Math20912 Notes

Dr Peter Johnson
Department of Mathematics
University of Manchester
MATH 20912:
Introduction to Financial
Mathematics
Semester 2 2023-24
Course Outline
MATH20912 Introduction to Financial Mathematics
Dr Peter Johnson
Department of Mathematics
1.129 Alan Turing Building
Office hours: Monday 10:00-12:00
E-mail: peter.johnson-3@manchester.ac.uk
Getting in Contact
• The best way to contact me is to speak to me after the Review Session.
• My ‘office hours’ are a time I will be in my office and available for a chat (in person or
via zoom) if you would like some extra help.
• An alternate way of asking questions about the course is to use the Blackboard Discussion
Board for that week. Here you can post questions anonymously and everyone will benefit
from the answer as many people will be struggling with the same things you are. I am
automatically notified every time someone posts and will try answer queries here as soon as
possible. Students are also encouraged to answer each others questions on the discussion
boards.
• When emailed about the course material, I will likely just post the query and response on
the blackboard discussion board (anonymised) or point you towards a similar question on
the forum - it is not easy to discuss mathematics via email.
i
MATH 20912 Course Outline
Assessment
• Mid-term test in Week 6: contributes 20% of the total mark.
• The mid-term test will consist of an online test.
• Final exam: contributes 80% of the total mark.
• The final exam format will be a traditional invigilated in person exam.
Weekly Schedule
The course will be delivered via the following lectures which cover the material in the Lecture
Notes:
• Monday 9:00: One weekly synchronous (live) lecture/Review Session (Weeks 1-12),
Engineering A Lecture ThB (2A.041).
• Wednesday 9:00: Material for following week uploaded to Blackboard, including asyn-
chronous (recorded) lectures.
• Tutorials: There will also be one tutorial per fortnight which will look at the solutions
to the Exercise Sheets (please see your personal timetable to see which of the 7 tutorials
you have been assigned). The Exercise Sheets form homework and students should make
an attempt at the examples before the tutorial.
Course materials
• Course materials (e.g. Lecture Notes, Example Sheets & Solutions, Past Papers, etc.) will
be uploaded to Blackboard (https://online.manchester.ac.uk).
• Asynchronous lectures will be made available the week before on Blackboard, to be

watched before Monday’s Review Session.
• Lecture Notes are available on Blackboard. It is encouraged that examples in the notes
are attempted while watching the asynchronous lectures.
• Lecture slides complement the notes and provide no extra information (apart from a few
pictures). It would be a waste of paper printing to print them out.
Some Textbooks
All the course material is self-contained but for further/supplementary reading the following two
books are suggested:
• J.C. Hull, “Options, Futures, and Other Derivatives,” 7th Edition, Prentice-Hall, 2008.
ii
• P. Wilmott, S. Howison, and J. Dewynne, “The Mathematics of Financial Derivatives: A

Student Introduction,” Cambridge University Press, 1995.
Blackwell’s Bookshop on University Green have stated previously that they will match (or beat)
the Amazon price of these titles. Hull’s book is extensive and covers all the material in this
course and beyond (it would be of use in third & fourth year courses) and the latest edition is
priced accordingly. Wilmott et al. is more of an introductory text that gives a mathematician’s
view of the material.
iii
Contents
1 Introduction to the Course 1

1.1 Background Economics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Money . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Modelling the Share Price . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 The Wiener process and Modelling Stock Price 9

2.1 Properties of SDEs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.1 Random Walk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Properties of the Wiener Process . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 An Approximation to the Model for the Share Price . . . . . . . . . . . . . . . . 16
3 Log-Normal Distribution 19
3.1 Itô’s Lemma for Continuous Stochastic Variables . . . . . . . . . . . . . . . . . . 19
3.2 Log-Normal Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3 Geometric Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3.1 Is it a Good Model? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4 Derivatives 26
4.1 Simple Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.2.1 Options In Stock Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.2.2 What Actually Happens at Expiry? . . . . . . . . . . . . . . . . . . . . . 33
4.3 How to Price an Option – Payoff Diagrams . . . . . . . . . . . . . . . . . . . . . 33
5 Trading with Portfolios 39

5.1 Portfolio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.2 Trading with Portfolios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2.1 Straddle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2.2 Profits from Portfolios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
iv
5.2.3 Bull Spread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

5.3 Risk-Free Investments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.3.1 Bond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6 Risk and No-Arbitrage 50

6.1 Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.2 No-Arbitrage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
6.2.1 The No-Arbitrage Principle . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6.2.2 No-Arbitrage Principle – Comparing Contracts . . . . . . . . . . . . . . . 53
6.2.3 No-Arbitrage Principle – Equivalent Contracts . . . . . . . . . . . . . . . 54
6.2.4 No-Arbitrage Principle – Return on Risk-Free Contracts . . . . . . . . . . 55
6.3 Put-Call Parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7 Arbitrage 59
7.1 The No-Arbitrage Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
7.2 Bounds on Put and Call Options . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
7.3 Put-Call Parity: Proof by Contradiction . . . . . . . . . . . . . . . . . . . . . . . 60
8 Binomial Trees 66
8.1 One-Step Binomial Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
8.2 Risk-Free Portfolio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
9 Two-Step Binomial Trees 71

9.1 Risk-Neutral Valuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
9.2 Two-Step Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
10 The Multi-step Binomial Model 77

10.1 A Discrete Model for the Share Price . . . . . . . . . . . . . . . . . . . . . . . . . 77
10.2 Binomial Share Price Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
10.3 Binomial Call Option Price Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
10.4 Approximating the Continuous Model . . . . . . . . . . . . . . . . . . . . . . . . 81
11 American Options and Replicating Portfolios 84

11.1 American Option Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
11.2 Replicating Portfolios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
11.2.1 Replicating the One-Step Binomial Tree . . . . . . . . . . . . . . . . . . . 89
12 The Black-Scholes Model 91

12.1 Model Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
12.2 Deriving the Black-Scholes Equation . . . . . . . . . . . . . . . . . . . . . . . . . 92
v
13 Exact Solution to the Black-Scholes Equation 96

13.1 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
13.2 Solution for a Call Option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
14 The “Greeks” 102

14.1 Calculating Delta . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
14.2 “Greeks” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
15 More on Replicating Portfolios 109

15.1 Replicating Portfolios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
15.2 Static Hedging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
15.3 Dynamic Hedging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
16 Extensions to the Black-Scholes Model 114

16.1 Dividends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
16.2 Options on Dividend Paying Shares . . . . . . . . . . . . . . . . . . . . . . . . . . 116
16.3 Early Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
17 Bonds and Interest Rates 121

17.1 Time Varying Interest Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
17.2 Zero-Coupon Bonds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
17.3 Coupon Bonds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
18 Term Structure of Interest Rates 126

18.1 Yield . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
18.2 Term Structure of Interest Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
18.3 Default Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
19 Exotic Options 131

19.1 Asian Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
19.2 Black-Scholes for Asian Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
vi
Chapter 1
Introduction to the Course
1.1 Background Economics

Financial Contract is a written agreement between two parties to exchange payments
according to some specified criteria. The two parties are normally called the holder and seller.
Contract Holder is normally the buyer, who pays money at the beginning in exchange
for receiving some payments at a later date. Because payments may not always be positive, a
contract could be free to enter at the start or even one where you pay to hold it.
Contract Seller is the opposite position to the holder, which normally means they receive
money at the beginning in exchange for giving out some payments at a later date.
Market is the collective term for all participants, both buyers and sellers, that are engaged in
trading financial contracts. It is used as a noun to describe the average, prevailing, or collective
belief of all participants, for example we might say “the market believes that the price will go
up.” Underlying everything in this course are economic assumptions, about the behaviour of the
market, which make the mathematical models work.
Exchange is where most financial contracts are bought and sold. The exchange acts as the
third party in all transactions. They do not set the prices of things, they merely allow the
buyers and sellers to meet or advertise products. They also perform the function of recording all
transactions and making that knowledge public, so that all participants in the market can see
the price at which things are sold.
Example 1.1. Draw up a contract to sell a phone for a fixed price at some future date.
1
MATH 20912 Chapter 1
Solution 1.1.
See figure 1.1.
Share or stock (can be used interchangeably) refers to the financial contract in which a
company sells a percentage of the ownership of the company to the holder, which entitles the
holder to a percentage share of all future profits. Stocks can be used to refer to contracts with
several different companies, while shares normally refers to a set of contracts with just one
company. This is why it is called the stock exchange (not the share exchange). We will mostly
use the term share from now on, as we are primarily concerned with models of the share price
of one company at a time.
Stock Market is the collective term for all participants buying and selling stocks (shares).
Stock Exchange is where stocks (shares) are bought and sold.
Bond is a debt guarantee in which the holder pays now to receive a guaranteed (in as much
as anything can ever be guaranteed) fixed payment in the future.
Bond Market is the collective term for participants buying and selling bonds and other
debt securities.
Asset implies ownership or property, so sometimes stocks and shares may be called assets
because they confer ownership of the company.
Price is the amount of money required in exchange to buy or sell a contract. You should note
that price is not necessarily the same as value, as the second term tends to incorporate more
than just numbers (see below).
Market Price is the highest (or lowest) amount at which at least one participant in the
market is willing to buy (or sell) a contract. We assume throughout this course that a unique
market price exists and that it is the same price to buy and to sell.
Value is a measure of worth to an individual. In almost any context it is extremely subjective,

and the idea of something being subjective does not sit well with mathematics. You might
note that the economics department in Manchester is considered a Social Science rather than
a Physical Science, and there are some who believe that we should not mix the two. What we
learn in this course are the techniques that have enabled academics and the finance industry
2
seller
Boris Johnson will underlying

asset
holder
sell his phone
to ????? for
£300 on 5th May expiry
date
2020. exercise
price
Signed
.......
....... both signatures
for valid contract
Figure 1.1: An example financial contract.
3
to remove the subjectivity (or risk; we will talk later about the value of risk) and develop a
framework under which complex financial contracts can be valued. Because certain economic
assumptions (which we will not go into here) apply to contracts sold in markets, it follows that
the value under that framework must be the same as the market price.
1.2 Money
Money is the circulating medium of exchange as secured by the government (and hence its
citizens). The most important thing we need to know about money is that it is engineered to
be inflationary, which means that the government strives to ensure that the value of the money
in the future will be worth less than it is now. This is because a government typically wishes
to encourage its citizens to spend, invest, and take part in the economy, and this inflationary
property is delivered through fiscal policy (which is something we won’t touch on in this course).
One consequence of the fact that money is inflationary is that banks will pay customers to deposit
money in bank accounts, and that payment is called the interest rate. This means that, as an
agent in the economy, I know that
£10 now > a promise of £10 in the future
because
£10 in the future plus interest > a promise of £10 in the future.
Again, because of the subjectivity involved in valuing sums of money received in the future (even
if guaranteed) we must compare apples with apples. So if we want to value a payment from a
contract in the future, we can only compare it to money in the future. Therefore we need to
know how money changes with time.
Interest rate is a payment made in exchange for being in debt to another party. Once you
deposit money with a bank they owe you the money back, so they are in debt to you and will
pay you interest in return. The rate part refers to the passage of time, i.e. interest is generally
expressed as a % payment per annum.
Time value of money we define the time value of money using the interest rate. There
are several ways in which interest rate may be paid.
• Simple Interest Rate If there is a single payment made at time t, then we call it a
simple interest rate. The value of an investment V (t) at time t is
V (t) = P × (1 + rt)
where V (0) = P is the initial investment, r is the interest rate and t is the current time.
4
• Compound Interest Rate If there are multiple payments made at fixed intervals over
the life of the investment, then we call it a compound interest rate. The value of an
investment V (t) at (integer) time t after mt payments are made is
r mt
V (t) = P × 1 +
m
where P is the initial investment, r is the interest rate, m is the number of payments per
year, and t is the current time (integer years).
• Continuously Compounded Interest Rate For many reasons (mainly practical) it is

