2}) P({1, 3})
= − = −1/3.
P({1, 2, 3}) P({1, 2, 3})
Similarly we find
E[X|Y = 6] = 1, E[X|Y = 24] = −1, E[X|Y = 60] = 1,
hence E[X|Y ] is the random variable

 −1/3 if ω = 1, 2 or 3
E[X|Y ](ω) = 1 if ω = 4 or 6
−1 if ω = 5.


The following theorem collects a few important properties of the conditional expectation
that will be used later on.
Theorem 5.3. Let X, Y, Z : Ω → R be random variables on the finite probability space
(Ω, P). Then

1. The conditional expectation is a linear operator, i.e.,

E[αX + βY |Z] = αE[X|Z] + βE[Y |Z],
for all α, β ∈ R;

88
 2. If X is independent of Y , then E[X|Y ] = E[X];
3. If X is measurable with respect to Y , i.e., X = g(Y ) for some function g, then
E[X|Y ] = X;
4. E[E[X|Y ]] = E[X];
5. If X is measurable with respect to Z, then E[XY |Z] = XE[Y |Z];
6. If Z is measurable with respect to Y then E[E[X|Y ]|Z] = E[X|Z].

These properties remain true if the conditional expectation is taken with respect to several
random variables.

The interpretation of 2-3 in the previous theorem is the following: If X is independent of

Y , then the information carried by Y does not help to improve our estimate on X and thus
our best estimate for X remains E[X]. On the other hand, if X is measurable with respect
of Y , then by knowing Y we also know X and thus our best estimate on X is X itself.
Exercise 5.16 (?). What is the interpretation of 4, 5, 6 in Theorem 5.3?

5.3 Stochastic processes. Martingales

Definition 5.14. Let T > 0. A one parameter family of random variables, X(t) : Ω → R,
t ∈ [0, T ], is called a stochastic process. We denote a stochastic process by {X(t)}t∈[0,T ]
and by X(t, ω) the value of the random variable X(t) on the sample ω ∈ Ω. For each fixed
ω ∈ Ω, the curve t → X(t, ω), is called a path of the stochastic process.

We refer to the parameter t as the time variable. If X(t, ω) = C(t), for all ω ∈ Ω, i.e., if
the paths are the same for all sample points, we say that the stochastic process is a non-
random (or deterministic) function of time. If t runs over a (possibly infinite) discrete
set {t1 , t2 , . . . } ⊂ [0, T ], then we say that the stochastic process is discrete. Note that a
discrete stochastic process is equivalent to a sequence of random variables:
X = {X1 , X2 , . . . } where Xi = X(ti ), i = 1, 2, . . . .
As a way of example, consider the following (discrete) stochastic process defined on the
N -coin toss probability space (ΩN , Pp ):

1 if γi = H
ω = (γ1 , . . . , γN ) ∈ ΩN , Xi (ω) =
−1 if γi = T
Clearly, the random variables X1 , . . . , XN are independent and have all the same distribution.
In particular,
Pp (Xi = 1) = p, Pp (Xi = −1) = 1 − p, for all i = 1, . . . , N .

89
From now on we assume that the coin is fair (p = 1/2); we then have
E[Xi ] = 0, Var[Xi ] = 1, for all i = 1, . . . , N .
Moreover, for n = 0, . . . , N we set
n
X
M0 = 0, Mn = Xi , for n ≥ 1.
i=1

The stochastic process (M0 , . . . , MN ) is called symmetric random walk. It is centered in

zero, which means that
E[Mn ] = 0, for all n = 0, . . . , N .
Moreover, since it is the sum of independent random variables, the symmetric random walk
has variance given by
Var(Mn ) = Var(X1 + X2 + · · · + Xn ) = n,
see Theorem 5.2.
To understand the meaning of the term “random walk”, consider a particle moving on the
real line in the following way: if Xi = 1 (i.e., if the ith toss is a head), at time t = i the
particle moves one unit of length to the right, while if Xi = −1 (i.e., if the ith toss is a tail)
it moves one unit of length to the left. Hence Mn gives the total amount of units of length
that the particle has traveled to the right or to the left up to the time n. The increments
of the random walk are defined as follows. Given 0 = k0 < k1 < · · · < km = N , we let
ki+1
X
∆i = Mki+1 − Mki = Xj , i = 0, . . . , m − 1,
j=ki +1

i.e., ∆0 = Mk1 , ∆1 = Mk2 − Mk1 , . . . , ∆m−1 = Mkm − Mkm−1 . Hence ∆j is the total displace-
ment of the particle from time kj−1 to time kj . It follows by Theorem 5.2 that the increments
of the random walk are independent random variables, that is to say, the distance traveled
by the particle in the time interval [kj−1 , kj ] is independent of the movements during any
earlier or later time interval. Moreover
E[∆i ] = 0, Var[∆i ] = ki+1 − ki . (5.23)
Exercise 5.17 (•). Let T > 0 and n ∈ N be given. Define the stochastic process
1
{Wn (t)}t∈[0,T ] , Wn (t) = √ M[nt] , (5.24)
n
where [z] denotes the greatest integer smaller than or equal to z and Mk = X1 + X2 +
· · · + Xk , k = 1, . . . , N , is a symmetric random walk. It is assumed that the stochas-
tic process (X1 , . . . , XN ) is defined for N > [nT ], so that Wn (t) is defined for all t ∈
[0, T ]. Compute E[Wn (t)], Var[Wn (t)], Cov[Wn (t), Wn (s)]. Show that Var(Wn (t)) → t and
Cov(Wn (t), Wn (s)) → min(s, t) as n → +∞.

90
 25

20

15

10

200 400 600 800 1000

-5

-10

Figure 5.1: A path of the stochastic process (5.24) for n = 1000.

Remark 5.3. For large n ∈ N, the process {Wn (t)}t∈[0,T ] can be used an approximation
for the Brownian motion, see Definition 5.18 below. An example of path of the stochastic
process {Wn (t)}t∈[0,T ] for n = 1000 is shown in Figure 5.1.
Exercise 5.18 (Matlab). Write a Matlab function that generates a random path of the
stochastic process {Wn (t)}t∈[0,T ] .

To conclude this section we introduce the fundamental concept of martingale. Roughly

speaking, a martingale is a stochastic process which has no tendency to rise or fall. The
precise definition is the following.
Definition 5.15. A discrete stochastic process {X1 , X2 , . . . } on the finite probability space
(Ω, P) is called a martingale if
E[Xi+1 |X1 , X2 , . . . Xi ] = Xi , for all i ≥ 1. (5.25)

The interpretation is the following: The variables X1 , X2 , . . . Xi contains the information

obtained by “looking” at the stochastic process up to the step i. For a martingale process,
this information is not enough to infer whether, in the next step, the process will raise or
fall. Note carefully that the martingale property depends on the probability being used: if
{X1 , X2 , . . . } is a martingale in the probability P and P0 is another probability measure on
the sample space Ω, then {X1 , X2 , . . . } need not be a martingale with respect to P0 .
Remark 5.4. Taking the expectation on both sides of (5.25) and using property 4 in
Theorem 5.3 we obtain
E[Xi+1 ] = E[Xi ], for all i ≥ 1.
Thus, iterating, E[Xi ] = E[X1 ], for all i ≥ 1, i.e., martingales have constant expectation.

91
An example of martingale is the random walk introduced above. In fact, using the linearity
of the conditional expectation we have

E[Mn |M1 , . . . , Mn−1 ] = E[Mn−1 + Xn |M1 , . . . , Mn−1 ]

= E[Mn−1 |M1 , . . . , Mn−1 ] + E[Xn |M1 , . . . , Mn−1 ].

As Mn−1 is measurable with respect to M1 , . . . , Mn−1 , then E[Mn−1 |M1 , . . . , Mn−1 ] = Mn−1 ,
see Theorem 5.3 (3). Moreover, as Xn is independent of M1 , . . . , Mn−1 , Theorem 5.3 (2)
gives E[Xn |M1 , . . . , Mn−1 ] = E[Xn ] = 0. It follows that E[Mn |M1 , . . . , Mn−1 ] = Mn−1 , i.e.,
the random walk is a martingale.

5.4 Applications to the binomial model

In this section we reformulate the binomial options pricing model using the language of
probability theory. We first show that the binomial stock price can be interpreted as a
stochastic process defined on the N -coin toss probability space (ΩN , Pp ), see Definition 5.2.
Recall that, for a given 0 < p < 1, the binomial stock price at time t is given by

S(t − 1)eu with probability p
S(t) = (5.26)
S(t − 1)ed with probability 1 − p,

where u > d. Now, for t ∈ I = {1, . . . , N }, consider the following random variable

u, if the tth toss in ω is H
Xt : ΩN → R, Xt (ω) = .
d, if the tth toss in ω is T
Hence we can write

S(t) = S(t − 1)eXt = S(t − 2)eXt−1 +Xt = · · · = S0 exp(X1 + X2 + · · · + Xt ) : ΩN → R,

where S(0) is the initial value of the stock price, which is a deterministic constant. Thus
S(t) is a random variable and therefore {S(t)}t∈I is a (discrete) stochastic process. For each
ω ∈ ΩN , the vector (S(1, ω), . . . , S(N, ω)) is a path for the stock price.
The value at time t of the risk-free asset is the deterministic function of time B(t) =
B0 exp(rt), where r is the (constant) interest rate of the money market and B0 is the initial
value of the risk-free asset. Recall that S ∗ (t) = e−rt S(t) is called the discounted price of the
stock, see Remark 2.3.
Theorem 5.4. If r ∈ / (d, u), there is no probability measure Pp on the sample space ΩN such
that the discounted stock price process {S ∗ (t)}t∈I is a martingale. For r ∈ (d, u), {S ∗ (t)}t∈I
is a martingale with respect to the probability measure Pp if and only if p = q, where

er − ed
q= .
eu − ed

92
Proof. By definition, {S ∗ (t)}t∈I is a martingale if and only if

E[e−rt S(t)|S ∗ (1), . . . , S ∗ (t − 1)] = e−r(t−1) S(t − 1), for all t ∈ I.

Clearly, taking the expectation conditional to S ∗ (1), . . . , S ∗ (t − 1) is the same as taking the
expectation conditional to S(1), . . . , S(t − 1), hence the above equation is equivalent to

E[S(t)|S(1), . . . , S(t − 1)] = er S(t − 1), for all t ∈ I, (5.27)

where we canceled out a factor e−rt in both sides of the equation. Moreover
S(t)
E[S(t)|S(1), . . . , S(t − 1)] = E[ S(t − 1)|S(1), . . . , S(t − 1)]
S(t − 1)
S(t)
= S(t − 1)E[ |S(1), . . . , S(t − 1)],
S(t − 1)
where we used that S(t − 1) is measurable with respect to the conditioning variables and
thus can be taken out from the conditional expectation (see property 5 in Theorem 5.3). As
 u
e with prob. p
S(t)/S(t − 1) =
ed with prob. 1 − p

is independent of S(1), . . . , S(t − 1), then by Theorem 5.3(2) we have

S(t) S(t)
E[ |S(1), . . . , S(t − 1)] = E[ ] = eu p + ed (1 − p).
S(t − 1) S(t − 1)

Hence (5.27) holds if and only if eu p + ed (1 − p) = er . The latter has a solution p ∈ (0, 1) if
and only if r ∈ (d, u) and the solution, when it exists, is unique and given by p = q.
Remark 5.5. Due to Theorem 5.4, Pq is called martingale probability measure. More-
over we can reformulate Theorem 2.3 as follows: a 1+1 dimensional binomial stock market
is arbitrage free if and only if there exists a martingale probability measure.

Since martingales have constant expectation (see Remark 5.4), we obtain the important
result
Eq [S(t)] = S0 ert . (5.28)
Thus in the martingale probability measure one expects the same return on the stock as on
the risk-free asset. For this reason, Pq is also called risk neutral probability. However,
as shown in the following exercise, the situation is very different in the physical (real world)
probability.
Exercise 5.19 (•). Let Ep [·] denote the expectation in the probability measure Pp . Show that

Ep [S(N )] = S(0)(eu p + ed (1 − p))N , Eq [S(N )] = S(0)erN .

Compute also Varp [S(N )].

93
The value of a portfolio position (hS , hB ) invested on hS shares of the stock and hB shares
of the risk-free asset is the stochastic process {V (t)}t∈I , where

V (t) = hB B(t) + hS S(t) : ΩN → R, t ∈ I,

while V (0) = hS S0 + hB B0 is a deterministic constant. If we change the portfolio position

depending on the price of the stock (i.e., depending on ω ∈ ΩN ), then we get a port-
folio (stochastic) process {(hS (t), hB (t))}t∈I . We recall that (hS (t), hB (t)) corresponds to
the portfolio position held in the interval (t − 1, t]. The portfolio process is predictable if
hS (t), hB (t) are measurable with respect to S(1), . . . , S(t − 1). The value {V (t)}t∈I of the
portfolio process is the stochastic process given by

V (t) = hB (t)B(t) + hS (t)S(t) : ΩN → R, t ∈ I. (5.29)

A portfolio process {(hS (t), hB (t))}t∈I is said to be self-financing if

V (t − 1) = hB (t)B(t − 1) + hS (t)S(t − 1). (5.30)

Theorem 5.5 (∗). Let {(hS (t), hB (t))}t∈I be a self-financing predictable portfolio with value
{V (t)}t∈I . Then the discounted portfolio value V ∗ (t) = e−rt V (t) is a martingale in the
risk-neutral measure. Moreover the following identity holds:

V ∗ (t) = Eq [V ∗ (N )|S(1), . . . , S(t)]. (5.31)

In particular at time t = 0 we obtain V (0) = e−rN Eq [V (N )], which is equivalent to (2.10).

Proof. The martingale claim is

Eq [V ∗ (t)|V ∗ (1), . . . , V ∗ (t − 1)] = V ∗ (t − 1).

We now show that this follows by

Eq [V ∗ (t)|S(1), . . . , S(t − 1)] = V ∗ (t − 1). (5.32)

In fact, by taking the conditional expectation in the risk-neutral measure of (5.32) with
respect to V ∗ (1), . . . , V ∗ (t − 1), we obtain

V ∗ (t − 1) = Eq [V ∗ (t − 1)|V ∗ (1), . . . , V ∗ (t − 1)]

= Eq Eq [V ∗ (t)|S(1), . . . , S(t − 1)]|V ∗ (1), . . . , V ∗ (t − 1)
 

= Eq [V ∗ (t)|V ∗ (1), . . . , V ∗ (t − 1)];

here we have used property 3 of Theorem 5.3 in the first step and property 6 in the last step.
The latter is possible because V ∗ (t) is measurable with respect to S(1), . . . , S(t) (being the
portfolio process predictable). Now we claim that (5.32) also implies the formula (5.31). We

94
argue by backward induction3 . Letting t = N in (5.32) we see that (5.31) holds at t = N − 1.
Assume now that holds (5.31) at time t + 1, i.e.,

V ∗ (t + 1) = Eq [V ∗ (N )|S(1), . . . , S(t + 1)].

Taking the conditional expectation with respect to S(1), . . . , S(t) in the risk neutral measure
we have, by (5.32),
h i
∗ ∗ ∗
V (t) = Eq [V (t + 1)|S(1), . . . , S(t)] = Eq Eq [V (N )|S(1), . . . , S(t + 1)]|S(1), . . . , S(t)
= Eq [V ∗ (N )|S(1), . . . , S(t)],

hence (5.32) ⇒ (5.31), as claimed. It remains to prove (5.32). As B(t) = B(t − 1)er , (5.30)
gives
hB (t)B(t) = er V (t − 1) − hS (t)S(t − 1)er .
Replacing in (5.29) we find

V (t) = er V (t − 1) + hS (t)[S(t) − S(t − 1)er ].

Taking the conditional expectation in the risk neutral measure with respect to the random
variables S(1), . . . , S(t − 1) we obtain

Eq [V (t)|S(1), . . . S(t − 1)] = er Eq [V (t − 1)|S(1), . . . , S(t − 1)]

+ Eq [hS (t)(S(t) − S(t − 1)er )|S(1), . . . , S(t − 1)]. (5.33)

As V (t − 1) and hS (t) are measurable with respect to the conditioning variables we have
Eq [V (t − 1)|S(1), . . . , S(t − 1)] = V (t − 1), as well as

Eq [hS (t)(S(t) − S(t − 1)er )|S(1), . . . , S(t − 1)]

= hS (t)Eq [S(t) − S(t − 1)er |S(1), . . . , S(t − 1)]
 
= hS (t) Eq [S(t)|S(1), . . . , S(t − 1)] − S(t − 1)er = 0,

where in the last step we used that {S ∗ (t)}t∈I is a martingale in the risk-neutral measure.
Going back to (5.33) we obtain

Eq [V (t)|S(1), . . . S(t − 1)] = er V (t − 1),

which is the same as (5.32).

Now we can use the martingale property of {V ∗ (t)}t∈I to give a simple proof that the
existence of a martingale probability implies the absence of arbitrage. In fact, assume that
{hS (t), hB (t)}t∈I is an arbitrage (see Definition 2.4). Then V ∗ (0) = 0 and since martingales
have constant expectation then E[V ∗ (t)] = 0, for all t ∈ {0, 1 . . . , N }. But V ∗ (N ) ≥ 0, hence
3
Note the similarity of this proof with the one of Theorem 2.2.

95
by 2 of Theorem 5.1 it must be V ∗ (N, x) = 0 along any path x ∈ {u, d}N . Thus the portfolio
is not an arbitrage.
Now let Y : ΩN → R be a random variable and consider the European-style derivative
with pay-off Y : ΩN → R at maturity time N . This means that the derivative can only be
exercised at time t = N (for standard European derivatives Y is a deterministic function of
S(N )). Let ΠY (t) be the binomial fair price of the derivative a time t. By definition, ΠY (t)
equals the value V (t) of self-financing, hedging portfolios. In particular, ΠY (t) is a random
variable and so {ΠY (t)}t∈I is a stochastic process. Using the hedging condition V (N ) = Y
(which means V (N, ω) = Y (ω), for all ω ∈ ΩN )) and (5.31), we have the following formula
for the fair price at time t of the financial derivative:
ΠY (t) = e−r(N −t) Eq [Y |S(1), . . . , S(t)]. (5.34)
Equation (5.34) is known as risk neutral pricing formula and it is the cornerstone of
options pricing theory. It holds not only for the binomial model but for any discrete—or
even continuum—pricing model for financial derivatives. It is used for standard as well as
non-standard European derivatives. In the special case t = 0, (5.34) reduces to
ΠY (0) = e−rN Eq [Y ]. (5.35)
Exercise 5.20 (Put-call parity for Asian options (•)). Consider a N -period arbitrage-
free binomial market with r 6= 0 and let S(t) denote the price of the stock at time t ∈
{0, . . . , N }. The Asian call, resp. put, with maturity T = N and strike K is the non-
standard European style derivative with pay-off
" N
# " N
#
 1 X   1 X 
Ycall = S(t) − K , resp. Yput = K − S(t) .
N + 1 t=0 N + 1 t=0
+ +

Denote by AC(0) and AP (0) the binomial price at time t = 0 of the Asian call and put,
respectively. Prove the following put-call parity identity:
h S(0) er(N +1) − 1 i
AC(0) − AP (0) = e−rN − K .
N + 1 er − 1
1−αN +1
HINT: For α 6= 1, N k
P
k=0 α = 1−α
.