much easier to deal with the continuously compounded interest rate rather than a discrete
set of payments. In the continuously compounded interest rate, we let the number of
z
payments per year increase (m → ∞) and, since e = limz→∞ 1 + z1 , we obtain the
formula
V (t) = P ert
where P is the initial investment, r is the interest rate, and t is the current time. In fact,
in real life, payments are always made discretely at fixed intervals, but the interest rates
can easily be adjusted so that they match up with real investments.
Example 1.2. Write down the value of an initial investment of £100 at t = 2 years with:
• a simple interest rate with r = 0.05 paid at t = 2
• a compound interest rate with r = 0.05 in two payments per year
• a continuously compounded interest rate with r = 0.05
Solution 1.2.
• Simple Interest Rate P = 100 is the initial investment, r = 0.05 is the interest rate,
and t = 2 is the time of payment. Then
V (t) = P × (1 + rt) = 100 × (1 + 0.05 × 2) = 110 .
• Compound Interest Rate Since P = 100 is the initial investment, r = 0.05 is the
interest rate, m = 2 is the number of payments per year, and t = 2 is the final date, we
have 2×2
0.05
V (t) = 100 × 1 + = 100 × 1.0254 = 110.38 .
2
5
• Continuously Compounded Interest Rate P = 100 is the initial investment, r = 0.05

is the interest rate, and t = 2 is the time of payment. Then
V (t) = 100 e0.05×2 = 110.52 .
Why buy shares? In a capitalist society, ownership of shares is encouraged through tax
breaks and other incentives. The general idea is that ownership of companies can be spread out
through the population, and the best performing companies will be able to get investment to
expand their operations. Most of you will end up owning shares, probably through investments
made on your behalf by pension companies. Most investors will look to the return on the
investment, which is how much money they make compared to the money they had to invest at
the start.
Return on investment (or return for short) will always be defined as the relative return
which can be calculated as
value at expiry − initial investment
return = .
initial investment
Example 1.3. Suppose an investor has £1000 to spend. They see Apple shares are trading at
£100 and they think it will go up to £125 by the end of the year. They also see Google shares
are trading at £50 and they think it will go up to £65 by the end of the year. Which shares
should they invest in?
Solution 1.3.
First, looking at the absolute increases shows that Apple is expected to have a better per-
formance (£25 vs £15), so should the investor buy Apple shares?
The obvious answer is no. The investor can use the £1000 to buy 10 Apple shares to get an
expected profit of £250, but if they invest in Google instead they can buy 20 shares to get an
expected profit of £300. So a much better metric to look at when comparing the performance of
shares is the relative increase in price. For Apple we have
125 − 100
relative increase = = 0.25
100
and for Google we have
65 − 50
relative increase = = 0.3
50
which shows clearly that Google is the better investment.
6
1.3 Modelling the Share Price

We let St denote the market price of a share at time t. There is an abundance of data available
for the historical prices of shares. Obviously the models that have been developed over the last
hundred years or so have been successful because they match in some way the properties of the
data. In a market there are many things which drive the changing of price, but they can be
broadly split into two parts.
The Deterministic Part contains all known or expected trends. These are long term
trends that can be estimated either by analyzing market data, or by using fundamental economic
knowledge and insight.
The Stochastic Part this contains all the uncertain parts. Uncertainty about what may
happen in the future plays a part, as even events such as extreme weather can have an effect. But
also the inherent complexity of the interactions between thousands and thousands of participants.
For example, the holder of some shares might lose his/her job and need to sell the shares quickly,
which will cause a short dip in the price. It is this second effect (market interactions) that drives
the price over the shortest time scales, and is what we really try to capture in this course.
A Stochastic Model for Shares

We model the return on a share as
dS
= µdt + σdW. (1.1)
S
Here µ is a measure of the deterministic expected rate of growth of the share price. In general,
µ = µ(S, t). In simple models µ is taken to be constant (e.g. µ = 0.1 yr−1 = 10% yr−1 ). The last
term σdW describes the stochastic change in the share price, where dW stands for
∆W = W (t + ∆t) − W (t)
as ∆t → 0. We will describe what W (t) is in the next chapter; it is a special random process
1 1
called the Wiener process. We call σ the volatility (e.g. σ = 0.2 yr− 2 = 20% yr− 2 ).
Volatility is a statistical measure of the standard deviation of returns for a share price. It
1
is normally given as a percentage per year 2 since it is relative to the current price of the share.
Example 1.4. What does this model mean in words? What should be the properties of the
Wiener process to make it consistent with the real data?
7
Solution 1.4.
In words, our model (1.1) can be written as
dS
relative change in price = = deterministic + stochastic.
S
Observations from the market mean that we need the following properties for the stochastic
part:
• Zero correlation – price change tomorrow is not related in any way to what happened
today, yesterday, or the day before;
• Growth in time – the uncertainty is expected to grow in time, so that the possible range
of prices a long way in the future is much greater than the range over the next few days;
• Normally distributed, over a long run.
It just so happens that all of these properties are satisfied by a special type of process, the Wiener
process.
8
Chapter 2
The Wiener process and

Modelling Stock Price
Background Material
Throughout this course we adopt the following notation
• ∆t is a small but finite change in time;
• ∆S, ∆V , etc. are changes in quantities that depend on t over a small finite change in time
∆t;
• dt represents an infinitesimally small change in time that can be used to build integrals
or differential equations;
• dS, dV , etc. are the corresponding changes in quantities over an infinitesimally small
change in time.
So if you see ∆ in front of a variable then we are talking about discrete approximations, whereas
a d will mean we are talking about the continuous limit.
Note: In later chapters when we look at trading with portfolios ∆ will also be used to describe
the number of shares held and so ∆S will mean we hold ∆ of shares S. This double use of ∆ is
unfortunate but common place - it will be obvious due to the context which is meant.
Normal Distributions
A normal distribution is a continuous probability distribution with the probability density

1 (x − µ)2
f (x; µ, σ) = √ exp − (2.1)
σ 2π 2σ 2
9
where µ is the mean and σ 2 is the variance of the distribution. If a random variable Y is drawn
from a normal distribution with mean µ and variance σ 2 , then we write
Y ∼ N (µ, σ 2 ) .
Adding normal distributions together

If Y1 and Y2 are independent random variables that are both normally distributed, say
Y1 ∼ N (µY1 , σY2 2 ) ,
Y2 ∼ N (µY2 , σY2 2 ) ,
then if we define Z as their sum, i.e.

Z = Y1 + Y2 ,
we have
Z ∼ N (µY1 + µY2 , σY2 1 + σY2 2 ) .
The next result we might need to know in this course is what happens when we construct a
new random variable by multiplying and/or adding constants. Suppose that
Z = a + bX
where
X ∼ N (0, 1)
i.e. X is given by the standard normal distribution, with zero mean and unit variance. Then it
follows that
Z ∼ N (a, b2 ).
2.1 Properties of SDEs

In the first chapter of the course, we proposed a continuous model for the share price of the form
dS
= µdt + σdW (2.2)
S
where the first term on the right-hand side captures the average growth rate of the share price,
or the trend, and the second term captures the unpredictable nature of share price. As a result
of the many observations of share prices over the past 100+ years, it has been found that one of
the best ways to model the dW term is to use a random walk model. It is easiest to explain a
random walk model in discrete steps, and we will do so next. In fact, the actual model we use is
the continuous version, called a Wiener process, which has special properties that make it easier
to work with.
10
2.1.1 Random Walk

In general, to create a random walk we take successive draws from a random distribution and
add them together. In our example the random distribution will be defined as ∆W , and we will
define Wk as the value of the random walk at the k-th step. To create our special random walk,
we take successive draws from a normal distribution with mean zero and variance ∆t, and add
each draw to our previous total. If we are working in discrete time, and we observe the process
Wk at the k-th step, then
Wk+1 = Wk + ∆W (2.3)
where W0 = 0 and
∆W ∼ N (0, ∆t) .
Example 2.1. Sketch this random walk (2.3) with ∆t = 1. You can use www.random.org to
generate random numbers.
Solution 2.1.
See figure 2.1.
Now if we write the discrete random walk (2.3) as

X
Wk = ∆W
k
it is trivial to show that

Wk ∼ N (0, k∆t).
Then if we write t = k∆t (so t is the time elapsed since the start), we have
W (t) ∼ N (0, t).
These results also hold for the continuous limit, but proving it is beyond the scope of this course.
Therefore in the limit ∆t → 0 we write
Z t
W (t) = dW (2.4)
0
where
dW ∼ N (0, dt).
11
W(t)
MATH 20912
W1
�W �W
W0=0
t
�t W3
W2
12
W4
At each time the next jump is an independent random draw from
the normal distribution.
�W ~ N(0,�t)
Here
W4 = �W + �W + �W + �W
so
Figure 2.1: Constructing a discrete random walk.

W4 ~ N(0, 12 + 12 + 12 + 12) ~ N(0,4)
Chapter 2
2.2 Properties of the Wiener Process

Wiener Process One way to describe the Wiener process W (t) is as the continuous limit
of a random walk process, e.g. given the random walk as defined above, we take the limit ∆t → 0.
It has the following properties
• W (0) = 0;
• W (t) has independent increments: if u ≤ v ≤ s ≤ t, then W (t) − W (s) and W (v) − W (u)
are independent;
• W (s + t) − W (s) ∼ N (0, t);
• W (t) has continuous paths, i.e. W (t) is continuous in t.
Example 2.2. Show that the following results hold:
E[W ] = 0 ; (2.5)
E[W 2 ] = t ; (2.6)
E[∆W ] = 0 ; (2.7)
E[(∆W )2 ] = ∆t ; (2.8)
1
∆W = X (∆t) 2 , where X ∼ N (0, 1) . (2.9)
Solution 2.2.
The first result (2.5) follows from the definition that W (s + t) − W (s) ∼ N (0, t). Simply put
s = 0, then
W (t) − W (0) ∼ N (0, t).
and since W (0) = 0 we get

W (t) ∼ N (0, t).
Therefore
E[W (t)] = 0,
and
var[W (t)] = t.
The second result (2.6) follows on from the standard definition of the variance
2
var[W ] = E W 2 − (E[W ])
13
Using the above we have

2
t = E W 2 − (0)
so

E W 2 = t.
Now if the increment ∆W = W (t+∆t)−W (t) is some discrete approximation to the process,
we have
∆W = W (t + ∆t) − W (t) ∼ N (0, ∆t).
This gives us the third result (2.7)
E[∆W ] = 0
and the fourth result (2.8)
var[∆W ] = ∆t
2
=⇒ E ∆W 2 − (E[∆W ]) = ∆t,
=⇒ E[(∆W )2 ] = ∆t.
Let
∆W
X= .
(∆t)1/2
We found above that ∆W ∼ N (0, ∆t), so by the properties of normal variables, we deduce the
fifth result (2.9)
X ∼ N (0, 1) .
Probability Density Function for the Wiener Process

Since W (t) ∼ N (0, t), the probability density function for a Wiener process is given by
2
1 y
f y; 0, t1/2 = √ exp − , (2.10)
2πt 2t
so the probability that W (t) is contained within some interval [a, b] at a future time t is given by
Z b
P(a ≤ W (t) ≤ b) = f y; 0, t1/2 dy. (2.11)
a
Example 2.3. Draw the probability distribution for the Wiener process. How does this distri-
bution change over time?
Solution 2.3.
See figure 2.2.
14
f(y;0,1)
Standard
MATH 20912
Normal W(t)
t1/2
N(0,1)
-1 1 y
s t
15
f(y;0,t1/2)
Wiener
Process
N(0,t)
-t1/2
Figure 2.2: The probability distribution for a Wiener Process.

-t1/2 t1/2
Chapter 2
2.3 An Approximation to the Model for the Share

Price
Recall our continuous model for the share price (1.1):
dS = µSdt + σSdW.
Over a small (finite) interval of time ∆t, we can write
∆S ≈ µS∆t + σS∆W ,
and substituting in the result (2.9) we obtain

1
∆S ≈ µS∆t + σS(∆t) 2 X , (2.12)
where X ∼ N (0, 1). So using the basic properties of a normal distribution, we know that

∆S ∼ N µS∆t, σ 2 S 2 ∆t (2.13)
and
S(t + ∆t) = S + ∆S ∼ N (S + µS∆t, σ 2 S 2 ∆t) . (2.14)
Note that although S(t + ∆t) is normal, we can’t simply add another ∆S and deduce the
distribution of S(t + 2∆t). Although the increments ∆W are all independent random draws the
∆S at the next step depends on the value of S(t + ∆t), which is itself drawn from a distribution.
The only way to get around this is by developing some stochastic calculus that we introduce in
Chapter 3.
Example 2.4. Consider a share that has volatility 30% and provides an expected return of 15%
p.a. (per annum; per year). Find an expression for the increase in share price after one week if
the initial share price is 100.
Solution 2.4.
We have
• σ = 0.3
• µ = 0.15
• ∆t = one week ≈ 1
52
• ∆S is the change in share price in this time
• S(t = 0) = 100
16
We insert these figures in (2.12)
∆S = µS∆t + σS(∆t)1/2 X
where X ∼ N (0, 1). We find

1
∆S = 0.15 × 100 × + 0.3 × 100 × (1/52)1/2 X .
52
Answer: ∆S ≈ 0.2885 + 4.160 X.
Note: 0.2885 is the expected increase, and 4.160 is the standard deviation of the increase.
Example 2.5. Given the result from Example 2.4, what is the distribution of the share price in
that case after one week?
Solution 2.5.
We have
S(∆t) − S(0) = ∆S
S(∆t) = S(0) + ∆S
S(∆t) ≈ 100.2885 + 4.160X ,
or equivalently,
S(∆t) ∼ N (100.2885, 17.31) .
∆S
Example 2.6. Show that the return S is normally distributed with mean µ∆t and variance
2
σ ∆t.
Solution 2.6.
We have
∆S = µS∆t + σS(∆t)1/2 X
so
∆S
= µ∆t + σ(∆t)1/2 X .
S
This is a linear function of a standard normal random variable X ∼ N (0, 1), so ∆S/S is
itself normal. Its mean is
∆S
E = µ∆t + 0,
S
17
and its variance is

∆S
var = σ 2 ∆t .
S
Thus overall we have
∆S
∼ N µ∆t, σ 2 ∆t .
S
18
Chapter 3
Log-Normal Distribution
3.1 Itô’s Lemma for Continuous Stochastic Vari-

ables
Mathematical Finance is about pricing (or valuing) financial contracts, and in particular those
contracts which depend on the value of another contract. Now we are going to write down a
result which tells us how a function that depends on a random variable changes in time.
We assume that the function f (S, t) is a smooth1 function of S and t. Then, if S satisfies
the stochastic differential equation
dS = µSdt + σSdW ,
Itô’s Lemma states that the distribution of the change in f is given by

∂f ∂f 1 2 2 ∂2f ∂f
df = + µS + σ S dt + σS dW (3.1)
∂t ∂S 2 ∂S 2 ∂S
where
df = f (S + dS, t + dt) − f (S, t).
There are two things to note here. First, the resulting df is a distribution characterized by
the dW term. Because the function f depends on something that is random and unpredictable,
the value of the function itself will also be random and unpredictable in the future. Secondly,
df includes a quadratic term in S; where did this come from? If we consider a Taylor series
approximation, then usually the quadratic term dS 2 is much smaller than the linear terms, as
dS 2 dS. However, in stochastic calculus we found that
dS 2 = O(dt) ,
1
Sufficiently differentiable.
19
which comes from the earlier result that E[W 2 ] = t. Expanding the above we have
dS 2 = (µSdt + σSdW )(µSdt + σSdW )

= σ 2 S 2 dW 2 + o(dt)
≈ σ 2 S 2 dt ,
so we can now see where the quadratic term in S has come from.
You are not required to know how to derive this result for the purposes of this course, only
how to apply Itô’s Lemma in practical situations. Later on in the course we will need to apply
it in order to derive the value of contracts.
Example 3.1. Find the stochastic differential equation satisfied by f = S 2 .
Solution 3.1.
First we need to find all the partial derivatives of f (S, t) and then substitute them into Itô’s
Lemma. We have
∂f ∂2f ∂f
= 2S , 2
= 2, = 0.
∂S ∂S ∂t
Substituting into Itô’s Lemma, we find

1
df = 0 + µS(2S) + σ 2 S 2 (2) dt + σS(2S)dW
2
2
=⇒ df = 2µ + σ S dt + 2σS 2 dW .
2
To obtain the required SDE, we replace S 2 with f :
df = (2µ + σ 2 )f dt + 2σf dW .
Note that we prefer to write the solutions to these problems in terms of the original function
f rather than keeping (for example) the S 2 in here.
3.2 Log-Normal Distribution

Example 3.2. Show that the stochastic differential equation (SDE) for f = ln S is:

σ2
ln (S(t)) = ln (S0 ) + µ − t + σW (t) . (3.2)
2
20
Hint: remember that adding together normal distributions results in another normal distri-
bution. This means that SDE’s with constant coefficients can be integrated.
Solution 3.2.
We apply Itô’s Lemma to f (S, t) = ln S. The required partial derivatives are

∂f 1 ∂2f 1 ∂f
= , =− 2, = 0.
∂S S ∂S 2 S ∂t
Substitute these into Itô’s Lemma to obtain

1
df = 0 + µS(1/S) + σ 2 S 2 (−1/S 2 ) dt + σS(1/S)dW
2
and note that S cancels throughout:

σ2
df = µ− dt + σdW .
2
This is an SDE with constant coefficients, assuming that µ and σ are constant.
Therefore we may integrate

Z Z Z
σ2
df = µ− dt + σ dW
2
and using time limits from 0 to t, we have

σ2
f (S(t), t) − f (S(0), 0) = µ− t + σ(W (t) − W (0))
2
and since W (0) = 0 we may write

σ2
f (S(t), t) − f (S(0), 0) = µ− t + σW (t) .
2
Now we use the fact that f (t) = ln(S(t)) to write
f (S(t), t) − f (S(0), 0) = ln (S(t)) − ln (S(0)) ,
and hence
σ2
ln (S(t)) = ln (S0 ) + µ − t + σW (t) ,
2
where S0 = S(0) is the initial share price.
From (3.2), we can deduce that the

logarithm of the share price ln (S(t)) is normally dis-
σ2
tributed, with mean ln (S0 ) + µ − 2 t, and variance σ 2 t, i.e.

σ2
ln (S(t)) ∼ N ln (S0 ) + µ − t, σ 2 t . (3.3)
2
21
Example 3.3. Consider a share with an initial price of 40, an expected return of 16%, and a
volatility of 20%. Find the probability distribution of ln S after six months.
Solution 3.3.
We have
• S0 = 40,
• µ = 0.16,
• σ = 0.2,
• t = 0.5 (note we measure time in years unless stated otherwise).
Now we need to find the distribution of ln (S(t)). First we get
E [ln (S(t))] = ln S0 + (µ − σ 2 /2)t

= ln 40 + (0.16 − 0.5 × (0.2)2 ) × 0.5
≈ 3.759 ,
and
var [ln (S(t))] = σ 2 t

= (0.2)2 × 0.5
= 0.02 .
So altogether we have the distribution
ln (S(t)) ∼ N (3.759, 0.02) .
Note that this solution works over ANY time period, whereas the normal approximation in
Chapter 2 only works for t 1.
3.3 Geometric Brownian Motion

Definition A stochastic process S(t) is said to follow a Geometric Brownian Motion if
dS = µSdt + σSdW ,
where W is a Wiener process, and µ and σ are constants.

Given that our share price model is a GBM, we have some properties and results that can
be useful:
22
1. We have an exact formula for the share price at time t given by

2
S(t) = S0 e(µ−σ /2)t+σW (t)
.
This comes from taking the exponential of equation (3.2).

2. There is a closed form expression for the probability density function of S at time t given
by
2 !
1 ln s − ln(S0 ) − (µ − σ 2 /2)t
fSt (s; µ, σ, t) = √ exp − .
sσ 2πt 2σ 2 t
The two results stated here are particularly important when it comes to evaluating financial
contracts depending on S using numerical methods. The first result is particularly useful when
generating random samples for S at some future time t, as we only need a single random number
to be generated for each path. The second result is used in more complex numerical methods
that take advantage of the fact that expectations can be evaluated as some integral involving the
probability density function.
Example 3.4. Write down a formula for S(t) in terms of the standard normal distribution.
Solution 3.4.
From (3.2) we have

S(t) 1 2
ln = µ − σ t + σW (t)
S0 2
and now take exponential of both sides

S(t) 1 2
= exp µ − σ t + σW (t)
S0 2
and so
1
S(t) = S0 exp µ − σ 2 t + σW (t) .
2
Given that W (t) ∼ N (0, t), we can write
W (t) = t1/2 X
where X ∼ N (0, 1). Hence we can write

1
S(t) = S0 exp µ − σ 2 t + σt1/2 X
2
where X has a standard normal distribution.
Example 3.5. Try to draw a sketch of the log-normal distribution for S(t).
Solution 3.5.
See Figure 3.1.
23
A sketch of the log-normal distribution has the following
MATH 20912
properties:
looks roughly normal in the middle
24
Skew
mode mean S(t)
Figure 3.1: Drawing a log-normal distribution.

Probability of S(t)=0 is zero
Chapter 3
3.3.1 Is it a Good Model?

How good is Geometric Brownian Motion as a model of the share price? Well, it certainly has
a lot of the properties that we would like to see in a good model of the share price, such as
independent increments, and it works very well for the most part, but there are two main flaws.
They are:
• the observed volatility σ in real markets is not constant in time;
• there are often large non-normal fluctuations in the price, when external events cause a
large number of participants to want to either buy or sell at the same time.
Trying to come up with models that can overcome these flaws has been the focus of much
research over the years. However, in light of recent market failures, academics and practitioners
are beginning to question whether any simple model can ever really capture everything about
the market.
25
Chapter 4
Derivatives
4.1 Simple Derivatives

Financial contracts have been written and exchanged between parties for thousands of years,
even back to ancient Greece. Before shares were traded in a market, the most common type of
financial contract would be one involving a commodity, such as food (crops, livestock) or fuel
(wood/coal/oil), in which the holder would agree a price now for which they would buy some
amount of the commodity at some future time.
The trading of such contracts has, over time, become more and more popular, not just
because there is money to be made, but also because they can have some real positive social
benefits. For example, imagine that farmers are able to “lock in” the price of their crops now
before they start growing the crops; this means that the farmers can be more certain about the
future, perhaps encouraging them to invest in new equipment to increase yields. In order for
farmers to be more certain about the prices they receive, somebody has to be on the other side
of the contract, which is a less certain position.
Take for instance the following scenario: a Spanish farmer is able to forward sell 100 tonnes
of oranges to Tesco that must be delivered in June 2020 for e600 per tonne. Is this a good deal
or not? Well, let us consider two extreme cases:
• The weather is good, crop yields are high, and as a result the price of oranges in June
drops much lower than e600 per tonne. In this case, the farmer is happy, and Tesco is
sad; Tesco would have been better off buying oranges from the market in June.
• The weather is bad, crop yields are low, and as a result the price of oranges in June rises
much higher than e600 per tonne. In this case, the farmer is sad, and Tesco is happy;
the farmer would have been better off selling the oranges on the market in June.
26
So it seems as though there is always one side that loses – so why would either side agree
to this? Well, there are benefits to both parties from the certainty about the price to be paid,
and therefore cash flows in the future. And to think about how both sides agree on a price, you
might consider that it depends on the probability of the price in June being above or below
e600 per tonne.
This type of contract is called a Derivative – it is a contract whose value depends on the
values of other underlying variables or assets. Other names are: financial derivative, derivative
security, derivative product. A share option, for example, is a derivative whose value is dependent
on the share price. Examples of such contracts are forward contracts, futures, options, swaps,
etc. In the example of Tesco and the farmer, they trade a forward contract and the underlying
asset is oranges (or, the capacity to grow oranges).
Example 4.1. Draw a cartoon to illustrate the buying and selling of this contract:
“Boris wishes to sell his phone for £300 on 5th May 2020.”
Solution 4.1.
See Figure 4.1.
Example 4.2. Now on 5th May 2020, the same model of phone is trading on the market at
either £250 or £350. Who is happy in each case?
Let E be the exercise price (300 here), and S the market price of this model of phone.
Solution 4.2.
See Figure 4.2.
Example 4.3. Now the holder wishes to include an option to cancel the purchase. What happens
when this new contract is agreed, in the two scenarios above?
Say the fee for this cancellation option is £10 (non-refundable), paid upfront.
Solution 4.3.
See Figure 4.3.
27
Boris wishes to sell Someone is willing to buy