5.4.1 Binomial distribution

Let us now derive the distribution of the stock binomial price S(t), t ∈ I. First we notice
that the image of S(t) is Im S(t) = {s0 , . . . , st }, where
sk = S(0) exp(ku + (t − k)d), k = 0, . . . , t.
Hence fS(t) (x) = 0, if x 6= sk , for all k = 0, . . . t and t ∈ I. The value of S(t) is sk if there
are exactly k heads in the first t tosses of the N -toss, which shows that fS(t) is given by the

96
so-called binomial distribution:
 
t k
fS(t) (sk ) = Pp [S(t) = sk ] = p (1 − p)t−k , k = 0, . . . , t, t ∈ I.
k
In particular, in the case of a fair coin p = 1/2, we have
 
t 1
fS(t) (x) = δ(x − S(0) exp(ku + (t − k)d)) (fair coin),
k 2t
where δ(z) = 1 if z = 0 and δ(z) = 0 for z 6= 0. It is convenient to express this result
in terms of the log-price of the stock log S(t), as this can take both positive and negative
values. For a fair coin we get
 
t 1
flog(S(t)/S(0)) (x) = δ(x − (ku + (t − k)d). (5.36)
k 2t
An example of this distribution is depicted in Figure 5.2. As it is clear from the picture,
for large values of N ∈ N, the distribution of log(S(N )/S(0)) = log S(N ) − log S(0) can be
approximated by a normal distribution, i.e., a distribution of the form (5.39) below4 . One
of the main critics to the binomial model is that it assigns very low probabilities to large
variations of the (log-)stock price.
The binomial distribution can be used to compute the probability that the price of the stock
lies in an interval [a, b] at time t. By (5.7) we have
X X t
P(a ≤ S(t) ≤ b) = fS(T ) (sk ) = pk (1 − p)t−k k = 0, . . . , t, t ∈ I.
k:a≤s ≤b k:a≤s ≤b
k
k k
(5.37)
Moreover if g : R → R then, by (5.8),
 
X X t k
P(a ≤ g(S(t)) ≤ b) = fS(T ) (sk ) = p (1−p)t−k k = 0, . . . , t, t ∈ I.
k
k:a≤g(sk )≤b k:a≤g(sk )≤b
(5.38)
The formula (5.38) can be used to compute the probability that a derivative on the stock is
in the money at time t. Consider a standard European or American derivative with pay-off
Y = g(S(N )) at the expiration date T = N . Then the probability that the derivative is in
the money at time t is given by
X t
P(Y (t) > 0) = P(g(S(t)) > 0) = pk (1 − p)t−k .
k
k:g(sk )>0

These formulas can be easily implemented with Matlab.

Exercise 5.21 (Matlab). Write a Matlab function ProbDerivative(S,t,g,p) which computes
the probability that a standard derivative with pay-off function g is in the money at time t.
4
A rigorous proof of the fact that the distribution of log S(N )/S(0) converges to a normal distribution in
the time-continuum limit is given in Theorem 6.3.

97
 f S HN L HxL

0.08

0.06

0.04

0.02

x
-100 -50 50 100

Figure 5.2: The distribution of log(S(N )). It is assumed S(0) = 1, u = −d = 1, p = 1/2

and N = 100.

5.4.2 General discrete options pricing models

Equation (5.34) is the most important result of this chapter. It leads to the following general
strategy to introduce discrete options pricing models for European derivatives:

1. Fix a finite probability space (Ω, P(λ)). The probability measures depends on a set of
parameters λ = (λ1 , . . . , λn ) (e.g., λ = p in the N -coin toss probability space)
2. Define the stock price as a positive stochastic process on the probability space (Ω, P(λ)).
3. Prove the existence of a probability measure Q = P(λ0 ) which makes the stock price
process a martingale. The existence of a martingale measure ensures that the market
is arbitrage-free
4. Provided the martingale measure is unique (i.e., provided λ0 is unique), define the fair
price of European derivatives by (5.34). If the martingale measure is not unique, then
the price of the derivative is not uniquely defined and the model is called incomplete.

In fact, the above strategy is applied for all models in options pricing theory, even for time-
continuum models, which are defined on uncountable probability spaces. An example of

98
time-continuum model is the Black-Scholes model considered in the next chapter (see [7] for
an introduction to general time-continuum models). The important lesson to be learned here
is the following: probability theory is the right framework where to formulate the mathematical
models in options pricing theory.
Exercise 5.22. Formulate a discrete options pricing model in which, at any time step, the
stock price can go up, down or stay the same (trinomial model). Use the language of
probability theory. Show that this model is incomplete (i.e., the martingale measure is not
unique). HINT: For the last part of the exercise one can restrict to the one-period model.

5.4.3 Quantitative vs fundamental analysis of a stock

To continue this section we discuss briefly the difference between the fundamental analysis
and the quantitative (or statistical) analysis of a stock. Performing a quantitative anal-
ysis means that the investor tries to estimate the price of the stock using a mathematical
probabilistic model, such as the binomial model. As opposed to this, a fundamental analysis
of the stock consists in a careful evaluation of the performance of the underlying company
and of other economical factors which may affect the value of the stock. As the quantitative
analysis, also the fundamental analysis can be formulated in the language of probability
theory (although in practice nobody does that!). Precisely, let S(T ) be the price of the stock
at some future time T > 0 (e.g., T = 1 year from now) and let us start by making a list
of all events which we think may affect the value of S(T ). Let us call {ω1 }, {ω2 }, . . . , {ωM }
such events; for instance
{ω1 } = [the company will make a profit this year]
{ω2 } = [there will be an oil crisis this year]
{ω3 } = [the company will open a new branch]
etc.
In this way we have constructed a sample space Ω = {ω1 , . . . , ωM }. Next we have to assign a
probability to each of these atomic events, say P({ωi }) = pi . Finally we have to specify how
these events affect the price of the stock, that is to say, we have to give S(T, ω), for ω ∈ Ω.
After all of this we can compute for instance
X
E[S(T )] = S(T, ω)P({ω}), Var[X] = E[X 2 ] − E[X]2 ,
ω∈Ω

which are our “expected value” for the stock price and our “expected error” of this value.
Clearly all the steps that led us to these estimates are quite subjective.
In the quantitative analysis we only try to “guess” what the distribution of the stock price
will be at time T , i.e., we assign fS(T ) : (0, ∞) → [0, 1] and then we derive our estimates
using the formulas
X
E[S(T )] = s fS(T ) (s), Var[X] = E[X 2 ] − E[X]2 ,
s∈I

99
where I is a (finite) set of possible values that we admit for the stock price. Note that in the
quantitative approach the only arbitrariness is in the choice of the distribution function of
the price. In the binomial model we assume that log S(T ) follows a binomial distribution, but
other choices are possible and can be justified by looking at the historical data for the stock
price. In fact, the advantage of the quantitative method versus the fundamental approach,
is that, by employing the stock price as the only relevant information, it provides us with
an objective (quantitative) way to justify, monitor and adjust our analysis.

5.5 Infinite Probability Spaces

In this last section we consider briefly probability spaces consisting of an infinite number of
sample points. The discussion can be complemented with Chapter 4 in [3].
Assume first that Ω is a countable set. This means that

Ω = {ωn }n∈N .

For countable sample spaces the definitions given in the previous sections for finite sets
extend straightforwardly. Precisely, given a sequence
X
p = (pn )n∈N such that 0 < pn < 1, pn = 1,
n∈N

we define the probability of the atomic events as

P({ωn }) = pn .

If A ∈ 2Ω , then we define X X
P(A) = pi = P({ω}).
i:ωi ∈A ω∈A

If X : Ω → R is a random variable, then

X X
E[X] = X(ωn )pn = X(ω)P({ω}).
n∈N ω∈Ω

The remaining definitions introduced in the finite case (variance, covariance, independent
random variables, etc.) continue to be valid for countable probability spaces.
When Ω is uncountable, there is no general procedure to construct a probability space, but
only an abstract definition (which we shall not give). The problem is that in the uncountable
case Ω admits very irregular (“wild”) sets and thus defining a probability over the whole 2Ω
becomes complicated. Moreover the occurrence of non-trivial events with zero probability
poses some technical problems. We restrict ourselves to present some examples.

100
• Let Ω = (0, 1) and let the admissible events be given by the sets which can be written
as the union of countably many open subintervals of (0, 1). For any admissible event
A ⊆ (0, 1) we define Z 1
P(A) = IA (x) dx,
0

where IA (x) = 1 if x ∈ A and zero otherwise.

• Let Ω = R and p : R → [0, ∞) be a continuous function such that

Z
p(x) dx = 1.
R

As admissible events we consider all real sets which can be written as the union of
countably many open intervals. For any admissible event A ∈ R we define
Z
P(A) = p(x) dx.
A

• In this example we construct a probability space that extends (ΩN , Pp ) (the N -coin
toss probability space) to N = ∞. Consider Ω∞ = {(γn )n∈N ; γi = H or T }. Thus
an element of Ω∞ is the outcome of the experiment “tossing a coin infinitely many
times”. It can be shown (using a standard Cantor diagonal argument), that the set
Ω∞ is uncountable. Now, given ω̄ = (γ̄1 , . . . , γ̄N ) ∈ ΩN , consider the events

AN (ω̄) ⊂ Ω∞ , A = {(γn )n∈N ∈ Ω∞ : γi = γ̄i , i = 1, . . . , N },

that is to say, AN (ω̄) contains all infinity-tosses whose first N -tosses coincide with
(γ̄1 , . . . , γ̄N ). Of course, AN (ω̄) is also uncountable. We define the probability of
AN (ω̄) as the probability of (γ̄1 , . . . , γ̄N ) in the probability space (ΩN , Pp ). For instance,
assuming that the coin is fair, the event

[the first 5 tosses in the infinite sequence are heads]

has probability (1/2)5 . By letting ω̄ varies in ΩN , we get a collection AN of 2N events

in Ω∞ ,
AN ⊂ 2Ω∞ , AN = {AN (ω̄); ω̄ ∈ ΩN },
whose probability has been defined. This definition can be extended to all events
that can be written as the disjoint union of events in the set AN , using the rule
P(A∪B) = P(A)+P(B) if A, B are disjoint. In this way we have assigned a probability
to all events in the space Ω∞ that depends only on the first N tosses, and letting N
arbitrarily large, we are able to define a probability to any event which depends on
a (arbitrary) finite number of tosses (e.g., the event “there is a tail in the first 10100
tosses”). This probability space might seem quite large, but actually is not! For
instance, the event “there exists at least one head in the infinite-toss” does not belong

101
 to this space, hence we cannot assign a probability to it. The inclusion of events
which are resolved by tossing the coin infinitely many times requires advanced tools in
probability theory (in particular, Carathéodory’s theorem), which will not be discussed
here.

Fortunately for most applications (and in particular for those in financial mathematics) the
knowledge of the full probability space is usually not necessary, as in the applications one
is typically concerned only with random variables and their distributions, rather than with
generic events. More precisely, we are only interested in assigning a probability to events of
the form {X ∈ I}, where X is a random variable on the (abstract) probability space and
I ⊂ R, that is to say, events which can be resolved by one (or more) random variables, cf.
the discussion in Section 5.4.3.
Definition 5.16. Let fX : R → [0, ∞) be a continuous function, except possibly on finitely
many points. A random variable X : Ω → R is said to have probability density fX if
Z
P(X ∈ A) = fX (x) dx,
A

for all regular sets A ⊆ R.

Note that if fX is the probability density of a random variable X, then necessarily

Z
fX (x) dx = 1.
R

Moreover the cumulative distribution FX (x) = P(X ≤ x) satisfies

Z x
dFX
FX (x) = fX (y) dy, for all x ∈ R, hence fX = .
−∞ dx

Examples

• A random variable X : Ω → R is said to be normal, or to have a normal distribution,

with mean m ∈ R and variance σ 2 > 0 if
|x − m|2
 
1
fX (x) = √ exp − . (5.39)
2πσ 2 2σ 2

We denote N (m, σ 2 ) the set of all such random variables. A variable X ∈ N (0, 1) is
called a standard normal random variable. The cumulative distribution of standard
normal random variables is denoted by Φ(x) and is called the standard normal
distribution, i.e., Z x
1 1 2
Φ(x) = √ e− 2 y dy.
2π −∞

102
 • A random variable X : Ω → R is said to be exponential, or to have an exponential
distribution, with intensity λ > 0 if
 −λx
λe x ≥ 0,
fX (x) =
0 x < 0.
We denote E(λ) the set of all exponential random variables with intensity λ.

5.5.1 Joint distribution. Independence

Definition 5.17. Two random variables X, Y : Ω → R are said to have the joint proba-
bility density fX,Y : R2 → [0, ∞), if
Z Z
P(X ∈ A, Y ∈ B) = fX,Y (x, y) dx dy,
A B

for all regular sets A, B ⊆ R.

The generalization of the previous definition to n variables is straightforward. Note that if

fX,Y is a joint probability density, then
Z Z
fX,Y (x, y) dx dy = 1.
R R

Example: Jointly normally distributed random variables. Let m ∈ R2 and C =

(Cij )i,j=1,2 be a symmetric, positive definite 2×2 matrix. Two random variables X1 , X2 :
Ω → R are said to be jointly normally distributed with mean m and covariance matrix C
if they admit the joint density
 
1 1 −1
fX1 ,X2 (x) = p T
exp − (x − m) · C · (x − m) , for all x = (x1 , x2 ) ∈ R2 ,
(2π)2 det C 2
(5.40)
where “ · ” denotes the row-by-column product of matrices. This definition extends straight-
forwardly to n variables.
The following theorem, which we give without proof, shows that the probability density,
when it exists, provides all the relevant statistical information on a random variable.
Theorem 5.6. The following holds for all sufficiently regular5 functions g : R → R:

(i) Let X : Ω → R be a random variable with density fX . Then for all regular sets A ⊆ R,
Z
P(g(X) ∈ A) = fX (x) dx,
x:g(x)∈A

which extends (5.8) to general probability spaces.

5
In particular, for all functions g such that the integrals in the theorem are well-defined.

103
 (ii) Let X : Ω → R be a random variable with density fX . Then
Z
E[g(X)] = g(y)fX (y) dy,
R

which extends (5.13) to general probability spaces.

(iii) Let X, Y : Ω → R be random variables with joint density fX,Y . Then

Z
E[XY ] = xyfX,Y (x, y) dx dy,
R2

which extends (5.20) to general probability spaces.

As an example of application of (i), suppose X ∈ N (0, 1). Then

Z 1
1 x2
2
P(X ≤ 1) = √ e− 2 dx ≈ 0.683,
2π −1

that is to say, a standard normal random variable has about 68,3% probability to take value
on the interval (−1, 1). Let us see some further applications of Theorem 5.6. By (ii), the
expectation and the variance of a random variable X with density fX are given by
Z Z Z 2
2 2 2
E[X] = xfX (x) dx, Var[X] = E[X ] − E[X] = x fX (x) dx − xfX (x) dx .
R R R

These formulas generalize (5.11) and (5.15) to random variables on general sample spaces
which admit a density distribution. In particular, for normal variables we obtain

X ∈ N (m, σ 2 ) =⇒ E[X] = m, Var[X] = σ 2 . (5.41)

Exercise 5.23. Prove (5.41). Compute the expectation and the variance of exponential
random variables.

By (iii) of Theorem 5.6, if X1 , X2 have the joint density fX1 ,X2 , then
Z
Cov(X1 , X2 ) = x1 x2 fX1 ,X2 (x1 , x2 ) dx1 dx2 − E[X1 ]E[X2 ],
R2

which generalizes (5.21). In particular, if X1 , X2 are jointly normal distributed with mean
m ∈ R2 and covariance matrix C = (Cij )i,j=1,2 , we find

m = (m1 , m2 ), Cij = Cov(Xi , Xj ). (5.42)

Exercise 5.24. Prove (5.42).

104
The next thing we need to know about general probability spaces is how to determine when
two random variables are independent. To this regard we have the following theorem.
Theorem 5.7 (∗). The following holds.

(i) If two random variables X, Y admit densities fX , fY and are independent, then they
admit the joint density
fX,Y (x, y) = fX (x)fY (y).
(ii) If two random variables X, Y admit a joint density fX,Y of the form
fX (x, y) = u(x)v(y),
for some functions u, v : R → [0, ∞), then X, Y are independent and admit densities
fX , fY given by
1
fX (x) = cu(x), fY (y) = v(y),
c
where Z  Z −1
c= v(x) dx = u(y) dy .
R R

Proof. As to (i) we have

Z Z
P(X ∈ A, Y ∈ B) = P(X ∈ A)P(Y ∈ B) = fX (x) dx fY (y) dy
ZA B

= fX (x)fY (y) dx dy.

A×B

To prove (ii), we first write

{X ∈ A} = {X ∈ A} ∩ Ω = {X ∈ A} ∩ {Y ∈ R} = {X ∈ A, Y ∈ R}.
Hence, by definition of joint density,
Z Z ∞ Z Z Z
P(X ∈ A) = fX,Y (x, y) dx dy = u(x) dx v(y) dy = cu(x) dx
A −∞ A R A
R
where c = R v(x) dx. Thus X admits the density fX (x) = c Ru(x). At the same fashion one
proves that Y admits the density fY (y) = c0 v(y), where c0 = R u(y)dy. Since
Z Z Z Z
1= fX,Y (x, y) dx dy = u(x) dx v(y) dy = c0 c,
R R R R
0
then c = 1/c. It remains to prove that X, Y are independent. This follows by
Z Z Z Z
P(X ∈ A, Y ∈ B) = fX,Y (x, y) dx dy = u(x) dx v(y) dy
ZA B Z A
Z B Z
1
= cu(x) dx v(y) dy = fX (x) dx fY (y) dy
A B c A B
= P(X ∈ A)P(Y ∈ B).

105
Exercise 5.25. Let X ∈ N (0, 1) and Y ∈ E(1) be independent. Compute P(X ≤ Y ).

As an application of Theorem 5.7 we now prove the following result.

Theorem 5.8. Let X1 , X2 : Ω → R be normal random variables which are jointly normally
distributed. Then the following properties are equivalent:

(a) X1 , X2 are independent;

(b) X1 , X2 are uncorrelated.

Proof. (a)=⇒(b) is always true, for all random variables. As to the implication (b)=⇒(a),
by (5.42) we have C12 = C21 = 0. Substituting in (5.40) we obtain the fX1 ,X2 has the form
fX1,X2 (x1 , x2 ) = u(x1 )v(x2 ), and so the claim follows by (ii) of Theorem 5.7.

The next theorem shows that independent normal random variables form a linear space.
Theorem 5.9. If X1 , X2 are independent random variables such that Xi ∈ N (αi , σi2 ), i =
1, 2, then X1 + X2 ∈ N (α1 + α2 , σ12 + σ22 ).

A way to prove Theorem 5.9 is suggested in the additional exercises for Chapter 5 in Ap-
pendix D.

5.5.2 Central limit theorem

The normal distribution is certainly the most important random variables distribution used
in finance (and in many other applications of probability theory). This is partially due to
historical reasons, but also to the following fundamental result.
Theorem 5.10 (Central Limit Theorem). Let {Xn }n∈N be a sequence of independent
and identically distributed random variables with E[Xi ] = 0 and Var[Xi ] = 1. Let Sn be the
sample average of the first n random variables, i.e.,
1
Sn = (X1 + X2 + · · · + Xn ).
n
√
Then, as n → ∞, nSn converges in distribution to a standard normal random variable,
that is Z x
1 1 2
√
lim F nSn (x) = √ e− 2 y dy, for all x ∈ R.
n→∞ 2π −∞
For example, Theorem 5.10 applied to√the random walk {Mn }n∈N implies that in the limit
n → +∞, the random variable Mn / n converges in distribution to a standard normal
random variable, which is consistent with√the result of Exercise 5.17 and the subsequent
remark that the stochastic process {M[nt] / n}t∈[0,T ] approximates a Brownian motion in the
limit n → ∞ (see Definition 5.18 below). The proof of the Central Limit Theorem can be
found in any textbook in probability theory (e.g., [5]).

106
Exercise 5.26. Prove the following. Let {Xn }n∈N be a sequence of independent and identi-
cally distributed random variables with E[Xi ] = µ and
√ Var[Xi ] = σ 2 . Let Sn be the sample
average of the first n random variables. Then σ −1 n(Sn − µ) converges in distribution to
a standard normal random variable. HINT: make a change of variables and apply Theo-
rem 5.10.

5.5.3 Brownian motion

Let Ω be a sample space and T > 0. Recall that a stochastic process is a one parameter
family {X(t)}t∈[0,T ] of random variables X(t) : Ω → R. We denote by X(t, ω) the value
attained by the random variable X(t) on the sample point ω ∈ Ω (i.e., X(t, ω) = X(t)(ω)).
Moreover, for each ω ∈ Ω fixed, the function t → X(t, ω) defined on [0, T ] is called a path
of the stochastic process. In the following we refer to the parameter t as the time variable,
since this is what it represents in the applications that we have in mind.
Definition 5.18. A Brownian motion, or Wiener process, is a stochastic process {W (t)}t∈[0,T ]
with the following properties:

1. For all 6 ω ∈ Ω, the paths are continuous (i.e., t → W (t, ω) is a continuous function
on [0, T ]) and W (0, ω) = 0;
2. For all 0 = t0 < t1 < · · · < tm = T , the increments
W (t1 ) = W (t1 ) − W (t0 ), W (t2 ) − W (t1 ), . . . , W (T ) − W (tm−1 )
are independent random variables and
E[W (ti+1 ) − W (ti )] = 0, Var[W (ti+1 ) − W (ti )] = ti+1 − ti , for all i = 0, . . . m − 1;

3. The increments are normally distributed, that is to say, for all 0 ≤ s < t ≤ T ,
Z
1 y2
P(W (t) − W (s) ∈ A) = p e− 2(t−s) dy,
2π(t − s) A
for all regular sets A ⊆ R.