his phone... the phone on 5th May.
Boris Boris
They sign a contract On 5th May they settle the
agreeing to the sale contract
Boris Boris
Figure 4.1: A cartoon to illustrate the buying and selling of a contract.
28
Agree on contract, F, Exchange on 5th May

with price E=300
Boris Boris
S-F F 300 S - 300
S=250 S=350
Boris Boris
300 S - 300 = -50 300 S - 300 = +50
Figure 4.2: A cartoon to illustrate different scenarios at expiry.
29
Agree on contract C,
with price E=300, and
Exchange on 5th May
cancellation option 10
? ?
? ?
Boris Boris
S - C + 10 C - 10 ? + 10 ? - 10
S=250 S=350
Boris Boris
S + 10 -10 E+10= 310 S-300-10=40
Figure 4.3: A cartoon to illustrate different scenarios at expiry when a cancellation

option is included.
30
4.2 Options
Options have been written into financial contracts for almost as long as they have been traded.
In our example of the farmer and Tesco, the farmer might be well-placed to mitigate bad weather
and would therefore like to include an option to cancel the contract with Tesco in order to get a
better price on the market in that scenario. If such an option is included in the contract, and it
is the farmer that makes the decision on whether to sell or not, then the farmer will be the one
buying the option from Tesco even though the farmer is selling the oranges. In order to agree
to the cancellation option, Tesco would most likely want the farmer to pay a (non-refundable)
upfront fee – the option price.
Options are very attractive to investors, both for speculation and for hedging.
Option Holder The option holder (buyer) is the one that is able to make a decision – to
exercise or not, to buy or not, to sell or not – and they pay the option seller a fee for the privilege.
Option Writer The option writer (seller) has a decision forced upon them at some future
time, so they receive some compensation in return.
Option Value is Positive By definition, because the option holder (buyer) is the one
that is able to make a decision, and as that usually includes doing nothing, the value of an option
is always at least zero.
4.2.1 Options In Stock Markets

All of the options that we describe in this course will be written on share prices. Because the
value of the option contracts will be derived from the price of the underlying share, the contract
is called a derivatives contract. Any financial contract whose price is derived from the value of
another fits into this category.
The most commonly traded options are call and put options.
European Call Option

The holder of a call option has the right (but not the obligation) to buy the underlying share
(sometimes we call this the asset) at a pre-agreed exercise price E on the specified expiry date
T of the contract. Mathematically, we will usually denote the price at time t of a call option by
C(S, t). Sometimes we may include some of the other parameters, such as an exercise price of
E = 10 and an expiry date T = 1 like this C(S, t; E = 10, T = 1).
The formula to derive the value of the contract on the expiry date (when t = T ) is
C(S, T ) = max(S − E, 0) ,
31
or equivalently, (
0 if S ≤ E ,
C(S, T ) =
S−E if S > E .
European Put Option

The holder of a put option can decide to sell or not sell at a pre-agreed exercise price on the
specified expiry date of the contract. Mathematically, we will usually denote the price at time t
of a put option by P (S, t). The formula to derive the value of the contract on the expiry date
(when t = T ) is
P (S, T ) = max(E − S, 0) ,
or equivalently, (
E−S if S ≤ E ,
P (S, T ) =
0 if S > E .
Example 4.4. Consider a three-month European call option on a BP share with exercise
price E = 15. If you enter into this contract, you have the right (but not the obligation) to buy
one BP share for E = 15 in three months time (at T = 0.25).
What happens at expiry if:
1. The share price is £25?
2. The share price is £5?
Solution 4.4.
1. If the share price is £25, then the price at which you can buy one share (E = 15) is lower
than the price at which you can sell one share (S = 25), so we can make an instant profit.
So we should exercise the option and make S − E = 25 − 15 = 10, i.e. £10 profit.
Using the notation described earlier, we have C(25, 0.25) = max(25 − 15, 0) = 10.
2. If the share price is £5, then the price at which you can buy one share (E = 15) is higher
than the price at which you can sell one share (S = 5), so we cannot make an instant
profit. It makes no sense to exercise the option – so we should do nothing.
Using the notation described earlier, we have C(5, 0.25) = max(5 − 15, 0) = 0.
Exercise Price This can be used interchangeably with strike price and is denoted E. The
pre-agreed strike price or exercise price E is written into the contract, and therefore the value of
the option will depend on this price.
32
Underlying Share/Asset Price The share price S may also be called the underlying
share price or the underlying asset price (the second is more generic).
Expiry Date/Maturity Date The final termination date as specified in the contract,
i.e. the date at which the option ends or must be exercised. The words expiry (or expiration) and
maturity are often used interchangeably. People tend to use expiry when talking about options,
and maturity when talking about bonds, but both can be used.
4.2.2 What Actually Happens at Expiry?

If you hold an option, and the option is going to make a profit, there are two possible ways you
could go:
• wait until the option expiry date and then exercise the option; if you have a call option,
you will need to pay E to buy the share. If you have a put option you will receive E in
exchange for selling the share (if you don’t currently hold a share then you will need to
go out and buy it on the market first!).
• close out your position (see below); i.e. a call option holder will sell an identical call option
and receive the profit on sale of the option.
The second way might be more attractive for a couple of reasons. For instance you might not
actually want the share, nor have the money E to pay for it. By closing out your position, you
can trade in options without ever owning a share! This is very popular and you will see a large
amount of trading in the last few days of an option’s lifetime, as people close out their positions.
Position The term position is used here to denote the amount of financial contracts held,
and “closing out a position” means restore that number to zero.
4.3 How to Price an Option – Payoff Diagrams

We need to look at the payoff to calculate the price. The price (or the value) of something does
NOT depend on how much was paid for it. The price is what someone else is willing to pay
for it, and they will only be interested in what the future profits might be.
In our earlier example, the farmer will buy a European put option from Tesco. This means
the farmer has an option to sell oranges at the pre-agreed exercise price of e600 per tonne in
June 2020. If for example, the market price of oranges drops down to e400 per tonne in June,
then the farmer will agree to sell the oranges to Tesco (exercising the option) and make a profit of
e600 − e400 = e200 per tonne above the market price. If however, the market price of oranges
33
rises to e800 per tonne in June, the farmer does not exercise the option (neither making nor
receiving payments to/from Tesco) and is free to sell the oranges on the market.
To know what price Tesco will charge the farmer for the cancellation option, we must look
at the profit/loss for the holder/writer of the option under different scenarios.
Payoff Diagrams The payoff diagram of a financial contract illustrates the cash flow in all
possible scenarios on the expiry date, i.e. for all possible share prices S at time t = T .
Example 4.5. Draw the payoff diagrams for a call option and a put option.
Solution 4.5.
See Figure 4.4 and Figure 4.5.
Profit This is not to be confused with payoff which only considers what happens at expiry.
Profit must take account of the price paid (or the amount received if we are selling) on the day
the option was bought or sold.
34
C(S,T) Payoff diagrams show the value of
the contract for ALL possible
MATH 20912
scenarios at expiry.
C(S,T) = max(S-E, 0)
= S-E if S≥E
{ 0 if S<E
C = S-E
35
Do nothing
Exercise
Figure 4.4: Payoff diagram for the call option.

C=0
E S
Chapter 4
P(S,T)
MATH 20912
P(S,T) = max(E-S, 0)
= E-S if S≤E
E
{ 0 if S>E
P=E-S
36
Do nothing
Exercise
Figure 4.5: Payoff diagram for the put option.

P=0
E S
Chapter 4
To compare cash flows at different times, we must compare like with like, so if we pay money
for an option now and receive cash in the future, we must compare with what we could have
earned by investing the same money in a bank paying interest at rate r over the same period of
time instead.
For the holder of a European call option, the profit at time T is
profit = C(S, T ) − C0 erT

= max (ST − E, 0) − C0 erT .
For the holder of a European put option, the profit at time T is
profit = P (S, T ) − P0 erT

= max (E − ST , 0) − P0 erT .
You may be asked to calculate an investor’s expected profit from an investment. You
can do this if you have information about the distribution of possible share prices in the future.
The formula for the expected profit at time T is given by (using a call option as an example)
investor’s expected profit = E[C(S, T )] − C0 erT . (4.1)
NOTE: Here we must calculate the expected payoff E[C(ST , T )] and NOT the payoff of the
expected share price C(E[ST ], T ), which is one of the most common mistakes in exams.
Example 4.6. A European call option, with an exercise price of £10 and an expiry date in three
months’ time, was bought for £2.25 financed by a loan with continuously compounded interest
at a rate of 5%. Calculate the share price on expiry which gives a profit of £14.
Solution 4.6.
We have
• E = 10
• T = 0.25 (3 months)
• C0 = 2.25
• r = 0.05
• profit = 14.
37
We substitute these figures into
profit = max (ST − E, 0) − C0 erT ,
noting that in order to make a positive profit we must have ST > E. We find
14 = ST − 10 − 2.25 × e0.05×0.25 ,
and this gives a share price on expiry of
ST ≈ 26.28 .
38
Chapter 5
Trading with Portfolios
How can I sell something I don’t own?

Often, market participants will wish to take negative positions in the share price; that is to say,
they will look to profit when the share price goes down. This practice is called short selling. But
how can you sell something if you don’t own it? Well, traders are able to do this by exchanging
an “IOU” (“I owe you”) contract with a share holder to gain ownership of the shares for a period
of time. They then sell the shares on the market and invest the cash, waiting for the share price
to drop so that they are able to buy them back for a lower price. Once the trader has bought
the shares back, they can restore ownership to the original holder and cancel out the “IOU.”
There are laws surrounding short selling, as it can be used to raise supply (the number of
shares for sale) artificially above the actual number of holders who want to sell, and an increase
in supply normally results in a drop in price, so there is obviously an incentive on the part of
the sellers to do this. It has been blamed for several market crashes and is normally restricted
or even completely banned under certain circumstances.
5.1 Portfolio
A portfolio is a range of investments held by an individual. We assume that portfolios may
contain both positive and negative positions in stocks, shares, bonds, call options, and put
options. Mathematically, we tend to denote portfolios as Π, and your position in each asset will
be positive if you are holding it (long) and negative if you are selling it (short).
Example 5.1. Starting from zero, create a portfolio that is long or short one share by borrowing
or investing money.
39
Solution 5.1.
With zero initial investment, we start with an empty portfolio
Π = 0.
Now, if we wish to buy a share, then we need to borrow money from the bank, so after the trip
to the bank we have
Π = cash − B ,
where the negative B indicates that we now owe money to the bank. We have sold the bank a
promise to give them back the money (plus interest) in the future. Now we can use the cash to
buy one share and we have
Π=S−B.
Overall, we hold a single share ownership contract, and we owe money to the bank. We say that
this portfolio is long one share and short one bond.
Now, suppose we start with zero initial investment again
Π = 0.
This time we wish to sell one share. We find somebody who wishes to buy a share and we give
them an “IOU.” They give us cash, so we have
Π = cash − S .
We have sold the other party a promise to give them one share in the future. We can also put
the cash in the bank, so our portfolio is updated as
Π=B−S.
We have bought a promise from the bank to give us our money back in the future (plus interest).
Overall we hold a bond contract with the bank, and we owe one share to the other party. We
say that this portfolio is long one bond and short one share.
Therefore the following portfolio
Π(S, t) = S + 2C − P
describes a situation where the investor is long one share, long two call options, and short one
put option.
Example 5.2. Draw the payoff diagrams for going both long and short on both a put and call.
Solution 5.2. See presentation slides for Chapter 5.
40
5.2 Trading with Portfolios