It can be shown that Brownian motions exist, yet a formal construction is technically quite
difficult and beyond the purpose of these notes. One particular way to construct a Brownian
motion is suggested in Exercise 5.17, namely by running a properly rescaled symmetric
random walk for infinitely many steps. In fact it is useful to think of Brownian motions
as time-continuum generalizations of the symmetric random walk. Note to this regard that
the increments of a symmetric random walk also satisfy the independence property 2 in
Definition 5.18. The fact that in the continuum limit they are normally distributed follows
by the Central Limit Theorem.
6
More precisely, for all ω ∈ A ⊆ Ω with P(A) = 1.

107
Exercise 5.27 (•). Let {W (t)}t∈[0,T ] be a Brownian motion. Show that Cov[W (s), W (t)] =
min(s, t), for all s, t ∈ [0, T ]. (Compare this with Exercise 5.17.)

108
Chapter 6

Black-Scholes options pricing theory

This final chapter is concerned with the most famous of all models in options pricing theory,
namely the Black-Scholes model, which first appeared in the seminal paper [1]. The way we
introduce it in this chapter is very different from the original argument in [1]. Our strategy
is to derive the Black-Scholes model from the binomial options pricing model in the time-
continuum limit. More precisely we let the number of steps N in the binomial model goes to
infinity and, at the same time, we let the length h of each time step tends to zero, keeping
constant the time of maturity T = N h.

6.1 Black-Scholes markets

Can we use a Brownian motion to model the evolution in time of a stock price, i.e., can we
set S(t) = W (t)? The answer is of course no, because W (t) can also take negative values.
In fact, since W (t) is normally distributed, then it has a positive probability to take value
in any interval within the real line. However this deficiency is easily corrected.

Definition 6.1. Let {W (t)}t∈[0,T ] be a Brownian motion, α ∈ R and σ > 0. The positive
stochastic process {S(t)}t∈[0,T ] given by

S(t) = S(0)eαt+σW (t) (6.1)

is called a geometric Brownian motion.

As the notation used in (6.1) suggests, we shall use geometric Brownian motions to model
the dynamics of stock prices in the time-continuum case. More precisely, a Black-Scholes
market is a market that consists of a risky asset, say a stock, whose price at time t ∈
[0, T ] is given by the geometric Brownian motion (6.1), and a risk-free asset with constant
interest rate r; in particular, the value of the risk-free asset at time t is given by B(t) =
B(0)ert . Within this application, we call α the instantaneous mean of log-return, σ the

109
instantaneous volatility and σ 2 the instantaneous variance of the geometric Brownian
motion. To justify this terminology we now show that α and σ satisfy the analogues of (5.16)
in the time-continuum case, namely, for all t ∈ [0, T ] and ε > 0 such that t + ε ≤ T we have
1 1
α = E[log S(t + ε) − log S(t)], σ 2 = Var[log S(t + ε) − log S(t)]. (6.2)
ε ε
In fact, by (ii) of Theorem 5.6, and since W (t) ∈ N (0, t),
E[log S(t + ε) − log S(t)] = E[αε + σW (t + ε) − σW (t)]
= αε + σ(E[W (t + ε)] − E[W (t)]) = αε.
The proof of the identity for σ 2 is left as an exercise.
Exercise 6.1. Prove the second identity in (6.2).

In the following theorem we show that the historical variance (1.6) of a stock is an unbiased
estimator for the instantaneous variance of the geometric Brownian motion used to model
its price.
Theorem 6.1. Suppose that at time t = 0 it is assumed that the stock price is described by
a geometric Brownian motion in the interval [0, T ]:
S(t) = S(0)eαt+σW (t) , t ∈ [0, T ].
Given any arbitrary subinterval [t0 , t] ⊂ [0, T ] with length τ = t − t0 , define the random
variable n
1 X
στ2 (t) = (Ri − R̄)2 , (6.3)
h(n − 1) i=1
where t0 < t1 < t2 < · · · < tn = t is a partition of [t0 , t] with h = ti − ti−1 and Ri , R̄ are
given by (1.4), (1.5). Then
E[στ (t)2 ] = σ 2 .
In other words, σ 2 is the expected value of the τ -historical variance at any time t ∈ [0, T ].

Proof. In (6.3) we replace

1 S(t) 1 σ
R̄ = log = (α(t − t0 ) + σ(W (t) − W (t0 )) = αh + (W (t) − W (t0 )), and
n S(t0 ) n n
S(ti )
Ri = log = α(ti − ti−1 ) + σ(W (ti ) − W (ti−1 ) = αh + σ(W (ti ) − W (ti−1 ).
S(ti−1 )
So doing we obtain
n  2
2 σ2 X 1
στ (t) = W (ti ) − W (ti−1 ) − (W (t) − W (t0 ))
h(n − 1) i=1 n
" n #
σ2 X 1
= (W (ti ) − W (ti−1 ))2 − (W (t) − W (t0 ))2 .
h(n − 1) i=1 n

110
Taking the expectation and using that E[(W (t) − W (s))2 ] = Var[W (t) − W (s)] = t − s we
obtain
" n #
2
σ2
 
2 σ X 1 1
E[στ ] = (ti − ti−1 ) − (t − t0 ) = (t − t0 ) 1 − = σ2.
h(n − 1) i=1 n h(n − 1) n

Next we derive the density function of the geometric Brownian motion.

Theorem 6.2. The density of the random variable S(t) is given by
(log x − log S(0) − αt)2
 
H(x) 1
fS(t) (x) = √ exp − , (6.4)
2πσ 2 t x 2σ 2 t
where H(x) is the Heaviside function, i.e.,

1 if x > 0
H(x) = .
0 if x ≤ 0

Proof. The density of S(t) is given by

d
fS(t) (x) = FS(t) (x),
dx
where FS(t) is the distribution of S(t), i.e.,

FS(t) (x) = P(S(t) ≤ x).

Clearly, fS(t) (x) = FS(t) (x) = 0, for x ≤ 0. For x > 0 we use that
 
1 x
S(t) ≤ x if and only if W (t) ≤ log − αt := A(x).
σ S(0)
Thus Z A(x)
1 y2
P(S(t) ≤ x) = P(−∞ < W (t) ≤ A(x)) = √ e− 2t dy,
2πt −∞
where for the second equality we used that W (t) ∈ N (0, t). Hence
Z A(x) !
d 1 y2 1 − A(x)2 dA(x)
fS(t) (x) = √ e− 2t dy = √ e 2t ,
dx 2πt −∞ 2πt dx

for x > 0, that is

(log x − log S(0) − αt)2
 
1 1
fS(t) (x) = √ exp − , x > 0.
2πσ 2 t x 2σ 2 t
The proof is complete.

111
Exercise 6.2. Express P(a < S(t) < b) in terms of the standard normal distribution Φ(x).

Finally we study the relation between the discrete binomial stock price and the time-
continuous geometric Brownian motion model. Consider an interval of time [0, t] and a
partition

0 = t0 < t1 < · · · < tN = t, ti+1 − ti = h > 0, for all i = 0, . . . , N − 1.

We define a binomial stock price on this partition by setting, for u > d,


S(ti−1 )eu with probability p
S(ti ) = d .
S(ti−1 )e with probability 1 − p

Now consider N independent and identically distributed random variables X1 , . . . XN satis-

fying 
1 with probability p
Xi = .
−1 with probability 1 − p
We can rewrite the binomial model above as
    
u+d u−d
S(ti ) = S(ti−1 ) exp + Xi .
2 2

Let us assume for simplicity that p = 1/2 (see Exercise 6.3 below for the general case). Then
iterating the previous identity we obtain
     
u+d u−d
S(t) = S(0) exp N + MN ,
2 2

where MN = X1 + . . . XN is the symmetric random walk stochastic process. Substituting

N = ht , the previous expression becomes
h √ i
S(t) = S(0) exp αt + σ hMt/h , (6.5)
√
where α = (u + d)/(2h) and σ = (u − d)/(2 h) are the instantaneous mean of log-return
and the instantaneous volatility of the binomial stock price, see (2.19). As indicated in
Exercise 5.17, we have √
hMt/h ≈ W (t), for h small.
Therefore for small h we can approximate (6.5) by the geometric Brownian motion

S(t) ≈ S(0) exp [αt + σW (t)] .

The following simple application of the Central Limit Theorem turns the above argument
into a rigorous result.

112
Theorem 6.3. Consider the binomial stock price stochastic process {SN (t)}t∈[0,T ] given by
√ M
αt+σ t √N
SN (t) = S(0)e N ,

where MN is the symmetric random walk. Then

SN (t) → S(t), in distribution as N → ∞,

where S(t) = S(0)eαt+σW (t) is the geometric Brownian motion with instantaneous mean of
log-return α and instantaneous volatility σ.

Proof. We have, for all x > 0,

x
MN log S(0) − αt
FSN (t) (x) = P(SN (t) ≤ x) = P( √ ≤ √ ).
N σ t

By the Central Limit Theorem, the random variable M√ N converges in distribution to a

N √
random variable X ∈ N (0, 1) as N → ∞. Letting X = W (t)/ t we then have
x
W (t) log S(0) − αt
FSN (t) (x) → P( √ ≤ √ ) = P(S(t) ≤ x) = FS(t) (x), as N → ∞.
t σ t

Exercise 6.3. Show that the binomial stock price converges in distribution to the geometric
Brownian motion even for p 6= 1/2. Remark: it can be shown that the value of p ∈ (0, 1) only
affects the rate of convergence. In particular the fastest convergence is obtained for p = 1/2,
which is the reason why we always make this choice in the binomial model.

Exercise 6.4 (Matlab). Test numerically the convergence of the binomial stock to the
geometric Brownian motion. Show that the fastest convergence is obtained for p = 1/2.

6.2 Black-Scholes price of standard European deriva-

tives
The ultimate purpose of this section is to introduce and justify the definition of fair price
for standard European derivatives on Black-Scholes markets.
In Section 5.4 we showed that the binomial fair price at time t = 0 of the standard
European derivative with pay-off Y = g(S(N )) at maturity T = N can be written as
ΠY (0) = e−rN Eq [g(S(N ))], where Eq [·] denotes the expectation in the risk-neutral probabil-
ity measure; see (5.35). As a first step to derive the Black-Scholes price of the same European

113
derivative, we first rewrite ΠY (0) in terms of the distribution of S(N ). Using the definition
of probability measure in the N -coin toss probability space we have
X
ΠY (0) = e−rN Eq [g(S(N ))] = e−rN q NH (ω) (1 − q)NT (ω) g(S(N ))
ω∈ΩN
N  
−rN
X N k
=e q (1 − q)N −k g(sk ),
k=0
k

where s0 , . . . , sN are the possible values of S(N ), i.e.,

sk = S(0) exp(ku + (N − k)d), k = 0 . . . , N, (6.6)

and
er − ed
q= u , r ∈ (d, u) (⇔ q ∈ (0, 1)). (6.7)
e − ed
Note the important formula
   
1 sk
k= log − Nd , (6.8)
u−d S(0)

which is obtained by inverting (6.6). We can rewrite ΠY (0) as

N  k  N −k  
−rN
X q 1−q N k
ΠY (0) = e p (1 − p)N −k g(sk ).
k=0
p 1 − p k

Recalling that the distribution of S(N ) is given by Nk pk (1 − p)N −k (see Section 5.4), we

obtain
XN
ΠY (0) = e−rN ZN (sk )fS(N ) (sk )g(sk ), (6.9)
k=0

where
1 sk 1 sk
 k 
q 1−q
N −k   u−d
q [log( S(0) )−N d]  1 − q N − u−d [log( S(0) )−N d]
ZN (sk ) = = , (6.10)
p 1−p p 1−p

the second equality following by (6.8). From now on we assume p = 1/2 for simplicity.
Denoting sk = x, we rewrite (6.9) as
X
ΠY (0) = e−rN ZN (x)fS(N ) (x)g(x), (6.11)
x∈ImS(N )

where ImS(N ) is the image of S(N ), fS(N ) is the probability distribution of S(N ), and
1 x 1 x
ZN (x) = (2q) u−d [log( S(0) )−N d] (2(1 − q))N − u−d [log( S(0) )−N d] . (6.12)

114
We have also seen in Section 6.1 that, in a suitable limit, the binomial stock price converges
to the geometric Brownian motion, see Theorem 6.3. We shall obtain the Black-Scholes
formula in the same limit applied to (6.11). To this purpose, let T > 0 and consider the
partition of the interval [0, T ] given by

0 = t0 < t1 < · · · < tN = T, tj − tj−1 = h > 0, j ∈ I.

Let us define the binomial model S(0) = S0 > 0,

S(tj−1 )eu , with probability p = 1/2

(
S(tj ) = , j ∈ I, (6.13)
S(tj−1 )ed , with probability p = 1/2

where √ √
u = αh + σ h, d = αh − σ h. (6.14)
Note that
u−d u+d
σ= √ , α= ,
2 h 2h
hence, by Theorem 6.3, in the limit h → 0 the price of the stock follows the geometric
Brownian motion S(t) = S(0) exp(αt + σW (t)). Moreover B(tj ) = B0 ertj = B0 erhj . Hence
the pair (S(j)
e = S(tj ), B(j)
e = B(tj )) is equivalent to a binomial market with parameters
(u, d, rh) and p = 1/2. In particular the parameter q defined by (6.7) becomes
√
erh − eαh−σ h
q= √ √ (6.15)
eαh+σ h − eαh−σ h
and since √ √
αh − σ h < rh < αh + σ h
holds for h small, then we can assume that q ∈ (0, 1) and that the market is arbitrage-free.
Therefore the initial price of the European derivative with pay-off Y = g(S(Ne )) = g(S(T ))
is given by (6.11). Using N h = T , we rewrite (6.11) as
X X
ΠY (0) = e−rhN ZN (x)fS(N
e ) (x)g(x) = e
−rT
Qh (x)fS(T ) (x)g(x), (6.16)
x∈ImS(N
e ) x∈ImS(T )

where Qh (x) = ZT /h (x); by (6.12), (6.15) and the definitions of u, d, we have Qh (x) =
ηh (x)ξh (x), where
√ ! 1√ x
(log S(0) T
−αT )+ 2h
erh − eαh−σ h 2σ h

ηh (x) = 2 √ √ ,
eαh+σ h − eαh−σ h
√ !− 1√ x
(log S(0) T
−αT )+ 2h
eαh+σ h
− erh 2σ h

ξh (x) = 2 √ √ .
eαh+σ h − eαh−σ h

115
Theorem 6.4. The following holds:
2
(α+ σ2 −r)2 T 2
 
1 x
− − (log S(0) −αT ) α+ σ2 −r
lim Qh (x) = e 2σ 2 e σ2 := Q(x).
h→0

Exercise 6.5. Prove the theorem.

We are now in the position to justify the definition of the Black-Scholes price of the derivative.
In view of the analogies between finite and uncountable probability spaces pointed out in
Theorem 5.6, the natural generalization of (6.16) in the time-continuum case is
Z
−rT
ΠY (0) = e Q(x)fS(T ) (x)g(x) dx,
R

where fS(T ) is now the probability density of the random variable S(T ) = S(0)eαT +σW (T ) ,
i.e.,
x
!
H(x) 1 (log S(0) − αT )2
fS(T ) (x) = √ exp − , (6.17)
2πσ 2 T x 2σ 2 T
see (6.4). Explicitly, ΠY (0) is given by
2 2
e−rT
Z ∞
−
(α+ σ2 −r)2 T
− 1 x
(log S(0)
2
−αT )(α+ σ2 −r) −
(log S(0)
x −αT
) g(x)
ΠY (0) = √ e 2σ 2 e σ2 e 2σ 2 T dx.
2πσ 2 T 0 x

Now, the exponents in the exponential functions inside the integral form a perfect square:
Z ∞
e−rT
 i2
−αT )+ α+ σ2 −r T g(x)
2
h 
− 12 (log S(0)
x
ΠY (0) = √ e 2σ T dx
2πσ 2 T 0 x
Z ∞
e−rT
 i2
+ σ2 −r T g(x)
h  2
− 12 log S(0)
x
= √ e 2σ T dx.
2πσ 2 T 0 x

Note that the dependence on α is gone! Finally, the change of variable

  2  
1 x σ
y= √ log + −r T
σ T S(0) 2

gives
e−rT √ 
Z
σ2 y2

ΠY (0) = √ g S(0)e(r− 2 )T eσ T y e− 2 dy.
2π R

The Black-Scholes price ΠY (t0 ) at a generic time t0 ∈ [0, T ] is obtained by assuming that the
present time is t = t0 (instead of t = 0) and replacing the expiration date T with the time
left to maturity, i.e., τ = T − t0 . The latter is of course quite reasonable and can be justified
by the same argument applied to derive the formula for ΠY (0). We are led to introduce the
following definition.

116
Definition 6.2. Consider the European derivative with pay-off Y = g(S(T )) and time of
maturity T > 0. Let r be the (constant) interest rate of the money market and assume
that the price of the stock is given by the geometric Brownian motion S(t) = S(0)eαt+σW (t) ,
t ∈ [0, T ]. The Black-Scholes price ΠY (t) of the derivative at time t ∈ [0, T ] is defined as

ΠY (t) = v(t, S(t)), (6.18)

where
e−rτ
Z  √ 
σ2 y2
v(t, x) = √ g xe(r− 2 )τ eσ τ y e− 2 dy, τ = T − t, (6.19)
2π R
is called the Black-Scholes price function of the derivative.
Remark 6.1. Of course we are tacitly assuming that the pay-off function g is such that the
integral in the right hand side of (6.19) is well-defined. Note also that the Black-Scholes price
at time t is a deterministic function of S(t) and thus can be computed with the information
available at time t.
Remark 6.2. The fact that the Black-Scholes price is independent of the mean of log-
return α of the stock is consistent with the numerical observation made in Section 3.3 that
the binomial price of European options is weakly dependent on the parameter α. In the
time-continuum limit, this dependence disappears completely.

Note that the definition (6.18)-(6.19) of the Black-Scholes price can be written also in the
probabilistic form
σ2
ΠY (t) = e−rτ E[g(S(t)e(r− 2 )τ +σ(W (T )−W (t)) ]. (6.20)
√
In fact, since G = (W (T ) − W (t))/ T − t ∈ N (0, 1), then
σ2 σ2 √
E[g(S(t)e(r− 2
)τ +σ(W (T )−W (t))
] = E[g(S(t)e(r− 2 )τ +σ τ G )]
Z  √ 
1 σ2 y2
=√ g S(t)e(r− 2 )τ eσ τ y e− 2 dy,
2π R
where we used Theorem 5.6.

Hedging portfolio
A portfolio process {hS (t), hB (t)}t∈[0,T ] invested in the Black-Scholes market is said to be
hedging the European derivative with pay-off Y and maturity T > 0 if V (T ) = Y , where
V (t) = hS (t)S(t) + hB (t)B(t) is the value of the portfolio process at time t ∈ [0, T ]. The
portfolio process is said to be replicating the derivative if V (t) = ΠY (t), where ΠY (t) is
the Black-Scholes price of the derivative. It can be shown that the Black-Scholes ΠY (t)
coincides with the value at time t ∈ [0, T ] of any self-financing portfolio processes hedging
the derivative. Thus Definition 6.2 can be motivated by the same argument used to justify
the definition of binomial price of European derivatives, see Remark 3.1. However this

117
approach requires the use of stochastic calculus and it is therefore beyond the purpose of
these notes. Moreover it can be shown that the portfolio process {(hS (t), hB (t))}t∈[0,T ] given
by
1
hS (t) = ∆(t, S(t)), ∆(t, x) = ∂x v(t, x), hB (t) = (ΠY (t) − hS (t)S(t)) (6.21)
B(t)
is self-financing and hedges the derivative. This portfolio is also predictable, as the position
at time t depends only on the stock price S(t). For a heuristic derivation of (6.21), see
Exercise 6.6 below. In the next two sections we compute the Black-Scholes price and the
hedging portfolio process of some simple derivatives.
Exercise 6.6. The goal of this exercise is to justify the formula (6.21) for a self-financing
and hedging portfolio process of standard European derivatives priced by the Black-Scholes
formula. Recall that in the time discrete case we have
1 ΠuY (t) − ΠdY (t)
hS (t) = , (6.22)
S(t − 1) eu − ed
see Theorem 3.3. Now, as for the derivation of the Black-Scholes price given in this section,
consider a partition {t0 , . . . , tN } of the interval [0, T ] and let
√ √
t − 1 = tj−1 , t = tj = tj−1 + h, u = αh + σ h, d = αh − σ h,
see (6.13) and (6.14). Let ΠY (tj ) = v(tj , S(tj )), where v is the Black-Scholes price function.
Show that (6.22) converges to ∂x v(t, x) as h → 0. Prove that (6.21) replicates the derivative.