There are various reasons why an investor may choose to buy a combination of options, shares,
and bonds, rather than just focus on one or another. The combination of financial contracts
that an investor holds is called the portfolio, and in this chapter we discuss some of the common
combinations that might be held by investors. We look at the payoff, and how to construct the
payoff diagrams for the resulting portfolios.
It is common for students to ask why an investor would take a certain position, particularly
if it looks as though the payoff at the end could be negative. To understand this you need to
remember that the price of taking a certain position will be related to the likely payoffs – there
is no easy way to make money on the stock market. If there is a large probability of a negative
payoff at the end of a contract (i.e. you pay out money rather than receiving it) then it is likely
that the contract will be very cheap (or even negative in price, so you receive money at the start)
to set up. The investor has to balance the risk of winning and losing against their view on the
market. Setting up different portfolios will allow them to maximize returns, given what they
think will happen in the future.
There are two main concerns for an investor who is trading in the market, and they are profit
and risk. In Chapter 6 we expand on the definition of risk, which is related to the variance of the
payoffs. When an investor buys or sells a contract, they wish either to increase their profit or
hedge their risks. If the investor wishes to make a profit, they will normally design a portfolio
with large positive payoffs, and that increases the risk. If the investor wishes to hedge or reduce
risk, they would design or adapt their portfolio such that the potential profit and losses are both
lower.
Hedge is something investors do to reduce their risk. This usually means that they are pro-
tecting themselves against a (potential) large financial loss. However, in Financial Mathematics
a reduction in risk might mean that both large losses and large profits are avoided, resulting in
a less risky portfolio.
5.2.1 Straddle
One of the most common portfolios is the straddle. This involves buying both a call option and
a put option, with the same strike price E and the same expiry date T , at the same time. The
value of this portfolio at time t is given by
Π(S, t) = C(S, t; E) + P (S, t; E) , (5.1)
and therefore the payoff at expiry (t = T ) is

(
E − S if S < E (you exercise the put);
Π(S, T ) = (5.2)
S − E if S ≥ E (you exercise the call).
41
We can see that this has large positive payoff if the share price has a large downward movement
or a large upward movement. Given that the investor wins both ways, it follows that the cost of
setting up such a portfolio will be relatively high.
The opposite position, that of a short straddle, is what you get if you sell the same two
options. The value of this portfolio at time t is
Π(S, t) = −C(S, t; E) − P (S, t; E) . (5.3)
At expiry (t = T ) the payoff is

(
−(E − S) if S < E (the buyer exercises the put);
Π(S, T ) = (5.4)
−(S − E) if S ≥ E (the buyer exercises the call).
Example 5.3. Draw the payoff diagrams for going long on a straddle, and short on a straddle.
Solution 5.3.
See Figure 5.1 and Figure 5.2.
Now, this short straddle contract has negative payoff in all scenarios (for all S), so why would
you do it? Well, if the payoff is always negative then the person who sells it to you will have to
give you a lot of money at the start. This means that you make money by hoping that there
isn’t a large movement up or down. As long as the movement in share price is small, you are
able to charge enough at the start to still make a potential profit in the end. This can be risky
as you may be likely to make small profits often, but there is always a small chance you might
make a big loss and potentially go bankrupt – making this a very risky strategy.
5.2.2 Profits from Portfolios

The formula for the expected profit at time T can be generalized to portfolios. Recall that the
expected profit for a call option was given in Chapter 4 as:
investor’s expected profit = E[C(S, T )] − C0 erT . (5.5)
So we can write the expected profit at time T for a portfolio of contracts is
investor’s expected profit = E[Π(S, T )] − Π0 erT . (5.6)
Similarly, we can calculate the expected return from a portfolio of contracts as
E[∆Π] E[Π(S, T )] − Π0
investor’s expected return = = . (5.7)
Π0 Π0
42
E-S if S<E we choose to exercise the put: sell at E
MATH 20912
�(S,T)=
{ S-E if S≥E we choose to exercise the call: buy at E
�(S,T)
43
E
Figure 5.1: Payoff diagram for a long straddle.

E S
Chapter 5
-E + S if S<E we are forced to buy the share at E
MATH 20912
�(S,T)=
-S + E if S≥E we are forced to sell the share at E
{
�(S,T)
44
E
-E
Figure 5.2: Payoff diagram for a short straddle.

Chapter 5
Example 5.4. An example of large profits.

For the long straddle portfolio (Π = C + P ), suppose that: S0 = 40, E = 40, C0 = 2, P0 = 2.
Find the investor’s expected return, if the investor believes that the share price at time T is
modelled by the following probability tree:
ST = 60
p= 3
4

S0 = 40 H

HH
p= 1
4
ST = 20
Solution 5.4.
p= 3Π = 60 − 40 = 20 as ST > E (exercise call option)

T
4
Π0 = 4
HH
H ΠT = 40 − 20 = 20 as ST < E (exercise put option)
p= 1
4
So we have
E[ΠT ] − Π0 20 − 4
investor’s expected return = = = 4,
Π0 4
which means a return of 400%!
5.2.3 Bull Spread

A bull spread is a slightly cheaper way to bet on an upward movement in the share price than
simply buying the call option. We create it by buying a call and then selling one with a slightly
higher exercise price, giving
Π(S, t) = C(S, t; E1 ) − C(S, t; E2 ) , (5.8)
where E2 > E1 . Why would you invest like this? Well, if you expect only a small movement in
share price, this is a cheaper way to exploit it than just buying the call option. The modification
can be considered as a hedge on the initial investment of buying the call, even though it is
limiting the potential winnings.
Example 5.5. Sketch the payoff diagram for a bull spread.
Solution 5.5.
45
The payoff of the portfolio is



 0 if S < E1 (neither call is exercised),

Π(S, T ) = S − E1 if E1 ≤ S < E2 (you exercise the 1st call),


 E −E if S ≥ E (you exercise the 1st call, buyer exercises the 2nd).
2 1 2
For the payoff diagram see Figure 5.3.
You may hear the term bull to describe market conditions; a bull market is one in which
confidence is high, and stocks and shares are growing in value. The term bear indicates the
opposite, and it refers to a situation where stocks and shares are losing value, so a bear spread
is used to exploit downward movements in price. We create a bear spread portfolio by buying
and selling a put option, i.e.
Π(S, t) = P (S, t; E1 ) − P (S, t; E2 ) , (5.9)
where E1 > E2 now.
Example 5.6. Sketch the payoff diagram for a bear spread.
Solution 5.6. Left as an exercise.
5.3 Risk-Free Investments

When dealing with contracts that make payments in the future, there is always a risk that you
won’t get paid, as even banks sometimes run out of money! In economics, we like to imagine that
there does exist an investment in the economy that is completely risk-free, and we can then use
this risk-free investment as a benchmark to compare against our options, portfolios, etc. Even
before we begin applying our model to the share price, there are things we can say about the
price of a financial contract, for instance the upper and lower bounds, that must be true under
any circumstances.
The risk-free investment is so important because the true market price (the price everyone
agrees on) is known, since the only thing that makes a value subjective is risk. We discuss this
further in the next chapter, but the main focus of this course is in describing a strategy by
which you can take risky financial derivatives and make them completely risk-free. This ability
to remove risk is what transformed the world of investment banking.
5.3.1 Bond
A bond is a contract that pays a known amount F , called the face value (or redemption value),
at a known time T called the maturity date (or redemption date). The authorized issuer/seller
46
�(S,T) = C(S,T;E1) - C(S,T;E2)
MATH 20912
E2-E1
Forced to sell one share at E2
47
Exercise option to buy one
share at price E1
Figure 5.3: Payoff diagram for bull spread.

E1 E2 S
Chapter 5
of the bond (for example, a country’s government) owes the holder a debt and is obligated to
repay the face value at maturity, and may also be obligated to make regular interest payments
(“coupons”). We will assume in this course that the only bonds traded are those issued by stable
governments that can be assumed to be risk-free. The term B(t) will denote the value at time t
of a risk-free investment in government bonds.
Face Value The final payment amount as described in the contract, usually denoted F . The
holder of the bond can collect the payment of F on the maturity date.
Maturity Date The maturity date is the date written in the contract; on this date the
holder receives the face value payment.
Interest Rate This is the growth rate of the bond, or how much value is added for the
holder over a period of time. If I deposit money into a restricted access savings account in a
bank, then the bank is effectively selling me a bond. Imagine I pay the bank (say) £1000 now
to receive £1100 in five years’ time: then I have bought a bond with face value F = 1100 and
maturity date T = 5, so the interest rate r (continuously compounded) could be worked out from
1000 er×5 = 1100 .
Meanwhile, the bank owes me money and I am hoping that they will have the money in five
years’ time to pay me!
Coupon These are payments from the issuer/seller of the bond to the holder/buyer. Usually,
if the maturity date is far in the future (they can be several decades!) the value of the bond
would be very low without these coupon payments. Coupons enable the seller of the bond to
receive more money from the initial sale.
Zero-Coupon Bond Simply a bond that does not pay any coupons.
Example 5.7. Given that the return on a risk-free bond can be defined as
dB
= rdt , (5.10)
B
where r is the risk-free interest rate, calculate the bond’s value at time t, if B(T ) = F and r is
constant.
Solution 5.7.
48
We are asked to calculate the function B(t). Integrating both sides of (5.10) with time limits
of t and T we get
Z B(T ) Z T
dβ
= rdτ ,
B(t) β t
where β and τ are dummy variables. Since r is constant we get
ln B(T ) − ln B(t) = r(T − t) .
We have been given the value of the bond at maturity, B(T ) = F , so
ln F − ln B(t) = r(T − t) ,
and, rearranging, we have the value of the bond
B(t) = F e−r(T −t) .
The term e−r(T −t) is often called the discount factor.
The No-Arbitrage Principle

In the next chapter we introduce the no-arbitrage principle, which is the bedrock of mathe-
matical finance. It allows us to draw equivalence between different types of financial contracts
and to construct arguments on how to price financial contracts.
49
Chapter 6
Risk and No-Arbitrage
6.1 Risk
Risk An investment is risky if its actual return will be different from its expected return.
Risky investments may go up or down, and they include the possibility of losing.
If a financial contract that is trading in a market is risky, economic theory tells us that
different individuals will value that contract differently according to their own current situation.
For instance, if one investor is rich and another investor is poor, then each investor will have a
different view on a particular investment according to how much they can afford to lose. The rich
investor may feel able to gamble for a big potential win, since a small potential loss is acceptable.
However, the poor investor cannot necessarily accept the same potential loss.
Dealing with risk in a market is extremely difficult because of this non-uniformity, and, before
we consider how to remove risk from valuations, it is important to note that valuing with risk is
possible but rather subjective.
Example 6.1. Two financial products are for sale on the market. Each product offers a certain
payoff at time T , based on the outcome of some future events. The whole market knows (and
agrees on) the probabilities associated with those future events.
Product A pays: Product B pays:

• £500 with 50% probability; • £0 with 99% probability;
• £1,500 with 50% probability. • £100,000 with 1% probability.
Suppose both products are available to buy for £750. Which is the better investment?
50
Solution 6.1.
We can calculate the expected payoff of each product at time T :
E(A) = 0.5 × 500 + 0.5 × 1500 = 1000 ,

E(B) = 0.99 × 0 + 0.01 × 100000 = 1000 .
Both products have equal expected payoff at T . Their prices were also equal (both £750), so
this means that their expected returns are equal: (1000 − 750)/750 = 1/3.
However, the two products feel very different because they have vastly different variances.
People would disagree over which product they would prefer to buy, based on their circumstances.
Typically, a rational investor would choose the one with the lowest variance, all else being equal.
Example 6.2. Suppose there are two similar products A and B for sale on the market, where
at time T :

• £1000 with 100% probability. • £250 with 100% probability.
Suppose A is available to buy for £800, and B for £200. Which is the better investment?
Solution 6.2.
We can calculate the return of each product:

1000 − 800 1
return(A) = = ,
800 4
250 − 200 1
return(B) = = .
200 4
Both products have the same return. Their variances are the same (both zero) because there is
no uncertainty in the payoff. Therefore we cannot choose between these two products on this
basis. Effectively, A is the same as 4B.
If one of the products had a different price, then there would be a way to make a risk-free
profit. For example, if A instead had a lower price of £600, then we could buy A and finance it
by selling 3 lots of product B. Our portfolio would be Π = A − 3B. Initially we have
Π0 = A − 3B = 600 − 3 × 200 = 0 ,
and then at time T we have
ΠT = A − 3B = 1000 − 3 × 250 = 250 ,
which equates to a £250 risk-free profit.
51
Example 6.3. Suppose there are two products A and B for sale on the market, where at time T :

• £1,000 with 100% probability. • £1,000 with 99% probability;
• £2,000 with 1% probability.
What can we say immediately about the prices of these two products?
Solution 6.3.
We can say that the price of B must be at least as much as the price of A, because B will
definitely payoff at least as much as A.
It is this sort of comparison between products that we are going to use in order to say
something about the prices of options in a market, by comparing them to other products in the
market.
6.2 No-Arbitrage
Arbitrage Opportunity An arbitrage opportunity is a way to make a risk-free profit.
Mathematically, we define such an opportunity as one which yields a positive payoff (in all
circumstances) with zero initial investment.
Suppose we have a portfolio Π(S, t) consisting of financial contracts that depend on a share S,
such that there is zero initial investment, i.e. Π(S, 0) = 0. Then an arbitrage opportunity exists
at time T if
Π(S, T ) > 0 for all S ∈ [0, ∞).
Note that we require the value of the portfolio at time T to be strictly greater than zero to
guarantee arbitrage.
In this course we are considering all circumstances (regardless of how we model the share
price) so this means all possible values of the share S at maturity, i.e. profit must be strictly
positive ∀S ∈ [0, ∞). For brevity, we will often use the notation
Πt = Π(S, t)
to denote the value of a portfolio at time t given a value of the share S at time t.
52
One of the key principles of financial mathematics is the No-Arbitrage Principle. This
says that there are never opportunities to make risk-free profit, i.e. there are no-arbitrage oppor-
tunities, and therefore any contract price which admits arbitrage can never exist in a market.
We will split our consideration of this principle into four key aspects:
• The No-Arbitrage Principle,
• Comparing Contracts,
• Equivalent Contracts,
• Return on Risk-Free Contracts.
6.2.1 The No-Arbitrage Principle

Here we define the No-Arbitrage Principle to be the result that a portfolio with a non-negative
payout at time T (i.e. ΠT ≥ 0) must be at least as valuable as holding nothing. Let us write this
as
ΠT ≥ 0 =⇒ Πt ≥ 0 ∀t ≤ T. (6.1)
This is a standard result to which we refer in order to derive conditions on the price of financial
contracts (or portfolios of contracts).
6.2.2 No-Arbitrage Principle – Comparing Contracts

Consider two financial contracts V and V̂ . If the payoff of each contract is such that
VT ≥ V̂T ,
then we can show that Vt ≥ V̂t for all t. We simply create a portfolio Π = V − V̂ , which has
ΠT ≥ 0, and then we use the no-arbitrage principle to deduce Πt ≥ 0 which gives us Vt ≥ V̂t .
Examples
Consider a contract V with payout (at time T ) bounded below by some fixed amount F , so that
VT ≥ F.
If B is the risk-free bond which pays face value F at maturity date T , then we have
V T ≥ BT .
Defining a portfolio Π = V − B, we have
ΠT = VT − BT ≥ 0 ,
53
and then by the no-arbitrage principle we have
Πt = Vt − Bt ≥ 0 ,
which gives us (referring to example 5.7 for the bond’s value Bt at time t)
Vt ≥ Bt = F e−r(T −t) . (6.2)
6.2.3 No-Arbitrage Principle – Equivalent Contracts

Consider two contracts that are identical in every way. It seems obvious that they must have
the same price in a market.
Suppose that V and V̂ are two financial contracts with the same payoff (upon expiry at
time T )
VT = V̂T ,
or in more detailed notation
V (S, T ) = V̂ (S, T ) for all S ∈ [0, ∞) .
If the no-arbitrage principle holds, then Vt = V̂t for all t ≤ T .
Example 6.4. Can you prove this result, using the no-arbitrage principle?
Solution 6.4.
Left as an exercise. Hint: VT = V̂T implies that VT ≥ V̂T and VT ≤ V̂T .
Examples
Suppose that there is a risk-free portfolio which has no profit or loss at maturity, in all circum-
stances, i.e. ΠT = 0 for all S. Then this is equivalent to holding nothing, so by the above it must
be true that
ΠT = 0 =⇒ Πt = 0 ∀t . (6.3)
Now suppose we construct a portfolio Π at time t, such that value of the portfolio at some
future time (t + dt) can be expressed as Πt+dt = K, where K is constant for all S. An equivalent
K
contract would be to invest in an amount F of bonds B with maturity date (t + dt) and face
value F . This means that we can write
K
Πt+dt = Bt+dt ,
F
and therefore, from the equivalent contracts result above, we have
K
Πt = Bt .
F
54
If we now take the limit dt → 0, we can deduce that the return on the investment for Π must be
the same as the return for the equivalent risk-free bonds, i.e.
dΠ dB
= = r dt, (6.4)
Π B
where r is the risk-free interest rate. This means that all risk-free portfolios must have the same
rate of return.
6.2.4 No-Arbitrage Principle – Return on Risk-Free Con-

tracts
Let Π be the value of a risk-free portfolio, and dΠ its increment during a small period of time dt.
Then, if the no-arbitrage principle holds, we have
dΠ
= r dt,
Π
where r is the risk-free interest rate.
Example 6.5. To see why this must be true practically, let us consider the opposite.
Suppose there are two risk-free products on the market offering to pay the holder £10 on
the same date T in the future. Let’s say that product A is selling for £5, and product B for £7.
Clearly, everyone (with a free choice) would choose to buy A, as it offers better return.
What happens to the demand (and hence price) of each product?
Solution 6.5.
• Demand for product A increases
• The seller(s) of A reacts by increasing the price of A
• Demand for product B decreases
• The seller(s) of B reacts by dropping the price of B
Ultimately, the two prices move until the price of A is equal to the price of B. In this sense, the
market adjusts itself.
6.3 Put-Call Parity

Using the above, we can now deduce a relationship between the market prices of put and call
options, that must hold at all times. In fact, often only the price of a call is calculated, and then
the corresponding put price follows using this relationship.
55
Example 6.6. Consider a portfolio of the form
Π=S+P −C. (6.5)
Calculate the payoff of Π upon maturity at time T . Then, by using the no-arbitrage principle,
derive the put-call parity relationship:
St + Pt − Ct = Ee−r(T −t) . (6.6)
Solution 6.6.
The payoff at maturity is given by

(
S + (E − S) − 0 = E if S ≤ E ,
ΠT =
S + 0 − (S − E) = E if S > E ,
so ΠT = E (constant) in all circumstances, and thus Π is risk-free.

In the previous section we found that the no-arbitrage principle implies that the return on a
risk-free portfolio satisfies
dΠ
= r dt .
Π
We now integrate, and use ΠT = E to set the constant of integration; we obtain
Πt = Ee−r(T −t) .
Thus, substituting back into (6.5), we have
St + Pt − Ct = Ee−r(T −t) .
The above shows that the price (at t = 0) of a European call option can be found from the
price of the European put option with the same exercise price E and expiry date T :
C0 = P0 + S0 − Ee−rT .
Example 6.7. Use the formula above to derive the lower bound for the price of a call option:
C0 ≥ S0 − Ee−rT . (6.7)
Solution 6.7.
The payoff for the corresponding put option is
PT = max (E − S, 0) ≥ 0.
56
Therefore, by the no-arbitrage principle we have P0 ≥ 0, so the above formula gives us
C0 ≥ S0 − Ee−rT .
This is the required lower bound for the call option.
The upper bound for the call option, and both lower and upper bounds for the put option,
are derived on Example Sheet 3. Together, the bounds on the call option are:
S0 − Ee−rT ≤ C0 ≤ S0 .
Example 6.8.
(a) Find a lower bound for a six month European call option with exercise price £35 given
that the initial share price is £40 and the risk-free interest rate is 5% per annum.
(b) Suppose that the European call option is available for £4. Show that there is an arbitrage
opportunity.
Solution 6.8.
(a) In this case the figures are
• S0 = 40,
• E = 35,
• T = 0.5 (six months),
• r = 0.05.
Using (6.7), the lower bound for the call option price is
C0 ≥ S0 − Ee−rT .
Inserting the figures above, we have
C0 ≥ 40 − 35 exp (−0.05 × 0.5) ≈ 5.864 .
(b) We have that

4 = C0 < S0 − Ee−rT ≈ 5.864 ,
57
so the call option is trading at a price that is too low. Thinking about this inequality as an
imbalance, we should buy the left-hand side that is too low and sell the right-hand side that is
too high. This tells us how to construct our portfolio:
Π = C − (S − B) = C + B − S .
We wish to demonstrate an arbitrage opportunity, so we want our portfolio to require zero initial
investment, i.e. Π0 = 0. This tells us the value of bonds B to include:
Π0 = C0 + B0 − S0 = 4 + B0 − 40 ,
=⇒ 0 = B0 − 36 ,
so we choose B0 = 36.
At the expiry date t = T , our portfolio has payoff:
ΠT = max (ST − E, 0) + B0 erT − ST ,
and inserting the figures, we have

( )
36.911 − ST , if ST ≤ 35
ΠT ≈ ≥ 1.911 > 0 .
ST − 35 + 36.911 − ST , if ST > 35
Thus our portfolio with Π0 = 0 can deliver positive payoff ΠT > 0 for all ST , and this represents
an arbitrage opportunity.
58
Chapter 7
Arbitrage
7.1 The No-Arbitrage Principle

“Arbitrage cannot exist in a market.”
No-arbitrage is a strong assumption about the prices that can exist in a real market. If we
set the price of a contract in the market, and our chosen price allows another participant in the
market to make a risk-free profit (meaning that we created an arbitrage opportunity for someone
else), then the price we chose was wrong and this could lead us to lose a lot of money. Conversely,
we assume that nobody else would make the same mistake.
We can use this idea to make deductions by contradiction; by looking for a set of possible
prices that would create an arbitrage opportunity, we can discount the possibility of those prices
from ever occurring in the real market.
For example, to prove bounds on an option price, say Ct ≥ 0, we suppose that we set an
opposite price and advertise it to the market, namely
Ct < 0. (7.1)
If we can show that any price satisfying (7.1) leads to an arbitrage opportunity, this contradicts
our no-arbitrage assumption. We then deduce that such prices must never be advertised, and so
the real price must in fact satisfy
Ct ≥ 0 .
59
7.2 Bounds on Put and Call Options

Example 7.1. Sketch the upper and lower bounds for a call option
St − Ee−r(T −t) ≤ Ct ≤ St , (7.2)
and a put option

Ee−r(T −t) − St ≤ Pt ≤ Ee−r(T −t) , (7.3)
as functions of St , at a fixed time t < T .
Solution 7.1.
Fill in Figure 7.1.
7.3 Put-Call Parity: Proof by Contradiction

We have already derived the put-call parity relationship, using the no-arbitrage principle applied
to the return of a risk-free contract or portfolio – see (6.6). Now we use a contradiction argument
to show that any deviation from this parity relationship will lead to an arbitrage opportunity.
Let us assume that the initial price of a put option P0 is too high relative to the call option
price C0 . If this is the case, then we can write
P0 > C0 − S0 + Ee−rT . (7.4)
Note here that it is not necessary for P0 or C0 to lie outwith the no-arbitrage bounds for their
individual prices; it is the combination which admits arbitrage.
Example 7.2. Justify to yourself why we will look for arbitrage using the following portfolio
Π=C −P −S+B. (7.5)
Hint: Use our mantra “Buy low, sell high”.
Solution 7.2.
We are assuming that P0 is priced too high. This suggests that we should sell the put option
(“buy low, sell high”). In order to take full advantage of the opportunity, we need to sell the put
option and simultaneously buy the other side of the inequality (7.4), i.e. we buy the call option
and sell the share. This means that we construct a portfolio of the form
Π = −P + (C − S) + B ,
60
C(S,t) C(S,t)>S - Ee-r(T-t)
C(S,t) < S
C(S,t)
C(S,T) = max(S-E, 0)
P(S,T) = max(E-S, 0)
P(S,t)
P(S,t) < Ee-r(T-t)
P(S,t)
S
P(S,t) > Ee-r(T-t) - S
Figure 7.1: A sketch of the bounds on option prices.
61
and we will use the bond B to set our initial investment Π0 to be zero.
This is continued in the next example.
Example 7.3. (continuing the previous example)

Show that this portfolio leads to arbitrage, and deduce that our initial assumption on the
put price was false, and hence the following must hold:
Pt ≤ Ct − St + Ee−r(T −t) . (7.6)
Solution 7.3.
To set our initial investment to zero, Π0 = 0, we should choose
B0 = P0 − C0 + S0 .
Given the initial assumption (7.4) about P0 , we can say something about the value of the bond
P0 > C0 − S0 + Ee−rT
=⇒ P0 − C0 + S0 > Ee−rT
=⇒ B0 > Ee−rT . (∗)
Now consider the payoff of the portfolio at maturity (t = T ):

( )
−(E − ST ) + 0 − ST + B0 erT if S < E
ΠT = −PT + CT − ST + BT = = −E + B0 erT .
−0 + (ST − E) − ST + B0 erT if S ≥ E
Now using the condition (∗) for the bond, we obtain
ΠT > 0 ,
and given that Π0 = 0, we have therefore demonstrated an arbitrage opportunity.