6.3 Black-Scholes price of European call and put op-

tions
In this section we focus the discussion on call/put options. We thereby assume that the
pay-off of the derivative is given by
Y = (S(T ) − K)+ , i.e., Y = g(S(T )), g(z) = (z − K)+ , for a call option,
Y = (K − S(T ))+ , i.e., Y = g(S(T )), g(z) = (K − z)+ , for a put option.
The function v given by (6.19) will be denoted by C, for a call option, and by P , for a put
option.
Theorem 6.5. The Black-Scholes price at time t of the European call option with strike
price K > 0 and maturity T > 0 is given by C(t, S(t)), where
C(t, x) = xΦ(d1 ) − Ke−rτ Φ(d2 ), (6.23a)
log Kx + r − 21 σ 2 τ



√
d2 = √ , d1 = d2 + σ τ , (6.23b)
σ τ

118
 Rx 1 2
and where Φ(x) = √12π −∞ e− 2 y dy is the standard normal distribution. The Black-Scholes
price of the corresponding put option is given by P (t, S(t)), where

P (t, x) = Φ(−d2 )Ke−rτ − Φ(−d1 )x. (6.24)

Moreover the put-call parity identity holds:

C(t, S(t)) − P (t, S(t)) = S(t) − Ke−rτ . (6.25)

Proof. We derive the Black-Scholes price of call options only, the argument for put options
being similar (see Exercise 6.7 below). We substitute g(z) = (z − K)+ into the right hand
side of (6.19) and obtain
e−rτ
Z  √ y2
1 2

C(t, x) = √ xe(r− 2 σ )τ eσ τ y − K e− 2 dy.
2π R +

1 2 )τ √
Now we use that xe(r− 2 σ eσ τy
> K if and only if y > −d2 . Hence
Z ∞ Z ∞
e−rτ

√ 2 2
(r− 21 σ 2 )τ σ τ y − y2 − y2
C(t, x) = √ xe e e −K e dy .
2π −d2 −d2
√ √ 2
Using − 12 y 2 + σ τ y = − 21 (y − σ τ )2 + σ2 τ and changing variable in the integrals we obtain
Z ∞ Z ∞
e−rτ
 
√ 2
rτ − 12 (y−σ τ )2 − y2
C(t, x) = √ xe e dy − K e dy
2π −d2 −d2
" Z d2 +σ√τ Z d2 #
e−rτ 1 2 y2
=√ xerτ e− 2 y dy − K e− 2 dy
2π −∞ −∞

= xΦ(d1 ) − Ke−rτ Φ(d2 ).

As to the put-call parity, we have

C(t, x) − P (t, x) = xΦ(d1 ) − Ke−rτ Φ(d2 ) − Φ(−d2 )Ke−rτ + xΦ(−d1 )

= x(Φ(d1 ) + Φ(−d1 )) − Ke−rτ (Φ(d2 ) + Φ(−d2 )).

As Φ(z) + Φ(−z) = 1, the claim follows.

Remark 6.3. Note that the Black-Scholes price of call and put options is strictly positive
for all t ∈ [0, T ), while the binomial price can also attain the (unrealistic) zero value prior
to maturity, see Section 3.3.
Exercise 6.7. Derive the Black-Scholes price P (t, S(t)) of European put options claimed in
Theorem 6.5.
Exercise 6.8 (Matlab). Write a Matlab function that computes the Black-Scholes price of
call/put options. Show numerically that the binomial price of call/put options converges to
their Black-Scholes price in the time continuum limit.

119
Exercise 6.9 (•). Prove that
lim C(t, x) = (x − Ke−rτ )+ , lim C(t, x) = x.
σ→0+ σ→∞

Compute also the following limits:

lim C(t, x), lim C(t, x), lim C(t, x), lim C(t, x).
K→0+ K→+∞ T →+∞ x→0+

Repeat all the above for put options.

Next we derive the hedging portfolio for call and put options.
Theorem 6.6. The following portfolio processes are self-financing hedging portfolios for
European call/put options on Black-Scholes markets:
Ke−rτ Φ(d2 )
hS (t) = Φ(d1 ), hB (t) = − for call options (6.26a)
B(t)
Ke−rτ Φ(−d2 )
hS (t) = −Φ(−d1 ), hB (t) = for put options. (6.26b)
B(t)

Proof. Recall that hS (t) = ∂x C(t, S(t)) for call options and hS (t) = ∂x P (t, S(t)) for put
options, see (6.21), while the number of shares of the risk-free asset in the hedging portfolio
is given by
hB (t) = (C(t, S(t)) − S(t)∂x C(t, S(t)))/B(t), for call options, (6.27a)
hB (t) = (P (t, S(t)) − S(t)∂x P (t, S(t)))/B(t), for put options. (6.27b)
Let us compute ∂x C:
∂x C = Φ(d1 ) + xΦ0 (d1 )∂x d1 − Ke−rτ Φ0 (d2 )∂x d2 .
1 2 √
As ∂x d1 = ∂x d2 = σ√1τ x , and Φ0 (x) = e− 2 x / 2π, we obtain
 
1 − 12 d21 K −rτ − 1 d22
∂x C = Φ(d1 ) + √ e − e e 2 .
σ 2πτ x
√
Replacing d1 = d2 + σ τ we obtain
1 2 
e− 2 d2

√ K −rτ
− 12 σ 2 τ −d2 σ τ
∂x C = Φ(d1 ) + √ e − e .
σ 2πτ x
Using the definition of d2 , the term within round brackets in the previous expression is easily
found to be zero, hence
∂x C = Φ(d1 ).
By the put-call parity we find also
∂x P = Φ(d1 ) − 1 = −Φ(−d1 ).
In both cases, the number of shares in the risk-free asset is computed using (6.27).

120
Remark 6.4. Note that ∂x C > 0, while ∂x P < 0. This agrees with the fact that call options
are bought to protect a short position on the underlying stock, while put options are bought
to protect a long position on the underlying stock.

The greeks

The Black-Scholes price of a call (or put) option derived in Theorem 6.5 depends on the price
of the underlying stock, the time to maturity, the strike price, as well as on the (constant)
market parameters r, σ (it does not depend on α). The partial derivatives of the price
function C with respect to these variables are called greeks. We collect the most important
ones (for call options) in the following theorem.

Theorem 6.7. The price function C of call options satisfies the following:

∆ := ∂x C = Φ(d1 ), (6.28)
φ(d1 )
Γ := ∂x2 C = √ , (6.29)
xσ τ
ρ := ∂r C = Kτ e−rτ Φ(d2 ), (6.30)
xφ(d1 )σ
Θ := ∂t C = − √ − rKe−rτ Φ(d2 ), (6.31)
2 τ
√
ν := ∂σ C = xφ(d1 ) τ (called “vega”), (6.32)
√ z2
where φ(z) = Φ0 (z) = ( 2π)−1 e− 2 . In particular:

• ∆ > 0, i.e., the price of a call is increasing on the price of the underlying stock;

• Γ > 0, i.e., the price of a call is convex on the price of the underlying stock;

• ρ > 0, i.e., the price of the call is increasing on the interest rate of the risk-free asset;

• Θ < 0, i.e., the price of the call is decreasing in time;

• ν > 0, i.e., the price of the call is increasing on the volatility of the stock.

Exercise 6.10. Use the put-call parity to derive the greeks of put options.

The greeks measure the sensitivity of option prices with respect to the market conditions.
This information can be used to draw some important conclusions. For instance, let us
comment on the fact that vega is positive. It implies that the wish of an investor with a long
position on a call option is that the volatility of the underlying stock increased. As usual,
since this might not happen, the buyer of the call may incur in a loss if the stock volatility
decreases (since the call option will loose value). This exposure to volatility can be secured
by adding volatility swaps into the portfolio.

121
6.4 The Black-Scholes price of other standard Euro-
pean derivatives
In this section we present a few more applications of the Black-Scholes formula (6.18)-(6.19).

Binary call option

A binary (or digital) call option with strike K and maturity T pays-off the buyer if and
only if S(T ) > K. If the pay-off is a fixed amount of cash L, then the binary call option is
said to be “cash-settled”, while if the pay-off is the stock itself then the option is said to be
“physically settled”. Let us compute the Black-Scholes price of the cash-settled binary call
option. The pay-off is Y = LH(S(T ) − K), where H(z) is the Heaviside function. Replacing
g(z) = LH(z − K) into (6.19) we obtain
e−rτ
Z √
σ2 y2
v(t, x) = √ L H(xe(r− 2 )τ eσ τ y − K)e− 2 dy
2π
ZR∞
e−rτ y2
=√ L e− 2 dy,
2π −d2

1 σ2
where we recall that d2 = √
σ τ
[log(x/K) + (r − 2
)τ ]. With the change of variable y → −y
we obtain
ΠY (t) = e−rτ LΦ(d2 ).
Exercise 6.11 (?). Compute the Black-Scholes price of the physically-settled binary call.
Exercise 6.12. Compute the hedging portfolio of physically/cash-settled binary options.

Butterfly options strategy

Given K, ∆K > 0, a butterfly option strategy is a constant portfolio which is

• Long 1 call with strike price K − ∆K and maturity T ;

• Short 2 calls with strike price K and maturity T ;
• Long 1 call with strike price K + ∆K and maturity T .

The portfolio value at expire is

V (T ) = (S(T ) − K + ∆K)+ − 2(S(T ) − K)+ + (S(T ) − K − ∆K)+ .
We may think of V (T ) as the pay-off Y of a standard European derivative, where Y =
g(S(T )) and
g(x) = (x − K + ∆K)+ − 2(x − K)+ + (x − K − ∆K)+ = g1 (x) + g2 (x) + g3 (x).

122
Note that the pay-off is positive, i.e., the option expires in the money, if and only if S(T ) ∈
(K − ∆K, K + ∆K). Hence a butterfly option strategy is lucrative when the price of the
stock at time T lies within an interval centered in K. As the integral in (6.19) is linear in
the pay-off function g, the value of a butterfly portfolio at time t is V (t) = v(t, S(t)), where
e−rτ
Z √ 
σ2 y2

v(t, x) = v1 (t, x) + v2 (t, x) + v3 (t, x), vi (t, x) = √ gi xe(r− 2 )τ eσ τ y e− 2 dy.
2π R
As g1 , g2 , g3 are pay-off functions of call options, we obtain

V (t) = ΠY (t) = C(t, S(t), K − ∆K, T ) − 2C(t, S(t), K, T ) + C(t, S(t), K + ∆K, T ),

where we denoted by C(t, S(t), K, T ) the Black-Scholes price of the call with strike K and
maturity T . This notation, and the analogous one P (t, S(t), K, T ) for put options, will be
used in the rest of this section.
Remark 6.5. We obtain the same result by applying the dominance principle to the butterfly
portfolio.

Chooser option
This is an example of “second derivative”, i.e., of a financial derivative whose underlying is
another derivative. More precisely, given T2 > T1 , a chooser option with maturity T1 is
a contract which gives to the buyer the right to choose at time T1 whether the derivative
becomes a call or a put option with strike K and maturity T2 . Hence the pay-off at time T1
for this derivative is

Y = max(C(t1 , S(T1 ), K, T2 ), P (t1 , S(T1 ), K, T2 )).

Using the identity max(a, b) = a + max(0, b − a) we obtain

Y = C(t1 , S(T1 ), K, T2 ) + max(0, P (t1 , S(T1 ), K, T2 ) − C(t1 , S(T1 ), K, T2 )).

By the put-call parity,

Y = C(t1 , S(T1 ), K, T2 ) + max(0, Ke−r(T2 −T1 ) − S(T1 )) = Z + U. (6.33)

Hence ΠY (t) = ΠZ (t) + ΠU (t). Since U is the pay-off of a put option with strike Ke−r(T2 −T1 )
expiring at time T1 then

ΠU (t) = P (t, S(t), Ke−r(T2 −T1 ) , T1 ). (6.34)

To find ΠZ (t) we need the following theorem:

Theorem 6.8 (∗). Consider the standard European derivative with pay-off Y = g(S(T )) at
maturity T and the derivative with pay-off Z = ΠY (t∗ ) at maturity t∗ < T . Then ΠZ (t) =
ΠY (t), t ∈ [0, t∗ ].

123
Proof. We have
√
Z  1 2 dy
σ2

−r(T −t∗ )
Z = ΠY (t∗ ) = e g S(t∗ )e(r− 2 )(T −t∗ )+σ T −t∗ y e− 2 y √ := h(S(t∗ )).
R 2π
This shows that the derivative with pay-off Z and maturity t∗ is a standard European
derivative with pay-off function h. Hence, applying again Black-Scholes’ formula, we obtain
√
Z   1 2 dz
σ2
−r(t∗ −t)
ΠZ (t) = e h S(t)e(r− 2 )(t∗ −t)+σ t∗ −tz e− 2 z √
R 2π
Z "Z  √ √
σ2 σ2

= e−r(t∗ −t) e−r(T −t∗ ) g S(t)e(r− 2 )(t∗ −t)+σ t∗ −tz e(r− 2 )(T −t∗ )+σ T −t∗ y
R R
#
1 2 1 2
dy dz
× exp − y − z √ √
2 2 2π 2π
√ √
Z Z   1 2 2 d(y, z)
σ2
= e−r(T −t) g S(t)e(r− 2 )(T −t)+σ( t∗ −tz+ T −t∗ y) e− 2 (y +z )
R R 2π
 σ2 √ √ 
= e−r(T −t) E[g xe(r− 2 )(T −t)+σ( t∗ −tZ+ T −t∗ Y ) ]x=S(t) ,

1 2 1 2
−2y −2z
where X, Y ∈ N (0, 1) are independent
√ and so their joint√ distribution is given by e /2π,
see Theorem 5.7. Let G1 := t∗ − t Z and G2 := T − t∗ Z. As G1 ∈ N (0, t∗ − t) and
G2 ∈ N (0, T − t∗ ), Theorem 5.9 gives G1 + G2 ∈ N (0, T − t), hence the above calculation
leads to
√
Z  2  1 2 dx
−r(T −t) (r− σ2 )(T −t)+σ T −tx
ΠZ (t) = e g S(t)e e− 2 x √ ,
R 2π
which is the claim.

Applying Theorem 6.8 with Z = C(t1 , S(T1 ), K, T2 ) = Π(S(T2 )−K)+ (T1 ) we obtain

ΠZ (t) = C(t, S(t), K, T2 ). (6.35)

Replacing (6.34) and (6.35) into (6.33) we finally obtain, for the Black-Scholes price of the
chooser option,

ΠY (t) = C(t, S(t), K, T2 ) + P (t, S(t), Ke−r(T2 −T1 ) , T1 ).

Exercise 6.13 (•). Consider the European derivative with maturity T and pay-off Y given
by
Y = k + S(T ) log S(T ),
where k > 0 is a constant. Find the Black-Scholes price of the derivative at time t < T and
the self-financing hedging portfolio. Find the probability that the derivative expires in the
money.

124
Exercise 6.14 (•). Consider the European derivative with pay-off Y = S(T )(S(T ) − K)
and time of maturity T , where K > 0 is a constant. Compute the Black-Scholes price ΠY (t)
of this derivative and the self-financing hedging portfolio. Finally, assume S(0) = K and
compute the expected relative return of a constant portfolio with 1 share of this derivative.
Exercise 6.15 (•). Let K > 0. A European style derivative on a stock with maturity T > 0
gives to its owner the right to choose between selling the stock for the price K at time T or
paying the amount K at time T . Draw the pay-off function of the derivative. Compute the
Black-Scholes price of the derivative. Show that there exists a value K∗ > 0 of K such that
the Black-Scholes price of the derivative is zero. What is the financial interpretation of K∗ ?.
Exercise 6.16 (?). Compute the Black-Scholes price ΠY (0) at time t = 0 of the European
derivative with pay-off Y = max(S(T ), B(T )), where B(t) is the price of the risk-free asset,
S(t) is the price of the underlying stock and T is the time of maturity of the derivative.
Derive the low volatility limit (σ → 0+ ) and the high volatility limit (σ → +∞) of ΠY (0).

6.5 Implied volatility

Let C(t, S(t), K, T ) denote the Black-Scholes price of the European call with strike price K,
maturity time T on a stock with price S(t) at time t. In fact, assuming that the underlying
stock pays no dividend before time T , this is also the Black-Scholes price of the corresponding
American call, which is the one most often traded in the market. Recall that in the derivation
of the Black-Scholes price it is assumed that the price of the stock follows the geometric
Brownian motion
S(t) = S(0)eαt+σW (t) ,
where {W (t)}t∈[0,T ] is a Brownian motion stochastic process, σ > 0 is the instantaneous
volatility and α ∈ R is the instantaneous mean of log-return. The function C(t, x, K, T ) is
given by
C(t, x, K, T ) = xΦ(d1 ) − Ke−rτ Φ(d2 ),
where r > 0 is the (constant) interest rate of the money market, τ = T − t is the time left
to the expiration of the call,
σ2
log Kx + (r − 2
)τ √
d2 = √ , d1 = d2 + σ τ
σ τ
and Φ denotes the standard normal distribution,
Z z
y2 dy
Φ(z) = e− 2 √ .
−∞ 2π
Remarkably, C(t, S(t), K, T ) does not depend on the mean of log-return α of the stock price.
However it depends on the parameters (σ, r) and since here we are particularly interested in
the dependence on the volatility, we re-denote the Black-Scholes price of the call as
C(t, S(t), K, T, σ).

125
 10

cHt, SHtL, K, T, ΣL
6

0
0 5 10 15 20

Σ
Figure 6.1: We fix S(t) = 10, K = 12, r = 0.01, τ = 1/12 and depict the Black-Scholes price
of the call as a function of the volatility. Note that in practice only the very left part of this
picture is of interest, because typically 0 < σ < 1.

Moreover, as shown in Theorem 6.7,

∂C x 1√
d2
= vega = √ e− 2 τ > 0.
∂σ 2π
Hence the Black-Scholes price of the option is an increasing function of the volatility. Fur-
thermore,
lim C(t, S(t), K, T, σ) = (S(t) − Ke−rτ )+ , lim C(t, S(t), K, T, σ) = S(t),
σ→0+ σ→+∞

see Exercise 6.9. Therefore the function C(t, S(t), K, T, ·) is a one-to-one map from (0, ∞)
into the interval I = ((S(t) − Ke−rτ )+ , S(t)), see Figure 6.1. Now suppose that at some
given fixed time t the real market price of the call is C(t).
e Clearly, the option is always
cheaper than the stock (otherwise we would buy directly the stock, and not the option) and
e > max(0, S(t)−Ke−rτ ). The latter is always true if S(t) < Ke−rτ
typically we also have C(t)
(the market price of options is positive), while if S(t) > Ke−rτ this follows by the fact that
S(t) − Ke−rτ ≈ S(t) − K and real calls are always more expansive than their intrinsic value.
e ∈ I.
This being said, we can safely assume that C(t)
Thus given the value of C(t)
e there exists a unique value of σ, which we denote by σimp , such
that
C(t, S(t), K, T, σimp ) = C(t).
e
σimp is called the implied volatility of the option. The implied volatility must be computed
numerically (for instance using Newton’s method), since there is no close formula for it.
Moreover it is usually computed using “nearly at the money” calls.