This contradicts the no-arbitrage principle for the market, so we deduce that the assumption
(7.4) about P0 must be false, and that the opposite holds. Finally, because the initial time is
arbitrary, we can deduce that the condition applies to all prior times t ≤ T , so we have
Pt ≤ Ct − St + Ee−r(T −t) .
Similarly, we can show that the opposite case will also admit arbitrage, namely if the put
option is priced too low. To do this, let us assume that the put option satisfies the following
condition
P0 < C0 − S0 + Ee−rT . (7.7)
62
Example 7.4. Justify to yourself why we will look for arbitrage using the following portfolio
Π=P −C +S−B. (7.8)
Solution 7.4.
We are now assuming that P0 is priced too low, relative to the call option (and the share).
This suggests that we should buy the put option (“buy low, sell high”), and sell the other side
of the inequality (7.7), i.e. sell the call and buy the share.
The resulting portfolio is
Π = P − (C − S) − B ,
where again we will use the bond B to set initial investment Π0 to be zero.
Example 7.5. Show that this portfolio leads to arbitrage, and deduce that our initial assumption
on the put price was false, and hence the following must hold:
Pt ≥ Ct − St + Ee−r(T −t) . (7.9)
Solution 7.5.
To set our initial investment to zero, Π0 = 0, we should choose
B0 = P0 − C0 + S0 .
Given the initial assumption (7.7) about P0 , we see that
P0 < C0 − S0 + Ee−rT
=⇒ P0 − C0 + S0 < Ee−rT
=⇒ B0 < Ee−rT . (∗)
Now consider the payoff of the portfolio at maturity (t = T ):

( )
(E − ST ) − 0 + ST − B0 erT if S < E
ΠT = PT − CT + ST − BT = = E − B0 erT .
0 − (ST − E) + ST − B0 erT if S ≥ E
Now using the condition (∗) for the bond, we obtain
ΠT > 0 ,
and given that Π0 = 0, we have therefore demonstrated an arbitrage opportunity.
63
This contradicts the no-arbitrage principle, so we deduce that the assumption (7.7) about
P0 must be false, and that the opposite holds. Finally, because the initial time is arbitrary, we
can deduce that the condition applies to all prior times t ≤ T , so we have
Pt ≥ Ct − St + Ee−r(T −t) .
Combining the two conditions (7.6) and (7.9) we have again established put-call parity:
Pt = Ct − St + Ee−r(T −t) . (7.10)
Example 7.6. Three month European call and put options with exercise price £12 are trading
at £3 and £6 respectively. The share price is £8 and the interest rate is 5%.
Show that there exists an arbitrage opportunity.
Solution 7.6.
Setting t = 0 in the put-call parity relationship (7.10) gives
P0 = C0 − S0 + Ee−rT ,
If the given market information does not satisfy this equality, then we will use the imbalance to
show the existence of arbitrage. In this question the figures are
• T = 0.25 (three months),
• E = 12,
• C0 = 3, P0 = 6,
• S0 = 8,
• r = 0.05,
so the right-hand side is
C0 − S0 + Ee−rT = 3 − 8 + 12 e−0.05×0.25 ≈ 6.851 ,
and the left-hand side is P0 = 6, so we have
P0 < C0 − S0 + Ee−rT .
This means the put option is trading too low, so we should buy the put and sell the other side.
Our portfolio is
Π=P −C +S−B.
64
We want Π0 = 0, so the bond investment should be
B0 = P0 − C0 + S0 = 6 − 3 + 8 = 11 .
At maturity (t = T = 0.25) the payoff of our portfolio is

( )
(E − ST ) − 0 + ST − B0 erT if ST < E
ΠT = rT
= E − B0 erT = 12 − 11 e0.05×0.25 ≈ 0.862 .
0 − (ST − E) + ST − B0 e if ST ≥ E
Therefore we have exhibited a portfolio with positive payoff ΠT > 0 for all values of the share
price S, and with zero initial investment Π0 = 0. This constitutes an arbitrage opportunity.
The arbitrage profit for this portfolio is 0.862.
65
Chapter 8
Binomial Trees
8.1 One-Step Binomial Model

Consider the following simple binomial model for the share price. Initially the share price is S0 .
At some future time t = T , the share price can either move up to S0 u (where u > 1) with
probability q, or it can move down to S0 d (where d < 1) with probability 1 − q.
Let C(S, t) denote the price of a call option at time t. Initially the option price is C0 =
C(S0 , 0). At time T , if the share price moves up to S0 u then the option price is Cu = C(S0 u, T ),
and if the share price moves down to S0 d then the option price is Cd = C(S0 d, T ).
Example 8.1. Sketch one-step tree diagrams to represent the share and option price.
Solution 8.1.
uS0 u uCu
q q

u

S0 H C0 u
H

HH HH
1 −H
q HHu 1 − q HHu
H
S0 d Cd
Figure 8.1: One-step tree: share price. Figure 8.2: One-step tree: option price.
66
From Figure 8.1 it is straightforward to calculate the expected value and the variance of the
share price at time T . For instance, the expected share price at time T is
E(ST ) = qS0 u + (1 − q)S0 d .
Similarly, the expected call option price at time T is
E(CT ) = qCu + (1 − q)Cd . (8.1)
Given that the expected value of the call option seems to be known at time T , this suggests that
we should be able to deduce something about the initial price of the option C0 at t = 0.
Example 8.2. Can we use (8.1) to price the call option?
Solution 8.2.
It turns out that we cannot. The reason is risk.
We could use (8.1) to price the option if everything on the right-hand side was known. However,
even if the values of u and d are known, the probability q is not necessarily known, or agreed
upon.
Different buyers and sellers in the market might have different ideas about what the probability q
should be, and we cannot assume that everyone will make the same assessment.
However, we would like there to be one single market price. So how do we come up with the one
‘true’ price? We eliminate the risk, i.e. we eliminate the probability q.
8.2 Risk-Free Portfolio

The trick here is to sell the call option and buy the share at the same time. We can balance any
likely increase/decrease in one of them by cancelling it with a decrease/increase in the other one,
in order to get the same payoff in both cases. In Chapter 6 we saw how to deal with portfolios
with guaranteed payoffs: by the no-arbitrage principle, they must have a risk-free return. This
will enable us to derive the true price of the option.
Consider the portfolio

Π = ∆S − C,
where ∆ is a coefficient that we can choose. Here we interpret ∆ to be the number of shares per call
in the portfolio, so we do not restrict ∆ to integer values; e.g. a large institution such as a bank
may be selling 10,000 call options, so if ∆ = 0.6251 then this means they hold 6,251 shares.
67
Example 8.3. Using this portfolio, show that if

Cu − Cd
∆= (8.2)
S0 (u − d)
then the portfolio has a guaranteed (constant) payoff at expiry time T .
Solution 8.3.
The portfolio’s payoff value at expiry time T is

(
∆S0 u − Cu with probability q,
ΠT = ∆ST − CT =
∆S0 d − Cd with probability 1 − q.
In order for the payoff to be guaranteed, the payoff value in each case must be the same, i.e.
∆S0 u − Cu = ∆S0 d − Cd .
Rearranging, we find that we should choose ∆ such that

Cu − Cd
∆=
S0 (u − d)
to have a risk-free portfolio Π = ∆S − C.
Can we now deduce the initial price of the call option? Now that there is no risk in the
payoff, the answer is yes! For a portfolio with a guaranteed constant payoff ΠT = K, we have
Π0 = Ke−rT ,
since the rate of return of a risk-free portfolio is the risk-free interest rate r, by equation (6.4).
Example 8.4. Derive the following formula for the initial call option price
C0 = e−rT [pCu + (1 − p)Cd ] , (8.3)
for some parameter p to be found in terms of the other model parameters.
Solution 8.4.
Based on example 8.3, the guaranteed payoff of our portfolio Π is
ΠT = ∆S0 u − Cu ,
so the initial value of this portfolio is
Π0 = (∆S0 u − Cu )e−rT .
68
But by definition Π = ∆S − C, so the initial value is
Π0 = ∆S0 − C0 .
Equating these two expressions for Π0 , we can rearrange for C0 to find
C0 = ∆S0 − (∆S0 u − Cu )e−rT .
This is the true market price of the call option, derived via a no-arbitrage argument with hedging.
Now we can substitute in the value of ∆ from (8.2) to find

Cu − Cd Cu − Cd
C0 = − u − Cu e−rT
u−d u−d
rT
−rT e −d erT − d
=e Cu + 1 − Cd
u−d u−d
= e−rT [pCu + (1 − p)Cd ] ,
where
erT − d
p= .
u−d
The pricing of options via formulæ such as (8.3) is called “Risk-Neutral Valuation.”
The probability q does not appear in formula (8.3) because the payoff of the portfolio was
the same in both cases, so the value of q did not matter. The parameter p can be interpreted1
as a risk-neutral probability. Under this interpretation, the formula (8.3) states that the
initial price of the call option is equal to the expected value of its payoff under the risk-neutral
probability measure, discounted back to t = 0, i.e.
C0 = e−rT Ep (CT ).
In practical terms, in order to interpret the meaning of the risk-neutral probability p, we

need to think about the model in reverse: p is not the probability of the share moving up or
down, it is the implied probability we would need to put in the formula to obtain the same price
that we observe on the market.
We assume that the agents on the market who effectively set the price (i.e. banks, large
institutions) are always hedging away their risk. Therefore, they set the price using the no-
arbitrage hedging strategy. An individual observer will see that price on the market and may
ask what probability in the model would produce that price – the answer will be p, not q.
1
Note that in order for p to be interpreted as a probability, we must have p ∈ [0, 1] so we
must have d ≤ erT ≤ u in the model; see Chapter 9.
69
Example 8.5. A share price is currently £40. In three months’ time, the share price will be
either £44 or £36. The risk-free interest rate is 12%.
Calculate the price of a three-month European call option with an exercise price of £42.
Solution 8.5.
44 36
In this example S0 = 40, T = 0.25, u = 40 = 1.1, d = 40 = 0.9, r = 0.12, and E = 42.
We calculate the two possible values of the call option payoff:
Cu = C(S0 u, T ) = max(S0 u − E, 0) = max(44 − 42, 0) = 2,
and
Cd = C(S0 d, T ) = max(S0 d − E, 0) = max(36 − 42, 0) = 0.
We also calculate the risk neutral probability
erT − d e0.12×0.25 − 0.9

p= = ≈ 0.65227.
u−d 1.1 − 0.9
Then using formula (8.3), the initial price of the call option is
C0 = e−rT [pCu + (1 − p)Cd ] = e−0.12×0.25 [0.65227 × 2 + 0] ≈ 1.266.
70
Chapter 9
Two-Step Binomial Trees
9.1 Risk-Neutral Valuation