126
The implied volatility of an option (in this example of a call option) is a very important
parameter and it is often quoted together with the price of the option. If the market followed
exactly the assumptions in the Black-Scholes theory, then the implied volatility would be
constant (independent of time) and it would be the same for all call options on the same
stock with the same strike and maturity. However for real market options this turns out to
be false, i.e., the implied volatility depends on time, K and T . In this respect, σimp may
be viewed as a quantitative measure of how real markets deviate from ideal Black-Scholes
markets. The implied volatility may also be viewed as the market consensus on the expected
future value of the volatility of the underlying stock. That is to say, by pricing the option
at the price C(t)
e = C(t, S(t), K, T, σimp ), the market participants are telling us that they
believe that the volatility of the stock in the future will be σimp . (This of course provided
one assumes the real markets are fair.)

Volatility smile
As mentioned before, the implied volatility in real markets depends on the parameters K, T .
Here we are particularly interested in the dependence on the strike price, hence we re-denote
the implied volatility as σimp (K). If the market behaved exactly as in the Black-Scholes
theory, then σimp (K) = σ for all values of K, hence the graph of the function K → σimp (K)
would be a straight horizontal line. Given that real markets do not satisfy exactly the
assumptions in the Black-Scholes theory, what can we say about the graph of the function
K → σimp (K)? Remarkably, it has been found that there exists recurrent convex shapes for
the graph of this function, which are known as volatility smile and volatility skew, see
Figures 6.2-6.3. The minimum of the volatility smile is reached at the strike price K ≈ S(t),
i.e., when the call is at the money. This behavior indicates that the more the call is far
from being at the money, the more it will be overpriced. Volatility smiles and skews have
been found in the market especially after the crash in 1987 (Black Monday), indicating that
this event led investors to be more cautious when trading on options that are in or out
of the money. Devise mathematical models of stochastic volatility and asset prices able to
reproduce volatility curves is an active research topic in mathematical finance.

Exercise 6.17 (?). What can we infer about the investors behavior from the volatility skew?

6.6 Standard European derivatives on a dividend-paying

stock
In this section we consider Black-Scholes markets with a dividend-paying stock. This means
that at some time t0 ∈ (0, T ) the price of the stock decreases of a fraction a ∈ (0, 1) of its price

127
 Σimp
50

40 æ

æ
æ
æ
æ
30
æ æ
æ æ
æ æ
æ æ
æ æ æ
æ æ
20

120 125 130 135 K

Figure 6.2: Volatility Smile of a call option on Apple expiring May 15th , 2015. The data
were taken on May 12th , when the Apple stock quoted 126.34 dollars.

Σimp
10

Reverse Forward
8
Skew Skew
6

S-K
-4 -2 0 2 4

Figure 6.3: Volatility skews (not from real data!)

immediately before t0 , the difference being deposited in the account of the shareholders1 .
Letting S(t−
0 ) = limt→t−
0
S(t), we then have

S(t0 ) = S(t− − −
0 ) − aS(t0 ) = (1 − a)S(t0 ). (6.36)
1
The dividend is expressed in percentage of the price of the stock. For instance, a = 0.03 means that the
dividend paid is 3%.

128
We assume that on each of the intervals [0, t0 ), [t0 , T ], the stock price follows a geometric
Brownian motion, namely,

S(s) = S(t)eα(s−t)+σ(W (s)−W (t)) , t ∈ [0, t0 ), s ∈ [t, t0 ) (6.37)

S(s) = S(u)eα(s−u)+σ(W (s)−W (u)) , u ∈ [t0 , T ], s ∈ [u, T ]. (6.38)

Theorem 6.9 (∗). Consider the standard European derivative with pay-off Y = g(S(T ))
(a,t )
and maturity T . Let ΠY 0 (t) be the Black-Scholes price of the derivative at time t ∈ [0, T ]
assuming that the underlying stock pays the dividend aS(t−0 ) at time t0 ∈ (0, T ). Then

(a,t0 ) v(t, (1 − a)S(t)), for t < t0 ,
ΠY (t) =
v(t, S(t)), for t ≥ t0 ,

where v(t, x) is the pricing function in the absence of dividends, which is given by (6.19).

S(T )
Proof. Using S(t)
= eατ +σ(W (T )−W (t)) , we can rewrite (6.20) in the form
σ2
ΠY (t) = e−rτ E[g(S(T )e(r− 2
−α)τ
)]. (6.39)

We want to express S(T ) in this formula in terms of S(t) in the cases 0 ≤ t < t0 and
t0 ≤ t ≤ T . Taking the limit s → t−
0 in (6.37) and using the continuity of the paths of the
Brownian motion we find

S(t−
0 ) = S(t)e
α(t0 −t)+σ(W (t0 )−W (t))
, t ∈ [0, t0 ).

Replacing in (6.36) we obtain

S(t0 ) = (1 − a)S(t)eα(t0 −t)+σ(W (t0 )−W (t)) , t ∈ [0, t0 ).

Hence, letting (s, u) = (T, t0 ) and (s, u) = (T, t) into (6.38), we find

(1 − a)S(t)eατ +σ(W (T )−W (t)) for t ∈ [0, t0 ),
S(T ) = (6.40)
S(t)eατ +σ(W (T )−W (t)) for t ∈ [t0 , T ].

By the√definition of Black-Scholes price in the form (6.39) and denoting G = (W (T ) −

W (t))/ τ , we obtain
σ2 √
(a,t0 )
ΠY (t) = e−rτ E[g((1 − a)S(t)e(r− 2
)τ +σ τ G
], for t ∈ [0, t0 ),
σ2 √
(a,t0 )
ΠY (t) = e−rτ E[g(S(t)e(r− 2
)τ +σ τ G
], for t ∈ [t0 , T ].
As G ∈ N (0, 1), the result follows.

We conclude that for t ≥ t0 , i.e., after the dividend has been paid, the Black-Scholes price
function of the derivative is again given by (6.19), while for t < t0 it is obtained by replacing
x with (1 − a)x in (6.19). To see the effect of this change, suppose that the derivative is a

129
call option; let C(t, x) be the Black-Scholes price function in the absence of dividends and
Ca (t, x) be the price function in the case that a dividend is paid at time t0 . Then, according
to Theorem 6.9, 
C(t, (1 − a)x), for t < t0 ,
Ca (t, x) =
C(t, x), for t ≥ t0 .
Since ∂x C > 0 (see Theorem 6.7), it follows that Ca (t, x) < C(t, x), for t < t0 , that is to say,
the payment of a dividend makes the call option on the stock less valuable (i.e., cheaper)
than in the absence of dividends until the dividend is paid.
Exercise 6.18 (?). Give an intuitive explanation for the property just proved for call options
on a dividend paying stock.
Exercise 6.19 (•). A standard European derivative pays the amount Y = (S(T ) − S(0))+
at time of maturity T . Find the Black-Scholes price ΠY (0) of this derivative at time t =
0 assuming that the underlying stock pays the dividend (1 − e−rT )S( T2 −) at time t = T2 .
Compute the probability of positive return for a constant portfolio which is short 1 share of
the derivative and short S(0)e−rT shares of the risk-free asset (assume B(0) = 1).
Exercise 6.20. Derive the Black-Scholes price of the derivative with pay-off Y = g(S(T )),
assuming that the underlying pays a dividend at each time t1 < t2 < · · · < tM ∈ [0, T ].
Denote by ai the dividend paid at time ti , i = 1, . . . , M .

6.7 Optimal exercise time of American calls on dividend-

paying stocks
The Black-Scholes pricing theory for American derivatives is quite complicated and beyond
the purpose of these notes. In this section we prove a property of American calls on a
dividend-paying stock, which depends only on the following facts:

(i) An American derivative is at least as valuable as its European counterpart and its
intrinsic value;

(ii) If it is not optimal to exercise the American derivative for t ∈ [t0 , T ], where t0 ∈ [0, T ],
then the fair value of the American derivative and of its European counterpart are the
same for t ∈ [t0 , T ].

These facts can be proved rigorously and hold regardless of whether the underlying stock
pays or not a dividend. They are of course quite intuitive properties (see Chapter 1) and will
be used without further comment in this section. We also assume throughout this section
that the interest rate r of the money market is non-negative.
Let C
ba (t, S(t), K, T ) denote the Black-Scholes price at time t of the American call with strike
K and maturity T assuming that the underlying stock pays the dividend aS(t− 0 ) at time t0 ∈

130
(0, T ). We denote by Ca (t, S(t), K, T ) the Black-Scholes price of the corresponding European
call. We omit the subscript a to denote prices in the absence of dividends. Moreover replacing
the letter C with the letter P gives the price of the corresponding put option. We say that
it is optimal to exercise the American call at time t if its Black-Scholes price at this time
ba (t, S(t), K, T ) = (S(t) − K)+ .
equals the intrinsic value of the call, i.e., C
Theorem 6.10 (∗). Consider the American call with strike K and expiration date T and
assume that the underlying stock pays the dividend aS(t−
0 ) at the time t0 ∈ (0, T ). Then

ba (t, S(t), K, T ) > (S(t) − K)+ ,

C for t ∈ [t0 , T ),

i.e., it is not optimal to exercise the American call prior to maturity after the dividend is
paid. Moreover, there exists δ > 0 such that, if
δ
S(t−
0 ) > max( , K),
1−a
then the equality
ba (t−
C − −
0 , S(t0 ), K, T ) = (S(t0 ) − K)+

holds, and so it is optimal to exercise the American call “just before” the dividend is to be
paid.

Proof. For the first claim we can assume (S(t) − K)+ = S(t) − K, otherwise the American
call is out of the money and so it is clearly not optimal to exercise. By Theorem 6.9 we have

Ca (t, S(t), K, T ) = C(t, S(t), K, T ), Pa (t, S(t), K, T ) = P (t, S(t), K, T ), for t ≥ t0 .

Hence, by Theorem 6.5, the put-call parity holds after the dividend is paid:

Ca (t, S(t), K, T ) = Pa (t, S(t), K, T ) + S(t) − Ke−r(T −t) , t ≥ t0 .

Thus, for t ∈ [t0 , T ),

ba (t, S(t), K, T ) ≥ Ca (t, S(t), K, T ) > S(t) − K = (S(t) − K)+ ,
C

where we used that P (t, S(t), K, T ) > 0 and r ≥ 0. This proves the first part of the theorem,
i.e., the fact that it is not optimal to exercise the American call prior to expiration after the
dividend has been paid. In particular

C
ba (t, S(t), K, T ) = Ca (t, S(t), K, T ), for t ≥ t0 . (6.41)

Next we show that it is optimal to exercise the American call “just before” the dividend is
ba (t−
paid, i.e., C − −
0 , S(t0 ), K, T ) = (S(t0 ) − K)+ , provided the price of the stock is sufficiently
high. Of course it must be S(t− b − − −
0 ) > K. Assume first that Ca (t0 , S(t0 ), K, T ) > S(t0 ) − K;
then, owing to (6.41), C ba (t− − − −
0 , S(t0 ), K, T ) = Ca (t0 , S(t0 ), K, T ) (buying the American call
just before the dividend is paid is not better than buying the European call, since it is

131
never optimal to exercise the derivative prior to expiration). By Theorem (6.9) we have
Ca (t− − − − −
0 , S(t0 ), K, T ) = C(t0 , (1 − a)S(t0 ), K, T ) = C(t0 , (1 − a)S(t0 ), K, T ), where for the
latter equality we used the continuity in time of the Black-Scholes price function in the
absence of dividends. Since (1 − a)S(t− 0 ) = S(t0 ), then

ba (t− , S(t− ), K, T ) > S(t− ) − K ⇒ C

C ba (t− , S(t− ), K, T ) = C(t0 , S(t0 ), K, T ).
0 0 0 0 0

Hence
ba (t− , S(t− ), K, T ) > S(t− )−K ⇒ C(t0 , S(t0 ), K, T ) > S(t− )−K = S(t0 )+(1−a)S(t− )−K.
C 0 0 0 0 0

Therefore, taking the contrapositive statement,

C(t0 , S(t0 ), K, T ) ≤ S(t0 ) + (1 − a)S(t− b − − −

0 ) − K ⇒ Ca (t0 , S(t0 ), K, T ) = S(t0 ) − K. (6.42)

Next we remark that the function x → C(t, x, K, T ) − x is decreasing (since ∆ = ∂x C =

Φ(d1 ) < 1, see Theorem 6.7), and

lim C(t, x, K, T ) − x = 0,
x→0+

lim C(t, x, K, T ) − x = lim P (t, x, K, T ) − Ke−r(T −t) = −Ke−r(T −t) ,

x→+∞ x→+∞

see Exercise 6.9. Thus if (1 − a)S(t− 0 ) − K > −Ke

−r(T −t)
, i.e., S(t− −1
0 ) > (1 − a) K(1 −
e−r(T −t) , there exists ω such that if S(t0 ) > ω, i.e., S(t−0 ) > ω/(1 − a), then the inequality
−
C(t0 , S(t0 ), K, T ) ≤ S(t0 ) + (1 − a)S(t0 ) − K holds. It follows by (6.42) that for such values
of S(t− −
0 ) it is optimal to exercise the call “at time t0 ”. Letting δ = max(ω, K(1 − e
−r(T −t)
)
concludes the proof of the theorem.

Exercise 6.21. Prove that it is not optimal to exercise the American call at time t ∈ [0, t0 )
if S(t) < Ka (1 − e−r(T −t) ).

132
Appendix A

The Markowitz portfolio theory

Consider a constant portfolio position in some time interval [0, T ]. We assume that the
portfolio is invested in n risky assets U1 , . . . , Un with prices ΠUi (t), i = 1, . . . , N and a risk-
free asset Un+1 with value B(t) = B0 ert , where r > 0 is the instantaneous interest rate. For
notational convenience we let ΠUn+1 (t) = B(t), so that ΠUi (t), i = 1, . . . , n + 1 now denotes
the price of every asset in the portfolio. The relative return of the asset i is defined by
ΠUi (T ) − ΠUi (0)
Ri = , i = 1, . . . , n + 1.
ΠUi (0)
For i = 1, . . . , n, Ri is a random variable, while for i = n + 1 it is the deterministic constant:
B0 erT − B0
Rn+1 = = erT −1 := ρ > 0.
B0
As usual, the price of the assets at time t = 0 are supposed to be known.
The problem under study in this chapter is the following: Given an investor with initial
capital K > 0, what is the best way to distribute this wealth among the n + 1 assets in order
to maximize the expected return and, at the same time, minimize the expected risk? To
solve this problem we need first to introduce some notation. Let ai ∈ R denote the number
of shares of the asset Ui in the portfolio. The initial value of the portfolio is
n+1
X
V (0) = ai ΠUi (0) = K.
i=1

Letting
ai ΠUi (0)
πi = , i = 1, . . . , n + 1, (A.1)
K
we get
n+1
X
πi = 1. (A.2)
i=1

133
The value of the portfolio at time t = T is
n+1
X n+1
X
Ui
V (T ) = ai Π (T ) = ai ΠUi (0)(1 + Ri ).
i=1 i=1

The relative return of the portfolio in the interval [0, T ] is given by

V (T ) − V (0) V (T )
R= = − 1.
V (0) K

After simple calculations we obtain

n+1
X
R= πi Ri , (A.3)
i=1

where we used (A.1)-(A.2). The relative return of the portfolio is a random variable and
taking its expectation we obtain
n+1
X n
X
E[R] = πi E[Ri ] = πi E[Ri ] + πn+1 E[Rn+1 ].
i=1 i=1

Using (A.2) and that E[Rn+1 ] = Rn+1 = ρ we obtain

n n
!
X X
E[R] = π i µi + 1− πi ρ, (A.4)
i=1 i=1

where we set
µi = E[Ri ], for i = 1, . . . , n.
Hence the portfolio should be chosen to maximize (A.4) and at the same time to minimize
the portfolio risk. We measure the latter with the variance Var[R] of the relative return. To
compute Var[R] we first observe that
n+1
X n+1
X n
X
R − E[R] = πi R i − πi E[Ri ] = πi (Ri − E[Ri ]),
i=1 i=1 i=1

hence
n
!2 n
! n
!
X X X
(R − E[R])2 = πi (Ri − E[Ri ]) = πi (Ri − E[Ri ]) πj (Rj − E[Rj ])
i=1 i=1 j=1
n
X
= πi πj (Ri − µi )(Rj − µj ).
i,j=1

134
Letting C = (cij )i,j=1...n , cij = Cov(Ri , Rj ), be the covariance matrix of the assets returns
and π = (π1 π2 . . . πn )T , we obtain
n
X
2
Var[R] = E[(R − E[R]) ] = πi πj cij = π T Cπ. (A.5)
i,j=1

Our purpose is then to find a portfolio which maximizes (A.4) and minimizes (A.5). To this
end we shall need the following simple result.

Theorem A.1. Let X1 , . . . , Xn : Ω → R be random variables and cij = Cov(Xi , Xj ). The

covariance matrix C = (cij )i,j=1...,n is symmetric and positive-semidefinite, i.e., C = C T and
ξ T Cξ ≥ 0, for all ξ ∈ Rn .

Exercise A.1. Prove the theorem.

Definition A.1. Given θ > 0, let

gθ (π1 , . . . , πn ) = E[R] − θVar[R]. (A.6)

A portfolio (a1 , . . . , an , an+1 ) is called a Markowitz portfolio if the function gθ has a max-
imum at π b = (bπ1 , . . . , π
bn ), where

a1 S1 (0) an Sn (0)
π
b1 = ,...,π
bn = .
K K

The parameter θ > 0 is called the risk aversion of the investor. We shall see later that
the higher is θ the more portfolio value is invested in the risk-free asset. Hence a large θ
corresponds to a cautious investor, who favors “small risks” to “higher returns”.

Theorem A.2. Assume that the covariance matrix of the assets returns is positive-definite,
i.e., ξ T Cξ > 0, for all ξ 6= 0. Let Ω be the vector
1
Ω= (µ1 − ρ µ2 − ρ . . . µn − ρ)T .
2θ
b = C −1 Ω.
Then the function (A.6) has a unique maximum, which is attained at π

Proof. Using (A.4) and (A.5) the function gθ is

n n
! n
X X X
gθ (π) = µ i πi + 1− πi ρ−θ πi πj cij .
i=1 i=1 i,j=1

Hence n
∂gθ X
= µk − ρ − 2θ ckj πj .
∂πk j=1

135
 b of ∇π gθ (b
Thus the stationary points, i.e., the solutions π π ) = 0, satisfy
n
X 1
ckj π
bj = (µk − ρ), i.e., Cb
π = Ω.
i=1
2θ

As C is assumed to positive definite, then it is invertible and thus Cb

π = Ω has the unique
b = C −1 Ω. To show that this stationary point is a maximum of gθ we compute
solution π
∂ 2g
= −2θckl .
∂πk ∂πl
Since C is positive definite, then the Hessian ∇2π gθ is negative definite and therefore the
point π
b is a maximum. This concludes the proof of the theorem.
Remark A.1. Note that π bi → 0, and so also ai → 0, as θ → +∞, for all i = 1, . . . , n. Hence
in this limit the entire capital K is invested in the risk-free asset: πn+1 → 1, as θ → +∞,
which is equivalent to an+1 → K/ΠUn+1 (0) = KB0 , as θ → +∞. Conversely, the smaller is
θ, the larger is the exposure of the investor on the risky assets.

As a special case, assume that ΠU1 (T ) , . . . , ΠUn (T ) are independent. Then R1 (T ), . . . , Rn (T )

are also independent and therefore Cov(Ri , Rj ) = 0, for i 6= j. Hence

C = diag(Var[R1 ], Var[R2 ], . . . , Var[Rn ]),

where diag(z1 , . . . , zn ) is the diagonal n × n matrix A with diagonal elements aii = zi ,

i = 1, . . . n. Therefore in this case the maximum of gθ is attained at
E[Ri ] − ρ
π
bi = . (A.7)
2θVar[Ri ]

Example
Suppose that the risky assets are stocks and that the prices ΠUi (t) = Si (t) are independent
and that each of them follows a binomial model:
Si (j − 1)eui , with probability pi
(
Si (j) = , j ∈ I = {1, . . . , N }, i = 1, . . . n.
Si (j − 1)edi , with probability = 1 − pi

Let us compute the Markowitz portfolio in the N days period. We have

1
E[Ri ] = E[Si (0)−1 (Si (N ) − Si (0))] = E[Si (N )] − 1.
Si (0)

By Theorem ?? we have E[Si (N )] = Si (0)(edi pi + eui (1 − pi ))N . Hence

E[Ri ] = (edi pi + eui (1 − pi ))N − 1. (A.8)

136
Moreover
E[Si (N )2 ] = Si (0)2 (e2di pi + e2ui (1 − pi ))N ,
since Si (t) follows a binomial model with parameters 2u, 2d. Hence

Var[Si (N )] = E[Si (N )2 ] − E[Si (N )]2 = Si (0)2 [(e2di pi + e2ui (1 − pi ))N − (edi pi + eui (1 − pi ))2N ]

and therefore
1
Var[Ri ] = 2
Var[Si (N )] = (e2di pi + e2ui (1 − pi ))N − (edi pi + eui (1 − pi ))2N . (A.9)
Si (0)

Replacing (A.8) and (A.9) into (A.7), and then inverting (A.1), we obtain the desired
Markowitz portfolio (a1 , . . . , an+1 ).

Exercise A.2 (•). Consider a 2-period binomial model with the following parameters
4 2 1
eu = , ed = , p= .
3 3 2
Assume further that S(0) = 36, and that the interest rate of the money market is zero.
Consider also the European derivative with pay-off

Y = (S(2) − 28)+ − 2(S(2) − 32)+ + (S(2) − 36)+

and time of maturity T = 2. Compute the fair value of the derivative at t = 0. Assume now
1
that an investor with risk aversion θ = 36 wants to distribute the initial wealth K = 1000 in
the following assets: the stock, the derivative, and a risk-free asset with interest r such that
e2r = 10/9. Derive the corresponding Markowitz portfolio.

Exercise A.3 (?). Consider two stocks with prices

S1 (t) = S1 (0)eα1 t+σ1 W1 (t) , S2 (t) = S2 (0)eα2 t+σ2 W2 (t)

where σ1 , σ2 > 0, α1 , α2 ∈ R and W1 (t), W2 (t) are two Brownian motions. Let T > 0 and
assume that W1 (T ) and W2 (T ) are independent random variables. Compute the Markowitz
portfolio of an investor with initial capital K > 0 and risk aversion θ who wants to invest in
the stocks and in a money market with interest r > 0 during the interval of time [0, T ].

137
Appendix B

Solutions to selected exercises

Exercise 1.7
To solve this exercise it is useful to begin by drawing the pay-off as a function of S(T ). Since
 
Y = min (S(T ) − K1 )+ , (K2 − S(T ))+ ,
where K2 > K1 and (x)+ = max(0, x), then we first draw the functions S(T ) → (S(T )−K1 )+
and S(T ) → (K2 − S(T ))+ and then we take their minimum.
Now let A be a portfolio that consists of one share of the derivative and let VA (t) be its value
at time t ∈ [0, T ]. The exercise asks to derive a portfolio B consisting of European calls and
puts which replicates the value of A, i.e., such that VA (t) = VB (t), for all t ∈ [0, T ]. By
the dominance principle (in particular by (b) of 1.1) it is enough to find the portfolio B in
such a way that VA (T ) = VB (T ). Clearly VA (T ) = Y , since the value of a derivative at the
expiration date always equals the pay-off. So we we have to find a combination of pay-off
functions of puts and calls such that their sum equal Y . By using the graph of the pay-off
function it is easy to see that
K1 + K2
Y = (S(T ) − K1 )+ − 2(S(T ) − )+ + (S(T ) − K2 )+
2
Even without making any drawing, the previous identity can be verified by direct calculation.
Hence, letting
U1 = European call with strike K1
U2 = European call with strike (K1 + K2 )/2
U3 = European call with strike K2

where for all calls the expiration date is T , we take

B = (1, −2, 1).
This concludes the solution of the exercise.

138
Exercise 1.9
Consider the following assets:

• U1 ≡ Contract

• U2 ≡ Risk-free asset with initial value 1

• U3 ≡ European Call with strike K and maturity T

Consider the two portfolios

A = (0, N e−rT , αN ), B = (1, 0, 0).

We want to show that the portfolio A on the call and the risk-free asset replicates the value
of the contract. To this purpose we observe that

VA (t) = N e−r(T −t) + αN C(t, S(t), K, T ), (B.1)

while VB (t) = N by assumption. Letting t = T in (B.1) we obtain

VA (T ) = N + αN (S(T ) − K)+ = VB (T ),

hence, by the dominance principle, VA (t) = VB (t). Using (B.1) and VB (t) = N we obtain

N e−r(T −t) + αN C(t, S(t), K, T ) = N

and since C(t, S(t), K, T ) > 0 and N 6= 0, we obtain (1.11).

Exercise 3.2
The binomial tree for the price of the stock is

S(2)
7
= 49
u

S(1)
8
= 28
u d
'
S0 = 16 S(2) = 14
7
d u
&
S(1) = 8
d
'
S(2) = 4

139
 er −ed
Moreover qu = eu −ed
= 1/2 = qd . Using the recurrence formula ΠY (2) = Y ,

ΠY (t) = e−r (qu ΠuY (t + 1) + qd ΠdY (t + 1)), t = 0, 1,

we find that the price Πput (t) of the put option is

Πput (2) = 0
6
u

Πput (1) = 0
6
u d
'
128
Πput (0) = Πput (2) = 0
81 7
d u
(
32
Πput (1) = 9
d
'
Πput (2) = 8

The pay-off for the call on the put at time t = T1 = 1 is (Πput − K1 )+ . Note that since
Πput (1) = g(S(1)), then the call on the put can be treated as a standard European derivative
on the stock expiring at time T1 = 1. Hence the price Πcp (t) of the call on the put is

Πcp (1) = 0
7
u

4
Πcp (0) = 9
d
'
Πcp (1) = 1

This completes the first part of the exercise. Now, the return of a portfolio with +1 share
of the call on the put is path dependent. We have
4 4 4 4
R(u, u) = 0 − =− , R(u, d) = 0 − =−
9 9 9 9
4 23 4 23
R(d, u) = 0 − − = −3, R(d, d) = 8 − − = 5.
9 9 9 9
Hence the probability of positive return for the buyer is is
 2
(u,u) 3
P(R > 0) = P(S )= = 75%.
4

140
Exercise 3.3
We have the general formula
X
ΠY (0) = e−rN (qu )Nu (x) (qd )Nd (x) Y (x), (B.2)
x∈{u,d}N

where (qu , qd ) is the risk neutral probability, Nu (x) is the number of “u” in the path x and
Nd (x) = N − Nu (x) is the number of “d” in the path x. Moreover Y (x) denotes the pay-off
as a function of the path of the stock price. The exercise tells us that
Y (x) = S(N, x), if x = x∗ = (u, u, . . . u), i.e., xi = u for all i = 1, . . . N ,
while Y (x) = S(0) for x 6= x∗ . Moreover, since S(N, x∗ ) = S(0)eN u , then
Y (x∗ ) = S(0)eN u .
Since in addition Nu (x∗ ) = N , we can rewrite the sum (B.2) as
X
ΠY (0) = e−rN (qu )Nu (x∗ ) (qd )Nd (x∗ ) Y (x∗ ) + e−rN (qu )Nu (x) (qd )Nd (x) Y (x)
x6=x∗
X
= e−rN (qu )N S(0)eN u + e−rN (qu )Nu (x) (qd )Nd (x) Y (x). (B.3)
x6=x∗

Next we compute the sum on x 6= x∗ . First replacing Nd (x) = N − Nu (x) and Y (x) = S(0)
we have
X X  qu Nu (x)
Nu (x) Nd (x) N
(qu ) (qd ) Y (x) = S(0)(qd ) . (B.4)
x6=x x6=x
qd
∗ ∗

Now, Nu (x) takes value in {0, 1, . . . , N −1} in (B.4); it cannot be equal to N because the only
element in {u, d}N for which Nu (x) = N is x∗ , but this element is not taken into account in
the sum that we are computing. Using
that the number of x ∈ {u, d}N for which Nu (x) = k
is given by the binomial coefficient Nk , we obtain
X  qu Nu (x) N −1    k
X N qu
= .
x6=x
q d
k=0
k q d
∗

Adding and subtracting the term k = N (where we use that N

N
= 1), we find
X  qu Nu (x) N −1    k
X N qu
=
x6=x∗
qd k=0
k qd
N   k  N
X N qu qu
= −
k=0
k qd qd
 N  N
qu qu
= 1+ − ,
qd qd

141
 PN N


where for the last equality we used the binomial theorem: (1 + a)N = k=0 k
ak . Substi-
tuting into (B.4) and using that qu + qd = 1 we obtain
X
(qu )Nu (x) (qd )Nd (x) Y (x) = S(0) 1 − (qu )N .


x6=x∗

Finally, replacing in (B.3) we find

ΠY (0) = e−rN S(0) (qu )N eN u + 1 − (qu )N .

 

Exercise 3.5
X
ΠY (0) = e−rN (qu )Nu (x) (qd )Nd (x) log(S(N )/S(0)),
x∈{u,d}N

by Definition 3.3 (at t = 0). Replacing S(N ) = S(0)eNu (x)u+Nd (x)d we obtain
X
ΠY (0) = e−rN (qu )Nu (x) (qd )Nd (x) (Nu (x)u + Nd (x)d).
x∈{u,d}N

Replacing Nd (x) = N − Nu (x) we obtain

X qu
ΠY (0) = e−rN (qd )N ( )Nu (x) (Nu (x)(u − d) + N d)
qd
x∈{u,d}N
X qu X qu Nu (x)
= e−rN (qd )N (u − d) ( )Nu (x) Nu (x) + N de−rN (qd )N ( )
N
qd qd
x∈{u,d} x∈{u,d}N

Letting Nu (x) = k we obtain

N   N  
−rN N
X N qu k −rN N
X N qu k
ΠY (0) = e (qd ) (u − d) k( ) + N de (qd ) ( ) (B.5)
k=0
k qd k=0
k qd

Using the binomial theorem, the second sum is

N  
X N qu k qu 1
( ) = (1 + )N =
k=0
k qd qd (qd )N

In the first sum we use the identity

   
N N −1
k =N .
k k−1

142
Hence
N   N  
X N qu k X N qu
k( ) = k( )k
k=0
k qd k=1
k qd
N   N −1  
X N − 1 qu k X N − 1 qu j+1
=N ( ) =N ( )
k=1
k − 1 qd j=0
j qd
N −1  
qu X N − 1 q u j qu qu
=N ( ) = N (1 + )N −1
qd j=0 j qd qd qd
qu
=N
(qd )N
Replacing in (B.5) we find
qu 1
ΠY (0) = e−rN (qd )N (u − d)N N
+ N de−rN (qd )N
(qd ) (qd )N
= N e−rN (qu (u − d) + d) = N e−rN (qu u + d(1 − qu )) = N e−rN (qu u + qd d).

Exercise 3.7
We start by writing down the diagram of the stock price and the value of the derivative at
time of maturity T = 3 (which is equal to the pay-off)

S(3) =5 5 ⇒ ΠY (3) = 5
u

S(2)
8
=4
u d
)
16
S(1) = S(3) = 2 ⇒ ΠY (3) = 2
7 5 5
u d u
&
64 8
S(0) = S(2) =
25 8 5
d u d
' )
32 4
S(1) = S(3) = ⇒ ΠY (3) = 0
25 5 5
d u
&
16
S(2) = 25
d
)
8
S(3) = 25
⇒ ΠY (3) = 0

143
The parameters of the binomial model are such that
2 1
qu = , qd = , r = 0.
3 3
To compute the price of the derivative at the times t ∈ {0, 1, 2} we use the recurrence formula
2 1
ΠY (t) = e−r (qu ΠuY (t + 1) + qd ΠdY (t + 1)) = ΠuY (t + 1) + ΠdY (t + 1), t ∈ {0, 1, 2}.
3 3
Hence at time t = 2 we have
2 1
S(2) = 4 ⇒ ΠY (2) = · 5 + · 2 = 4
3 3
8 2 1 4
S(2) = ⇒ ΠY (2) = · 2 + · 0 =
5 3 3 3
16 2 1
S(2) = ⇒ ΠY (2) = · 0 + · 0 = 0.
25 3 3
At time t = 1 we have
16 2 1 4 28
S(1) = ⇒ ΠY (1) = · 4 + · =
5 3 3 3 9
32 2 4 1 8
S(1) = ⇒ ΠY (1) = · + · 0 =
25 3 3 3 9
and at time t = 0 we have
2 28 1 8 64
ΠY (0) = · + · =
3 9 3 9 27
Hence we obtain the following diagram for the derivative price
Π7Y (3) = 5
u

ΠY (2) = 4
7
u d
'
28
ΠY (1) = ΠY7 (3) = 2
7 9
u d u
'
64 4
ΠY (0) = ΠY (2) =
27 7 3
d u d
' '
8
ΠY (1) = ΠY (3) = 0
9 7
d u
'
ΠY (2) = 0
d
'
ΠY (3) = 0

144
This concludes the first part of the exercise. To compute the number of shares of the
underlying asset in the hedging portfolio we use the formula
1 ΠuY (t + 1) − ΠdY (t + 1)
hS (t + 1) =
S(t) eu − ed
for t = 1, 2 and hS (0) = hS (1), where we recall that hS (t + 1) is the position in the interval
(t, t + 1]. Letting t = 2 we obtain
4 ΠuY (3) − ΠdY (3)
hS (3) = ,
3 S(2)
whence
4 5−2
S(2) = 4 ⇒ hS (3) = · =1
3 4
8 4 2−0 5
S(2) = ⇒ hS (3) = · =
5 3 8/5 3
16
S(2) = ⇒ hS (3) = 0.
25
Likewise
4 ΠuY (2) − ΠdY (2)
hS (2) = .
3 S(1)
Hence
16 4 4 − 4/3 10
S(1) = ⇒ hS (2) = · =
5 3 16/5 9
32 4 4/3 − 0 25
S(1) = ⇒ hS (2) = · =
25 3 32/25 18
and finally
4 ΠuY (1) − ΠdY (1) 4 28/9 − 8/9 125
hS (0) = hS (1) = = · = .
3 S(0) 3 64/25 108
This concludes the second part of the exercise. Consider now a constant portfolio with -1
shares of the derivative. The return of this portfolio is positive if the value of the derivative
at the expiration date is smaller than the initial value. This happens along all paths except
x = (u, u, u), hence the probability that the return of this portfolio be positive is 1 − p3 =
1 − 1/8 = 7/8 = 87.5%.

Exercise 4.4
With the given values of the parameters u, d, r, we have
er − ed 1 − 23 1
qu = = 4 2 = = qd .
eu − ed 3
− 3
2

145
The fair price Π
b Y (t) of the American derivative satisfies

b Y (t) = max(Y (t), e−r (qu Π

Π b uY (t + 1) + qd Πb dY (t + 1)))
1 bu b dY (t + 1))),
= max(Y (t), (Π (t + 1) + Π
2 Y

where Πb u (t) (resp. Π

b d (t)) is the price of the derivative at time t assuming that the stock
Y Y
price goes up (resp. down) at time t. The diagram of the stock price is

S(3) = 64
7
u

S(2) = 48
7
u d
'
S(1)
7
= 36 S(3)
7
= 32
u d u
'
S(0) = 27 S(2)
7
= 24
d u d
' '
S(1) = 18 S(3)
7
= 16
d u
'
S(2) = 12
d
'
S(3) = 8

146
to which there corresponds the following diagram for the intrinsic value:

Y (3) = 0 = Π
b Y (3)
6
u

Y (2) = 0
8
u d
(
Y (1) = 0 Y (3) = 0 = Π
b Y (3)
7 6
u d u
&
Y (0) = 0 Y (2) = 0
8
d u d
' (
Y (1) = 6 Y (3) = 8 = Π
b Y (3)
6
d u
&
Y (2) = 12
d
(
Y (3) = 8 = Π
b Y (3)

Therefore

S(2) = 48 ⇒ Π
b Y (2) = 0, S(2) = 24 ⇒ Π
b Y (2) = 4, S(2) = 12 ⇒ Π
b Y (2) = 12

S(1) = 36 ⇒ Π
b Y (1) = 2, S(1) = 18 ⇒ Π
b Y (1) = 8,

147
and Π
b Y (0) = 5. We thereby obtained the following diagram for the price of the derivative:

Π
b Y (3) = 0
7
u

Π
b Y (2) = 0
7
u d
'
Π
b Y (1) = 2
7
Π
b Y (3) = 0
7
u d u
'
Π
b Y (0) = 5 Π
b Y (2) = 4
7
d u d
' '
Π
b Y (1) = 8 Π
b Y (3) = 8
8
d u
&
Π
b Y (2) = 12
d
&
Π
b Y (3) = 8

This completes the first part of the exercise. The only case in which the price of the derivative
equals its intrinsic value prior to expiration is at time t = 2 when the price of the stock is
S(2) = 12 (i.e., the stock price goes down in the first two steps). This is indicated in the
previous diagram by putting the price of the derivative in a box. In this case, and only in
this case, it is optimal to exercise the derivative prior to expiration. If the buyer does not
exercise the derivative at the optimal moment, the writer can withdraw the amount

b Y (2) − e−r [qu Π

C(2) = Π b dY (3)] = 12 − [ 1 8 − 1 8] = 4.
b Y (3)u + qd Π
2 2
Note that after withdrawing this cash, the value of the portfolio is 8, which is exactly what
the writer needs to hedge the derivative!

148
Exercise 4.5
The binomial trees for the stock price S(t) and the intrinsic value Y (t) are as follows.

S(3)
7
=5
u

S(2)
8
=4
u d
&
16
S(1) = S(3) = 2
7 5 8
u d u
&
64 8
S(0) = S(2) =
25 8 5
d u d
' &
32 4
S(1) = S(3) =
25 8 5
d u
&
16
S(2) = 25
d
&
8
S(3) = 25

Y (3) = 2
8
u

Y (2) = 1
8
u d
&
1
Y (1) = Y 8 (3) = 1
7 5
u d u
&
11 7
Y (0) = Y (2) =
25 8 5
d u d
' &
Y (1) = 0 Y (3) = 0
8
d u
&
Y (2) = 0
d
&
Y (3) = 0

149
The binomial price of the American derivative is defined as
Π
b Y (3) = Y (3), b Y (t) = max[Y (t), e−r (qu Π
Π b u (t + 1) + qu Π
b u (t + 1))], t = 0, 1, 2,
Y Y

where qu = 2/3 and qd = 1/3. Applying the above formula one finds the following binomial
tree for Π
b Y (t).

Π
b Y (3) = 2
8
u

5
Π
b Y (2) =
8 3
u d
&
71
Π
b Y (1) = Π
b Y (3) = 1
7 45 8
u d u
&
184 7
Π
b Y (0) = Π
b Y (2) =
135 8 5
d u d
' &
14
Π
b Y (1) =
15
Π
b Y (3) = 0
8
d u
'
Π
b Y (2) = 0
d
&
Π
b Y (3) = 0

This concludes the first part of the exercise. The initial position on the stock in the hedging
portfolio is
1 Π b u (1) − Π
Y
b d (1)
Y 1 71
45
− 14
15 145
hS (0) = hS (1) = = 64 3 = ,
S(0) eu − ed 25 4
432
which answers the second question. As to the cash flow, observe that the only optimal
exercise time is t = 2 when S(2) = 8/5, as in this case, and only in this case, Π b Y (t) and
Y (t) are equal. If the buyer does not exercise the derivative at this instance, the seller can
withdraw the amount

C(2) = Π b Y (2) − e−r (qu Π b u (t + 1)) = 7 − 2 = 11 .

b u (t + 1) + qu Π
Y Y
5 3 15
This answers the third question. The probability that the derivative is in the money at time
t is P(Y (t) > 0). We have
P(Y (0) > 0) = 1, P(Y (1) > 0) = p,
P(Y (2) > 0) = p2 + 2p(1 − p) = p(2 − p), P(Y (3) > 0) = p2 + 3p2 (1 − p) = p2 (3 − 2p).

150
This answers the fourth question. Finally, the return for the buyer at time t is positive if and
only if Y (t) > ΠY (0) (the question is relevant because the buyer can exercise at any time).
This happens only at time t when S(2) = 8/5 (the optimal exercise time) and at maturity
t = 3 when S(3) = 5. Hence we have

P(R(0) > 0) = 0, P(R(1) > 0) = 0,

P(R(2) > 0) = 2p(1 − p) P(R(3) > 0) = p3 .

This completes the exercise.

Exercise 4.6
The binomial tree for the stock price and for the intrinsic value Y (t) of the American put
are
S(2)
8
=4
u

16
S(1) =
7 5
u d
&
64 8
S(0) = S(2) =
25 8 5
d u
'
32
S(1) = 25
d
&
16
S(2) = 25

Y (2) = 0
7
u

Y (1) = 0
7
u d
'
3
Y (0) = 0 Y (2) =
8 5
d u
'
23
Y (1) = 25
d
&
39
Y (2) = 25

151
 b put (t) be the price at time t ∈ {0, 1, 2} of the American put. We have the recurrence
Let Π
formula Π b put (2) = Y (2) and
b put (t) = max[Y (t), e−r (qu Π
Π b dput (t))],
b uput (t) + qd Π

where
er − ed 2 1
qu = u d
= , qd = 1 − qu = , er = 1.
e −e 3 3
Hence the binomial tree for Πb put (t) is

Π
b put (2) = 0
7
u

1
Π
b put (1) =
6 5
u d
'
11 3
Π
b put (0) = Π
b put (2) =
25 7 5
d u
(
23
Π
b put (1) =
25
d
'
39
Π
b put (2) =
25

The only optimal exercise time prior to maturity is t = 1 when S(1) = 32 25

. This concludes
the first part of the exercise. The compound option has maturity T = 1 and pay-off

Q = (Π b put (1) − 8 )+ .
25
Since Πb put (1) is a function of S(1), then we can treat the compound option as a standard
derivative on the stock. The compound option expires in the money if the stock price goes
down at time t = 1 and out of the money otherwise. Hence the price of the compound option
at time t = 0 is
1 23 8 1
Πcp (0) = ( − )+ = .
3 25 25 5
This answers the second question. As to the hedging portfolio, the compound option can be
hedged by investing on the stock and the risk-free asset. The number of shares in the stock
is
1 Πucp (1) − Πdcp (1) 25 0 − 35 5
hS = = 5 1 = −
S(0) eu − ed 64 4 − 2 16
The number of shares in the risk-free asset is obtained by solving the replicating equation
at time t = 0:
Πcp (0) − hS S(0) 31
hS S(0) + hB B(0) = Πcp (0) ⇒ hB = = .
B(0) 25

152
This answers the third question. Finally we compute the expected return E[R] for the owner
of the compound option as a function of p ∈ (0, 1). Clearly
1
R=− with prob. p,
5
which is the return when the stock price goes up at time t = 1. If the stock price goes down
at time t = 1, the owner of the compound option will buy the American put for K1 = 8/25.
If the American put is exercised at this optimal exercise time, then the return will be
23 1 8 2
R= − − = with prob. 1 − p.
25 5 25 5
Hence, if the American put is exercised at t = 1, the expected return is
1 2 2 3
E[R] = − p + (1 − p) = − p.
5 5 5 5
If the American put is exercised at t = 2, the expected return is
1 3 8 1
39 8 1
1
E[R] = − p + − − p(1 − p) + − − (1 − p)2 = (3p − 2)(8p − 13) = f (p).
5 5 25 5 25 25 5 25
Now, it is straightforward to verify that f (p) > 25 − 53 p when 0 < p < 2/3 and f (p) < 25 − 35 p
when 2/3 < p < 1. Hence the strategy which maximizes the expected return for the
compound option is: for 0 < p < 2/3, the American put should not be exercised at time
t = 1, while for 2/3 < p < 1 the American put should be exercised at time t = 1. For
p = 2/3 the two strategies lead to the same expected return. This answers the last question.

Exercise 5.9
Let A be an event that is resolved by both variables X, Y . This means that there exist
I, J ⊆ R such that A = {X ∈ I} = {Y ∈ J}. Hence, using the independence of X, Y ,

P(A) = P(A ∩ A) = P(X ∈ I, Y ∈ J) = P(X ∈ I)P(Y ∈ J) = P(A)P(A) = P(A)2 .

Therefore P(A) = 0 or P(A) = 1. In a finite probability space this implies A = ∅ or A = Ω,

respectively.
Now let a, b be two deterministic constants. Note that, for all I ⊂ R,

1 if a ∈ I
P(a ∈ I) =
0 otherwise

and similarly for b. Hence


1 if a ∈ I and b ∈ J
P(a ∈ I, b ∈ J) = = P(a ∈ I)P(b ∈ J).
0 otherwise

153
Finally we show that X and Y = g(X) are independent if and only if Y is a deterministic
constant. For the “if” part we use that

P(X ∈ J) if a ∈ I
P(a ∈ I, X ∈ J) = = P(a ∈ I)P(X ∈ J).
0 otherwise

For the “only if” part, let z ∈ R and I = {g(X) ≤ z} = {X ∈ g −1 (−∞, z]}. Then, using
the independence of X and Y = g(X),

P(g(X) ≤ z) = P(g(X) ≤ z, g(X) ≤ z) = P(X ∈ g −1 (−∞, z], g(X) ≤ z)

= P(X ∈ g −1 (−∞, z])P(g(X) ≤ z) = P(g(X) ≤ z)P(g(X) ≤ z).

Hence P(Y ≤ z) is either 0 or 1, which implies that Y is a deterministic constant.

Exercise 5.10
The exercise asks to prove the following:
Let X1 , X2 be independent random variables, g : R → R, and f : R → R. Then the random
variables
Y = g(X1 ), Z = f (X2 )
are independent.
Given I, J ⊆ R we have {Y ∈ I} = {X1 ∈ {g ∈ I}} and {Z ∈ J} = {X2 ∈ {f ∈ J}}.
Hence, using the independence of X1 , X2 ,

P(Y ∈ I, Z ∈ J) = P(X1 ∈ {g ∈ I}, X2 ∈ {f ∈ J})

= P(X1 ∈ {g ∈ I})P(X2 ∈ {f ∈ J}) = P(Y ∈ I)P(Z ∈ J).

Exercise 5.11
The statement X, Y independent ⇒ X, Y uncorrelated holds for random variables on gen-
eral probability spaces, but here we are only concerned with finite probability spaces. In
particular, X can only take a finite number of values x1 , . . . xN and Y a finite number of
values y1 , . . . yM . Letting Ai = {X = xi }, Bj = {Y = yj }, i = 1, . . . N , j = 1, . . . M , and
denoting IA the indicator function of the set A, we have
N
X M
X
X= xi IAi , Y = yj IBj .
i=1 j=1

Hence
N X
X M N X
X M
XY = xi yj IAi IBj = xi yj IAi ∩Bj
i=1 j=1 i=1 j=1

154
Hence, by the linearity of the expectation, and the assumed independence of X, Y ,
N X
X M
E[XY ] = xi yj E[IAi ∩Bj ]
i=1 j=1
N X
X M
= xi yj P(Ai ∩ Bj )
i=1 j=1
N X
X M
= xi yj P(Ai )P(Bj )
i=1 j=1
N
X M
X
= xi P(Ai ) P(Bj ) = E[X]E[Y ].
i=1 j=1

As an example of uncorrelated, but not independent, random variables X, Y , consider


 −1 with prob. 1/3
X= 0 with prob. 1/3 Y = X 2.
1 with prob. 1/3


The random variables X, Y are not independent, since Y is not a deterministic constant (see
Exercise 5.9 above). Moreover XY = X 3 = X and thus E[XY ] = E[X 3 ] = E[X] = 0. Since
E[X]E[Y ] = 0, then Cov(X, Y ) = 0, i.e., the two random variables are uncorrelated. As
to (5.18), we write

Var[X + Y ] = E[(X + Y )2 ] − E[(X + Y )]2 = Var[X] + Var[Y ] + 2(E[XY ] − E[X]E[Y ]),

by which the claim follows.

Exercise 5.12
To prove the inequality we first we notice that

Var[αX] = E[α2 X 2 ] − E[αX]2 = α2 E[X 2 ] − α2 E[X]2 = α2 Var[X],

Cov(αX, Y ) = E[αXY ] − E[αX]E[Y ] = αCov(X, Y )

and

Var[X + Y ] = E[(X + Y )2 ] − E[X + Y ]2 = E[X 2 ] + E[Y 2 ] + 2E[XY ]

− E[X]2 − E[Y ]2 − 2E[X]E[Y ] = Var[X] + Var[Y ] + 2Cov(X, Y ).

Hence letting a ∈ R we have

Var[Y − aX] = a2 Var[X] + Var[Y ] − 2aCov(X, Y ).

155
Since the variance of a random variable is always non-negative, the parabola

y(a) = a2 Var[X] + Var[Y ] − 2aCov(X, Y )

must always lie above the a-axis, or touch it at one single point a = a0 . Hence

Cov(X, Y )2 − Var[X]Var[Y ] ≤ 0,

which proves (5.19). Moreover Cov(X, Y )2 = Var[X]Var[Y ] if and only if there exists a0
such that Var[−a0 X + Y ] = 0, i.e., Y = a0 X + b0 , for some constant b0 . Note that a0 6= 0,
otherwise Y is a deterministic constant. Substituting in the definition of covariance, we see
that Cov(X, a0 X + b0 ) = a0 Var[X]. Hence if the right inequality in (5.19) is an equality we
have p
a0 Var[X] = Var[X]Var[a0 X + b0 ]), i.e., a0 Var[X] = |a0 |Var[X],
and thus a0 > 0. Similarly one shows that if the left inequality becomes an equality then
a0 < 0.

Exercise 5.17
By linearity of the expectation,
1
E[Wn (t)] = √ E[M[nt] ] = 0,
n
where we used that E[Xk ] = E[Mk ] = 0. Since Var[Mk ] = k, we obtain
[nt]
Var[Wn (t)] = .
n
Since nt ∼ [nt], as n → ∞, then limn→∞ Var[Wn (t)] = t. As to the covariance of Wn (t) and
Wn (s) for s 6= t, we compute

Cov[Wn (t), Wn (s)] = E[Wn (t)Wn (s)] − E[Wn (t)]E[Wn (s)] = E[Wn (t)Wn (s)]
 
1 1 1
= E √ M[nt] √ M[ns] = E[M[nt] M[ns] ]. (B.6)
n n n
Assume t > s (a similar argument applies to the case t < s). If [nt] = [ns] we have
E[M[nt] M[ns] ] = Var[M[ns] ] = [ns]. If [nt] ≥ 1 + [ns] we have
2
E[M[nt] M[ns] ] = E[(M[nt] −M[ns] )M[ns] ]+E[M[ns] ] = E[M[nt] −M[ns] ]E[M[ns] ]+Var[M[ns] ] = [ns],

where we used that the increment M[nt] − M[ns] is independent of M[ns] . Replacing into (B.6)
we obtain
[ns]
Cov[Wn (t), Wn (s)] = .
n
It follows that limn→∞ Cov[Wn (t), Wn (s)] = s.

156
Exercise 5.19
The second formula follows by the first one using that eu q + ed (1 − q) = er (or by letting
t = N in (5.28)). To prove the first formula we use
Ep [S(N )] = Ep [S(0) exp(X1 + · · · + XN )] = S(0)Ep [Y ],
where Y is the random variable
Y (ω) = exp(X1 (ω) + · · · + XN (ω)) = exp(uNH (ω) + dNT (ω)), ω ∈ ΩN .
Hence, using NT (ω) = N − NH (ω) and (5.10) we obtain
X
Ep [S(N )] = S(0) e(uNH (ω)+dNT (ω)) pNH (ω) (1 − p)NT (ω)
ω∈ΩN
 NH (ω)
Nd N
X
(u−d)NH (ω) p
= S(0)e (1 − p) e .
ω∈ΩN
(1 − p)

Now we use that for k = 0, . . . , N there exist exactly Nk sample points ω ∈ ΩN such that

NH (ω) = k. Hence we can write

N   k
Nd N
X N eu p
Ep [S(N )] = S(0)e (1 − p) d (1 − p)
.
k=0
k e
N
By the binomial theorem, (1 + a)N = k=0 Nk ak , hence
P

N
eu p

Nd N
Ep [S(N )] = S(0)e (1 − p) 1+ d = S(0)(ed (1 − p) + eu p)N .
e (1 − p)
Similarly one finds that Varp [S(N )] = S(0)2 [(e2u p + e2d (1 − p))N − (eu p + ed (1 − p))2N ].

Exercise 5.20
By the risk neutral pricing formula (5.35) at time t = 0 we have AC(0) = e−rN Eq [Ycall (x)]
and similarly for the Asian put. Thus
AC(0) − AP (0) = Eq [Ycall − Yput ]
Using
" N
# " N
# N
 1 X  1 X    1 X 
Ycall −Yput = S(t) − K − K− S(t) = S(t) −K
N + 1 t=0 N + 1 t=0 N + 1 t=0
+ +

we find !
N
e−rN X
AC(0) − AP (0) = Eq [S(t)] − Ke−rN Eq [1],
N +1 t=0

157
where Eq [·] denotes the expectation in the risk-neutral measure. As Eq [1] = 1 and Eq [S(t)] =
S(0)ert , we obtain
N
!
h S(0) X i
AC(0) − AP (0) = e−rN ert − K .
N + 1 t=0
Using the formula in the HINT concludes the exercise.

Exercise 5.27
As E[W (t)] = 0 for all t ≥ 0,
Cov[W (s), W (t)] = E[W (s)W (t)].
Assume t > s (for t < s the argument is identical). Using that the increments W (t) − W (s)
and W (s) = W (s) − W (0) are independent we have
E[W (s)W (t)] = E[W (s)(W (t) − W (s))] + E[W (s)2 ]
= E[W (s)]E[W (t) − W (s)] + Var[W (s)] = Var[W (s)] = s.

Exercise 6.9
Recall that
C(t, x) = xΦ(d1 ) − Ke−rτ Φ(d2 ), (B.7)
where
x
+ r − 12 σ 2 τ



log K
√
d2 = √ , d1 = d2 + σ τ , (B.8)
σ τ
Rx 1 2
and where Φ(x) = √1
2π −∞
e− 2 y dy is the standard normal distribution. As σ → 0+ we have
d1 → d2 and
1 x
d2 ∼ √ (log + rτ )σ −1 .
τ K
Hence
d2 → +∞, if x > Ke−rτ ,
d2 → −∞, if x < Ke−rτ ,
d2 → 0, if x = Ke−rτ ,
Thus
lim Φ(d1 ) = lim+ Φ(d2 ) = 1, if x > Ke−rτ ,
σ→0+ σ→0
lim Φ(d1 ) = lim+ Φ(d2 ) = 0, if x < Ke−rτ ,
σ→0+ σ→0
lim+ Φ(d1 ) = lim+ Φ(d2 ) = Φ(0), if x = Ke−rτ .
σ→0 σ→0

158
It follows that
lim C(t, x) = x − Ke−rτ if x > Ke−rτ ,
σ→0+
lim C(t, x) = 0, if x ≤ Ke−rτ ,
σ→0+

i.e., limσ→0+ C(t, x) = (x − Ke−rτ )+ . For σ → +∞ we have d2 → −∞ and d1 → +∞, hence

Φ(d1 ) → 1 and Φ(d2 ) → 0. Thus C(t, x) → x as σ → +∞. As K → 0+ , both d1 and d2
diverge to +∞, hence
lim+ C(t, x) = x.
K→0
For K → +∞, d1 , d2 diverge to −∞. Hence the first term in C(t, x) converges to zero. As
the first term in C(t, x) always dominates the second term (since C(t, x) > 0), then the
second term also goes to zero and thus
lim C(t, x) = 0.
K→+∞

For T → +∞ we obtain
lim C(t, x) = x.
T →+∞
Finally, for x → 0+ , both d1 , d2 diverge to −∞ and thus
lim C(t, x) = 0.
x→0+
To compute the limits for put options we use the put-call parity:
C(t, x) − P (t, x) = x − Ke−rτ ,
by which it follows that
lim+ P (t, x) = (Ke−rτ − x)+ , lim P (t, x) = Ke−rτ
σ→0 σ→+∞

lim P (t, x) = 0, lim P (t, x) = +∞,

K→0+ K→+∞

lim P (t, x) = 0, lim P (t, x) = Ke−rτ .

T →+∞ x→0+

Exercise 6.13
The pay-off function is g(z) = k + z log z. Hence the Black-Scholes price of the derivative is
ΠY (t) = v(t, S(t)), where
Z   2 √ 
r− σ2 τ +σ τ y y 2 dy
−rτ
v(t, x) = e g xe e− 2 √
2π
ZR 
σ2 √

2 √ y 2 dy
−rτ (r− σ2 )τ +σ τ y
=e k + xe (log x + (r − )τ + σ τ y) e− 2 √
R 2 2π
Z √
1 2 dy
= ke−rτ + x log x e− 2 (y−σ τ ) √
2π
Z R
2 √
√
Z √ 2 dy
σ − 21 (y−σ τ )2 dy 1
+ x(r − )τ e √ + xσ τ ye− 2 (y−σ τ ) √
2 R 2π R 2π

159
Using that
√
Z √
Z √
− 12 (y−σ τ )2 dy 1
τ )2 dy
e √ = 1, ye− 2 (y−σ √ = σ τ,
R 2π R 2π
we obtain
σ2
v(t, x) = ke−rτ + x log x + x(r + )τ.
2
Hence
σ2
ΠY (t) = ke−rτ + S(t) log S(t) + S(t)(r +
)τ.
2
This completes the first part of the exercise. The number of shares of the stock in the hedging
portfolio is given by
hS (t) = ∆(t, S(t)),
∂v σ2
where ∆(t, x) = ∂x
= log x + 1 + (r + 2
)τ . Hence
σ2
)τ + log S(t).
hS (t) = 1 + (r +
2
The number of shares of the risk-free asset is obtained by using that
ΠY (t) = hS (t)S(t) + B(t)hB (t),
hence
1
hB (t) = (ΠY (t) − hS (t)S(t))
B(t)
σ2 σ2
= e−rt (ke−rτ + S(t) log S(t) + S(t)(r + )τ − S(t) − S(t)(r + )τ − S(t) log S(t))
2 2
= ke−rT − S(t)e−rt .
This completes the second part of the exercise. To compute the probability that Y > 0,
we first observe that the pay-off function g(z) has a minimum at z = e−1 and we have
g(e−1 ) = k − e−1 . Hence if k ≥ e−1 , the derivative has probability 1 to expire in the money.
If k < e−1 , there exist a < b such that
g(z) > 0 if and only if 0 < z < a or z > b.
−1
Hence for k < e we have
P(Y > 0) = P(S(T ) < a) + P(S(T ) > b).
√
Since S(T ) = S(0)eαT +σ TG
, with G ∈ N (0, 1), then
S(0)
log a
+ αT log S(0)
b
+ αT
S(T ) < a ⇔ G < − √ := −A, S(t) > b ⇔ G > − √ := −B.
σ T σ T
Thus
Z −A Z +∞
dy − y2
2 y 2 dy
P(Y > 0) = P(G < −A) + P(G > −B) = e √ + e− 2 √
−∞ 2π −B 2π
= Φ(−A) + 1 − Φ(−B) = 1 − Φ(A) + Φ(B).
This completes the solution of the third part of the exercise.

160
Exercise 6.14
The pay-off function is g(z) = z(z−K); the Black-Scholes price is given by ΠY (t) = v(t, S(t)),
where
Z √
σ2 y 2 dy
−rτ
v(t, x) = e g(xe(r− 2 )τ −σ τ y )e− 2 √
ZR h 2π
σ 2 √ i h σ2 √ i y2 dy
= e−rτ xe(r− 2 )τ −σ τ y se(r− 2 )τ −σ τ y − K e− 2 √
2π
h R 2 i
= x xe(r+σ )τ − K .
This solves the first part of the exercise. The number of shares of the underlying stock in
the hedging portfolio is given by hS (t) = ∆(t, S(t)), where
∂v 2
∆(t, x) = = 2xe(r+σ )τ − K.
∂x
The number of shares of the risk-free is obtained using
1 e−rt (r+σ2 )τ
ΠY (t) = hS (t)S(t) + hB (t)B(t) ⇒ hB (t) = (ΠY (t) − hS (t)S(t)) = − e S(t)2 .
B(t) B(0)
This completes the second part of the exercise. The relative return of the portfolio is
ΠY (T )
R= − 1.
ΠY (0)
2 )T
Using ΠY (T ) = S(T )(S(T ) − K) and ΠY (0) = K 2 (e(r+σ − 1), we obtain
E[S(T )(S(T ) − K)]
E[R] = − 1.
K 2 (e(r+σ2 )T − 1)
Writing the geometric Brownian motion at time T as
σ2 σ2
√
S(T ) = S(0)e(µ− 2
)T +σW (T )
= Ke(µ− 2
)T +σ TG
,
√
where G = W (T )/ T ∈ N (0, 1), we get
√
Z
2 2 (2µ−σ 2 )T dy T y− 12 y 2
2
E[S(T ) ] = K e e2σ √ = K 2 e(2µ+σ )T ,
R 2π
√
Z
σ 2 1 2 dy
E[S(T )] = Ke(µ− 2 )T eσ T y− 2 y √ = KeµT .
R 2π
Therefore
2 )T
E[S(T )(S(T ) − K)] = E[S(T )2 ] − KE[S(T )] = K 2 eµT (e(µ+σ − 1).
We conclude that the expected relative return is given by
2
eµT (e(µ+σ )T − 1)
E[R] = − 1.
e(r+σ2 )T − 1
This completes the third part of the exercise.

161
Exercise 6.15
The graph of the pay-off looks like in the following picture (the numbers on the axes are
irrelevant).
10

K 2K
10 20 30 40 50

-5

-10

By this picture one can see that

g(x) = (K − x)+ − (x − K)+ + (x − 2K)+ = g1 (x) + g2 (x) + g3 (x).

As the Black-Scholes price is linear in the pay-off function, the Black-Scholes price of the
derivative is the sum of the Black-Scholes price of the derivatives with pay-off functions
g1 , g2 , g3 , hence

ΠY (t) = P (t, S(t), K, T ) − C(t, S(t), K, T ) + C(t, S(t), 2K, T )

= Ke−rτ − S(t) + C(t, S(t), 2K, T ),

where for the second equality we used the put-call parity. Here C(t, S(t), K, T ) denotes the
Black-Scholes price of the European call with strike K and maturity T . Hence

C(t, S(t), 2K, T ) = S(t)Φ(de1 ) − 2Ke−rτ Φ(de2 ),

where 2 2
log S(t)
2K
+ (r + σ2 )τ log S(t)
2K
+ (r − σ2 )τ
d1 =
e √ , d2 =
e √ .
σ τ σ τ
This completes the second part of the exercise. For the last part, denote

f (K) = Ke−rτ − S(t) + C(t, S(t), 2K, T )

the price of the derivative as a function of K. We want to prove that there exists K∗ > 0
such that f (K∗ ) = 0. First we observe that f (K) > 0 for K > S(t)erτ . Hence, if we prove

162
that f (K) can also take negative values, then, since f is continuous, there must exist K∗
such that f (K∗ ) = 0. To this purpose define K0 such that de2 = 0, that is
1 σ2
K0 = S(t)e(r− 2 )τ .
2
For this strike price we have Φ(de2 ) = 1/2 and so the function f evaluated at K0 is
f (K0 ) = (Φ(de1 ) − 1)S(t).
As Φ(de1 ) < 1, we find f (K0 ) < 0. Hence there exists K0 < K∗ < S(t)erτ such that
f (K∗ ) = 0. The value K∗ is the fair value of the strike price for the derivative. In fact, as
both the buyer and the seller of this derivative can loose money at maturity, the fair price
of the derivative should be zero. This concludes the solution of the exercise.

Exercise 6.19
Letting a = 1 − e−rT , the price at time t = 0 of a European derivative with pay-off function
g(x) = (x − S(0))+ is
√
Z
2 y 2 dy
−rT
ΠY (0) = e g((1 − a)S(0)e(r−σ /2)T +σ T y )e− 2 √
R3 2π
√
Z
σ 2 y 2
dy
= e−rT S(0) (e− 2 T +σ T y − 1)+ e− 2 √
3 2π
ZR+∞  √
1 2 y 2  dy
= e−rT S(0) √ e− 2 (y−σ T ) − e 2 √
σ T /2 2π
√ √
σ T σ T
= e−rT S(0)[Φ( ) − Φ(− )]
2√ 2
σ T
= S(0)e−rT (2Φ( ) − 1)
2
where Φ is the standard normal cumulative distribution. The value of the given portfolio at
times t = 0, T is
V (T ) = −g(S(T )) − S(0)e−rT erT = −(S(T ) − S(0))+ − S(0)
√
−rT −rT σ T
V (0) = −ΠY (0) − S(0)e = −2S(0)e Φ( )
2
Hence R(T ) = V (T ) − V (0) is given by
√
−S(T ) + 2S(0)e−rT

√ Φ(σ T /2) if S(T ) > S(0)
R(T ) = −rT
S(0)[2e Φ(σ T /2) − 1] if S(T ) < S(0)
In particular, for S(T ) > S(0) we have
√
R(T ) < S(0)[2e−rT Φ(σ T /2) − 1], for S(T ) > S(0).

163
 √
It follows√that for 2e−rT Φ(σ T /2) − 1 ≤ 0, the portfolio return is always negative. For
2e−rT Φ(σ T /2) − 1 > 0 we have
√
P(R(T ) > 0) = P(S(T ) > S(0)) + P(S(0) < S(T ) < 2S(0)e−rT Φ(σ T /2))
√
= P(S(T ) < 2S(0)e−rT Φ(σ T /2))
As √ √
2 /2)T +σ σ2
S(T ) = (1 − a)S(0)e(r−σ TG
= S(0)− 2
T +σ TG
,
where G is a standard normal random variable, then
√
S(T ) < 2S(0)e−rT Φ(σ T /2) ⇔ G ≤ δ
where 2 √
( σ2 − r)T + log(2Φ(σ T /2))
δ= √ .
σ T
Hence √
P(S(T ) < 2S(0)e−rT Φ(σ T /2)) = P(G ≤ δ) = Φ(δ).

Exercise A.2
The diagram of the stock price is
S(2)4 = 64 ⇒ ΠY (2) = 0
u

S(1)
7
= 48
u d
*
S(0) = 36 S(2)
4
= 32 ⇒ ΠY (2) = 4
d u
'
S(1) = 24
d
*
S(2) = 16 ⇒ ΠY (2) = 0

where we also indicated the value of the derivative at the expiration date (which is equal to
the pay-off). The parameters of the binomial model are
1
q u = qd = , r = 0.
2
To compute the price of the derivative at the times t ∈ {0, 1} we use the recurrence formula
1
ΠY (t) = e−r (qu ΠuY (t + 1) + qd ΠdY (t + 1)) = (ΠuY (t + 1) + ΠdY (t + 1)), t ∈ {0, 1}.
2
164
Hence at time t = 1 we have
1
S(1) = 48 ⇒ ΠY (1) = (0 + 4) = 2
2
1
S(1) = 24 ⇒ ΠY (1) = (4 + 0) = 2
2
and at time t = 0 we have
1
ΠY (0) = (2 + 2) = 2
2
This concludes the first part of the exercise. Next, let a1 be the number of shares of the
stock and a2 the number of shares of the derivative in the Markowitz portfolio and let
a1 S(0) a2 ΠY (0)
π1 = , π2 = .
K K
Then we have the formula    
π1 1 µ1 − ρ
= C −1 (B.9)
π2 2θ µ2 − ρ
where the various symbols have the following meaning: θ is the risk aversion of the investor
(which is 1/36 in this exercise), µi = E[Ri ], R1 , R2 are the relative returns of the stock and
the derivative, ρ = e2r − 1 (which is 1/9 in this exercise) and C −1 is the inverse of the
covariant matrix Cij = C(Ri , Rj ). Hence we first compute the random variables R1 , R2 ,
which are defined as
S(2) − S(0) ΠY (2) − ΠY (0)
R1 = , R2 = .
S(0) ΠY (0)

Since p = 1/2, then 

 64 with probability 1/4
S(2) = 32 with probability 1/2
16 with probability 1/4


and similarly 
 0 with probability 1/4
ΠY (2) = 4 with probability 1/2
0 with probability 1/4


Note that S(2) and ΠY (2) are not independent variables! In fact, the price of the derivative
at time 2 is uniquely determined by the price of the stock at time 2. Using the above we
find 
 7/9 with probability 1/4
R1 = −1/9 with probability 1/2
−5/9 with probability 1/4


 −1 with probability 1/4
R2 = 1 with probability 1/2
−1 with probability 1/4


165
Hence
7 1 1 1 5 1
µ1 = E[R1 ] = · − · − · =0
9 4 9 2 9 4
and similarly µ2 = E[R2 ] = 0. Next we compute the covariance matrix. We have
49 1 1 1 25 1 19
C11 = Cov(R1 , R1 ) = Var[R1 ] = E[R12 ] = · + · + · = ,
81 4 81 2 81 4 81
C22 = Cov(R2 , R2 ) = Var[R2 ] = 1,
7 1 1 1 5 1 1
C12 = C21 = Cov(R1 , R2 ) = E[R1 R2 ] = − · − · + · = − .
9 4 9 2 9 4 9
Hence the covariant matrix is !
19 1
81
− 9
C= 1
,
−9 1
whose inverse is !
9 1
2 2
C −1 = 1 19
,
2 18

Replacing what we found in (B.9) we obtain,

!  
9 1
  
π1 1 −10
=− = .
π2 1 19 1 −28/9
9

Hence the number of shares in the portfolio are

1000 · 10
π1 = −10 ⇒ a1 = − ≈ −278
36
1000 · 28
π2 = −28/9 ⇒ a2 = − ≈ −1556.
2·9
Hence the investor takes a short position on the risky assets and invest in the risk-free asset
the capital K = 1000 plus the income from the short position. This is natural, because the
risky assets have zero expected return in the present exercise. This concludes the second
part of the exercise.

166
 Appendix C

Answer to selected exercises

Exercise 2.2: V u (1) = 3, V d (1) = 9, V (0) = 15.

Exercise 2.3: hS (1) = −6, hB (1) = 21; if the price of the stock goes up at time t = 1, choose
hS (2) = −1/3, hB (2) = 4; if the price of the stock goes down at time t = 1, choose
hS (2) = −5/2, hB (2) = 14.

Exercise 3.4: ΠY (0) = 5/54. P(Y > 0) = 28, 125%.

Exercise 4.7: b hS (0) = −1/5, b

hS (1) = b hB (1) = b
hB (0) = 14/45; if the stock price goes up at time
1, then bhS (2) = bhB (2) = 0; if the price of the stock goes down at time 1, then
hS (2) = −4/5, hB (2) = 224/405.
b b

Exercise 5.6: E[X] = N (2p − 1).

87
Exercise 5.8: E[R] = − 112 , Var[R] = 7983/12544 ≈ 64%.

Exercise 6.11: ΠY (t) = S(t)Φ(d1 ).

√ √  
Exercise 6.16: ΠY (0) = S0 − S0 Φ(d − σ T ) + B0 Φ(d), where d = T + σ T log BS00 . Moreover
1
√1
2
σ
limσ→+∞ ΠY (0) = S0 + B0 and limσ→0+ ΠY (0) = max(S0 , B0 ).

Exercise A.3: For i = 1, 2 we have

1 eµi T − erT
πi = .
2θ e2µi T (eσi2 T − 1)

167
Appendix D

Additional exercises

The exercises in this appendix are taken from [3] and from old exams.

Chapter 1
For the following exercises assume that the dominance principle holds.

1. Let r > 0. Prove that C(t, S(t), K, T ) → S(t) as T → ∞

2. Prove that ∂K C(t, S(t), K, T ) ≥ − exp(−rτ ), provided the derivative exists
3. Prove that C(t, S(t), K, T ) → 0 as S(t) → 0
4. Prove that ∂s C(t, s, K, T ) − ∂s P (t, S(t), K, T ) = 1
5. Find a constant portfolio consisting of European calls and puts with expiration date
T such that the value of the portfolio at time T equals V (T ) = min(K, |S(T ) − K|)
6. Find a constant portfolio consisting of European calls and puts with expiration date T
such that the value of the portfolio at time T equals V (T ) = min[(S(T ) − K)+ , (L −
S(T ))+ , (L − K)/4], where 0 < K < L

Chapter 3
In the following exercises, {S(t), B(t)}t∈I , I = {1, . . . , N }, is an arbitrage free binomial
market with parameters d < r < u, p ∈ (0, 1). The value of the market parameters is
specified in each single exercise.

1. Assume d = −u, r = u/2 and N = 2 and let g(x) = 1 if x = 0, g(x) = 0 if

x 6= 0. Consider a European derivative with pay-off Y = g(S(2) − S(0)) at time

168
 of maturity T = N = 2. Find ΠY (0) and hS (0). ANSWER: ΠY (0) = 2e−u qu qd ,
−u/2 −qu
hS (0) = hS (1) = eS(0) equd−e−u

2. Let u > r > 0, d = −u and N = 3. A European style derivative with expiration

date T = N = 3 pays the amount 1 if the underlying stock moves always in the
same direction (up or down) and 0 otherwise. Compute ΠY (0). ANSWER: ΠY (0) =
e−3r (qu3 + qd3 )

3. Let N = 2 and er = (eu + ed )/2. Derive the hedging portfolio of a European derivative
paying the amount Y = |S(2)/S(1) − S(1)/S(0)| at time of maturity T = 2

4. (Hard) A European derivative pays the amount Y = N

P
t=1 (S(t)−S(t−1))+ at maturity
1−e−rN
T = N . Find ΠY (0). ANSWER: ΠY (0) = S0 A er −1 , where A = [qu (eu −1)+ +qd (ed −
1)+ ]

5. Let N = 3 and
5 1 64
eu = , ed = , er = 1, S0 = , p ∈ (0, 1).
4 2 25
Consider a European derivative with maturity T = 3 and pay-off
3
1 X 
Y = S(i) − 2 ,
4 i=0
+

which is an example of Asian call option. Compute the price of the derivative at
time t = 0. Compute the probability that the derivative expires in the money and the
probability that a long position in one share of the derivative gives a positive return.
ANSWER: ΠY (0) = 52/75, P(Y > 0) = p, P(R > 0) = p2

Chapter 4
In the following exercises, {S(t), B(t)}t∈I , I = {1, . . . , N }, is an arbitrage free binomial
market with parameters d < r < u, p ∈ (0, 1). The value of the market parameters is
specified in each single exercise.

1. Let N = 2, eu = 47 , ed = 12 and er = 89 . Consider an American put option with strike

K = 3/4 at the maturity time T = 2. Assume S(0) = B(0) = 1. Compute the initial
binomial price of the derivative and the initial position in the hedging portfolio. Verify if
the put-call parity holds at all times. ANSWER: Π b Y (0) = 1/9, hS (0) = hS (1) = −1/5,
hB (1) = hB (0) = 14/45

2. Assume eu = 74 , ed = 21 , S(0) = 1, p = 3/4, er = 9/8.

169
 a) Compute the fair price at t = 0, 1, 2 of an American put with strike K = 3/4 and
maturity T = 2
b) Compute the fair price at t = 0, 1, 2 of a European call with strike K = 3/4 and
maturity T = 2
c) A derivative U gives to its owner the right to convert U at time t = 1 into either
the European call or the American put defined above. Compute the fair price of
U at time t = 0
d) Describe the optimal strategy that the holder of U should follow
e) Compute the expected value at time t = 2 of a portfolio containing one share of U
at time t = 0 and assuming that the American put is not exercised at time t = 1

ANSWER: c) ΠU (0) = 16/27, e) E[V (2)] = 347/256

3. Consider the American option in Exercise 4.4. Assume either that (a) the buyer is not
allowed to exercise at time t = 1 or that (b) the buyer is not allowed to exercise at time
t = 2. In both cases compute the initial price of the option and discuss the difference
with the original American option, where exercise is allowed at all times. REMARK:
The new option where either (a) or (b) is enforced is an example of bermuda option.

Chapter 5
1. Consider a portfolio that is long 1 share of the American put option in Section 4.3.
Assume p = 1/2 and compute the expected relative return of the portfolio. ANSWER:
E[R] = 17/648

2. Repeat the previous exercise for the compound option in Exercise 3.2. ANSWER:
E[R] = 77/16

3. Suppose X, Y are independent random variables with density functions fX , fY respec-

tively. Prova that X + Y has the density function
Z
fX+Y (x) = fX (y)fY (x − y) dy, x ∈ R.
R

4. Use the result in the previous exercise to prove Theorem 5.9.

Chapter 6
In the following exercises, the stock price S(t) is given by a geometric Brownian motion with
instantaneous volatility σ > 0 and instantaneous mean of log-return α.

170
1. Find the Black-Scholes price of a European derivative with pay-off Y = S(T ) + S(T )−1
2
at maturity T > 0. ANSWER: ΠY (t) = S(t) + e(σ −2r)τ S(t)−1
2. Let 0 < a < b. A European derivative pays 1 if S(T ) lies in the interval (a, b) and zero
otherwise. (a) Compute the Black-Scholes price ΠY (t) of the derivative and (b) find
√ σ2
the value of S(t) for which ΠY (0) is maximal. ANSWER (b): S(0) = abe−(r− 2 )τ
3. A European derivative pays 1 if S(T ) > K and −1 otherwise, where K > 0. Determine
σ2
K such that ΠY (0) = 0. ANSWER: K = S(0)e(r− 2
)T

4. Let 0 < L < K. A European style derivative on a stock with maturity T > 0 pays
nothing to its owner when S(T ) > K, while for S(T ) < K it lets the owner choose
between 1 share of the stock and the fixed amount L. (a) Draw the pay-off function
of the derivative. (b) Compute the Black-Scholes price of the derivative. (c) Compute
the number of shares of the stock in the hedging self-financing portfolio. √ ANSWER
(c): hS (t) = ∆(t, S(t)),
√ where ∆(t, x) = Φ(d2 (K)) − Φ(d√ 2 (L)) − φ(d2 (K))/σ τ , where
2
d2 (a) = d1 (a) − σ τ and d1 (a) = (log xa − (r − σ2 )τ )/σ τ .
5. Consider a European derivative with pay-off Y = (S(T ) − S(0))2 /S(T ) at time of ma-
turity T > 0. Compute the Black-Scholes price ΠY (t) and the number of shares of the
2
stock in the hedging portfolio. ANSWER: ΠY (t) = S(t)−2S(0)e−rτ +S(0)2 e(σ −2r)r S(t)−1 ;
2
hS (t) = 1 − S(0)2 e(σ −2r)τ S(t)−2
6. Let a, K, T > 0 be given numbers and consider a European derivative with pay-off
Y = KH(a − S(T )), where H is the Heaviside function. (a) Compute the Black-
Scholes price of the derivative; (b) compute the greeks of the derivative
7. Let v(t, x) be the Black-Scholes price function of a European derivative with pay-off
Y = g(S(T )). Show that v is convex in the variable x; show also that v(t, x) ≥ g(x) if
g(0) = 0. Finally prove that v(t, x) satisfies the Black-Scholes PDE:
σ2 2 2
∂t v + rx∂x v + x ∂x v − rv = 0, 0 ≤ t < T, x > 0,
2
with terminal value v(T, x) = g(x).
8. Let C(t, x) be the Black-Scholes price function of a call option with strike K and
maturity T . Show that
∂ 2C xd1 d2 √
= φ(d 1 ) τ.
∂σ 2 σ
Conclude from this that the function σ → C(t, x) is convex in the interval (0, σ0 ] and
concave in the interval [σ0 , ∞), where
r
2 xerτ
σ0 = | log |
τ K

171
Bibliography

[1] F. Black, M. Scholes: The Pricing of Options and Corporate Liabilities. The Journal of
Political Economy 81, 637–654 (1973)

[2] Z. Bodie, A. Kane, A. J. Marcus: Investments, 5th ed. The McGraw-Hill Primis (2003)

[3] C. Borell: Introduction to the Black-Scholes theory.

[4] J. C. Cox, S. A. Ross, M. Rubinstein: Option pricing: A simplified approach. Journal

of Financial Economics 7, 229 (1979)

[5] W. Feller: An Introduction to Probability Theory and Its Applications (3rd edition).
Wiley series in probability and mathematical statistics (1968)

[6] Shreve, E. S.: Stochastic calculus for finance I. The binomial asset pricing model.
Springer Finance, New York (2004)

[7] Shreve, E. S.: Stochastic calculus for finance II. Continuous-time models Springer Fi-
nance, New York (2004)

172