In the previous chapter we derived the initial call option price in a one-step binomial model as
C0 = e−rT [pCu + (1 − p)Cd ] ,
where
erT − d
p= .
u−d
This formula is known as a risk-neutral valuation.
Example 9.1. Show that in order for the no-arbitrage principle to hold in a one-step binomial
model, we require
d ≤ erT ≤ u, (9.1)
and hence
0 ≤ p ≤ 1.
Solution 9.1.
We proceed by contradiction. Suppose that
erT > u. (?)
Since r is the risk-free interest rate, this implies that the share will definitely perform worse (in
the model) than a risk-free investment in all circumstances, since the upper share price S0 u is
less than S0 erT . In other words, the share’s current price S0 is too high, so to exploit this we
set up a portfolio
Π = B − S,
71
and we set Π0 = 0 by choosing B0 = S0 . Then at time T we have

(
B0 erT − S0 u with probability q,
ΠT =
B0 erT − S0 d with probability 1 − q,
(
S0 (erT − u) with probability q,
=
S0 (erT − d) with probability 1 − q.
Now, using (?), and the fact that u > d in the model, we have
ΠT > 0
in both cases. Therefore we have a portfolio with a strictly positive payoff ΠT > 0 at time T ,
for zero initial investment Π0 = 0, and this is an arbitrage opportunity. For the no-arbitrage
principle to hold, we therefore require
erT ≤ u,
which implies p ≤ 1.
Deriving p ≥ 0 is left as an exercise.
Given that (9.1) holds, we have 0 ≤ p ≤ 1, and we can interpret the parameter p as the
risk-neutral probability of an upward movement in the share price. The model probability q
of an upward movement seems to play no role whatsoever! Why?
Let us find the expected share price at time T under the risk-neutral probabilities:
rT
e −d erT − d
Ep (ST ) = pS0 u + (1 − p)S0 d = S0 u + 1 − S0 d = S0 erT .
u−d u−d
This shows that, under these probabilities, the share price grows at the risk-free interest rate r.
This is called a risk-neutral world.
In the Real World: In a Risk-Neutral World:

µT
E(ST ) = S0 e Ep (ST ) = S0 erT
In Chapter 8 we derived our risk-neutral valuation formula for the call option price:
C0 = e−rT Ep (CT ). (9.2)
This formula says that the call option price is the expected payoff of the option in a risk-neutral
world, discounted at the risk-free interest rate r. The expected payoff is calculated under the
risk-neutral probability measure which allows us to eliminate risk to derive a single price for the
option.
72
9.2 Two-Step Tree

We now consider a two-step extension of the previous one-step model. In the two-step model,
the share price changes twice, each time by a factor of either u > 1 or d < 1. Let us say that
the length of each time-step is δt, such that T = 2δt. After two time-steps, the share price will
be either S0 u2 , S0 ud, or S0 d2 .
Example 9.2. Sketch the two-step tree diagram for the share price.
Solution 9.2.
S0 u2
v

S0 u
v
PP
PP
PP
PP S0 ud
P
v v

S0 P

PP
P
Pv

P
PP
P PP S0 d
P
Pvv

PP
PP
PP
P PP S0 d2
P
Pv
The call option expires after two time steps, producing payoff Cuu , Cud , or Cdd respectively.
Example 9.3. Write down formulæ for Cuu , Cud , and Cdd in terms of S0 , E, u, and d.
Solution 9.3.
First note that

Cuu = C(S0 u2 , T ),
so
Cuu = max(S0 u2 − E, 0).
Similarly
Cud = max(S0 ud − E, 0),
and
Cdd = max(S0 d2 − E, 0).
73
Example 9.4. Sketch the two-step tree diagram for the call option price.
Solution 9.4.
Cuu

v

Cu

v

PP

PP
PP

P PP Cud
C0 v

P

v
P
Pv

PP
PP
PP
PP Cd
P
Pvv

PP
PP
PP
P PP Cdd
P
Pv
Our aim is to calculate the initial option price C0 , at the initial node of the tree. Before we
can find C0 , we must apply the risk-neutral valuation backwards in time, one step at a time, by
first finding Cu and Cd . To do this, we simply use the one-step formula to price the option at
these two intermediate nodes:
Cu = e−rδt (pCuu + (1 − p)Cud ) ,
and
Cd = e−rδt (pCud + (1 − p)Cdd ) .
Then we apply the one-step formula again to work backwards to the initial node:
C0 = e−rδt (pCu + (1 − p)Cd ) .
Example 9.5. Substituting to eliminate the intermediate values Cu and Cd , we find

C0 = e−2rδt p2 Cuu + 2p(1 − p)Cud + (1 − p)2 Cdd , (9.3)
for the initial option price. What do the coefficients p2 , 2p(1 − p), and (1 − p)2 represent?
Solution 9.5.
Here p2 , 2p(1−p), and (1−p)2 are the probabilities in a risk-neutral world that the upper,
middle, and lower final nodes are reached.
In other words, they are the risk-neutral probabilities of the share price going up twice, or
up once and down once, or down twice, respectively.
74
In summary, noting that T = 2δt, we can write the initial call option price as
The initial put option price can be found in the same way:

P0 = e−2rδt p2 Puu + 2p(1 − p)Pud + (1 − p)2 Pdd , (9.4)
which we can write as

P0 = e−rT Ep (PT ).
Example 9.6. Consider a six-month European put option with an exercise price of £32. The
initial share price is £40, and the risk-free interest rate is 10% per annum.
Calculate the initial put option price using a two-step model in which the share price either
moves up by 20% or down by 20% at each time-step.
Solution 9.6.
With a two-step model, the initial put option price is given by formula (9.4):

P0 = e−2rδt p2 Puu + 2p(1 − p)Pud + (1 − p)2 Pdd . (∗)
In our notation, the given figures correspond to:
T = 0.5, E = 32, S0 = 40, r = 0.1, u = 1.2, d = 0.8.
Using a two-step model, our time-step is δt = T /2 = 0.25.
The risk-neutral probability p is
erδt − d e0.1×0.25 − 0.8

p= = ≈ 0.56329.
u−d 1.2 − 0.8
The three possible payoffs of the put option at time T are
Puu = max (E − S0 u2 , 0) = max (32 − 40 × 1.22 , 0) = 0,

Pud = max (E − S0 ud, 0) = max (32 − 40 × 1.2 × 0.8, 0) = 0,
Pdd = max (E − S0 d2 , 0) = max (32 − 40 × 0.82 , 0) = 6.4.
Putting everything into (∗), the initial put option price is

P0 ≈ e−2×0.1×0.25 × 0 + 0 + (1 − 0.56329)2 × 6.4
≈ 1.161,
or £1.16 to the nearest penny.
75
76
Chapter 10
The Multi-step Binomial Model
10.1 A Discrete Model for the Share Price

Reminder: continuous model for the share price
Earlier in the course we considered the continuous random walk model for the share price:
dS = µSdt + σSdW.
Example 10.1. Given the initial share price S0 , what are the possible values of the share price S
at time t, in this continuous model?
Solution 10.1.
• In this continuous model, any positive share price S ∈ (0, ∞) is possible (has a positive
probability) at any time t > 0.
• In Chapter 3 we found the probability density function for the share price S, which has a
log-normal distribution.
Discrete model for the share price

The one-step and two-step binomial models we have considered in the previous two chapters
are examples of discrete models for the share price. They are both particular cases of a more
general multi-step binomial model, which we summarize as follows:
77
• let N be the number of steps
• the share price S changes only at discrete times: δt, 2δt, 3δt, . . . , N δt
• the price either moves up (S 7→ Su) or down (S 7→ Sd), where d ≤ erδt ≤ u
• the probability of each upward movement is q.
Example 10.2. If there are N steps in the tree, how many possible share prices can be observed?
Solution 10.2.
Each node in the tree corresponds to a different price, so we count the number of nodes.
In the one-step tree, there were three nodes in total.
In the two-step tree, the second step added three more nodes to make six in total.
Thinking inductively, the k-th step adds (k + 1) more nodes, so the total number of nodes in an
N -step tree is
N
X (N + 1)(N + 2)
(k + 1) = ,
2
k=0
and this is the number of possible share prices for the N -step discrete binomial model.
10.2 Binomial Share Price Tree

Consider the tree of possible share prices in the multi-step binomial model. The tree is called
a binomial tree, because the share price will either move up or down at each time-step. Each
node in the tree represents a possible future share price.
If the total time duration of the model is T , then we divide the time interval [0, T ] into
N equal time-steps of duration δt = T /N . Our aim is to have the capacity to increase N to a
large enough number such that the N -step binomial tree approximates the continuous model.
Example 10.3. Sketch the tree for a 4-step binomial model for the share price.
Solution 10.3.
See figure 10.1.
To enumerate the possible share prices in the multi-step binomial model, we introduce the
following notation:
• Let Snm denote the share price at time t = mδt if there have been n upward movements.
78
S0 u4
N =4 u

3
S0 u

u

PP
PP
2 3
S0 u PPS0 u d
PP
u
P u
PPP

PP 2
S0 u PP S0 u d

u

PP
PPu

PP

PP PP
2 2
S0 PPS0 ud PPS0 u d
PP PP

u
PP
P u
P u
PPP
PP
PP S0 ud2
PP
PPS0 d
PP
Pu

PP
PPu

PP

PP PP
2 3
PPS0 d PPS0 ud
PP PP
Pu
PP P
u
PP
3
PPS0 d
PP
Pu

PP
PP
4
PPS0 d
PP
Pu
0 δt 2δt 3δt 4δt = T
Figure 10.1: A sketch of the share price tree for a 4-step binomial model.
• Equivalently, Snm = S0 un dm−n where n ∈ {0, 1, 2, . . . , m}.
• S00 = S0 is the initial share price at time t = 0.
• Note that u and d are the same at every node in the tree.
For example, after the third time-step, at time t = 3δt, there are four possible share prices:
S03 = S0 d3 , S13 = S0 ud2 , S23 = S0 u2 d, and S33 = S0 u3 .
After the final time-step, at t = N δt, there are (N + 1) possible share prices.
10.3 Binomial Call Option Price Tree

Example 10.4. Sketch the tree for a 4-step binomial model for the call option price.
79
Solution 10.4.
See figure 10.2.
N =4 Ctuuuu = max (S0 u4 − E, 0)

C
tuuu

PP

C PP Cuuud = max (S0 u3 d − E, 0)
PP
t
P
uu Pt
PP
C
tu
PP C
PPt
uud

P PP P PP
2 2
PPtuudd = max (S0 u d − E, 0)
C0 PP C PP C
t
P
PPP
t
ud

PP PP
PP C PP C
td PPt
udd
PP
P PP PPP
3
PPtuddd = max (S0 ud − E, 0)
PP C PP C
PPtdd
PP
PP
PP Cddd
Pt
PP
PP Cdddd = max (S0 d4 − E, 0)
PP
Pt
0 δt 2δt 3δt 4δt = T
Figure 10.2: A sketch of the call option price tree for a 4-step binomial model.
Let Cnm denote the call option price at time t = mδt if there have been n upward movements.
In order to calculate the initial price C0 of the call option we must work recursively backwards,
just as we did with the two-step tree. Over each step in the tree we apply a Risk-Neutral
Valuation according to the one-step formula (derived in Chapter 9):

Cnm = e−rδt pCn+1
m+1
+ (1 − p)Cnm+1 .
Here 0 ≤ n ≤ m, 0 ≤ m < N , and

erδt − d
p= .
u−d
In order to solve the recurrence relation to find the initial option price, we need a boundary
condition at t = T , and this is given by the set of possible payoffs of the call option:
CnN = max (SnN − E, 0),
where 0 ≤ n ≤ N , and E is the exercise price.
80
Solving inductively, we find that the initial call option price C0 is again given by the expected
payoff under the risk-neutral probabilities, discounted at the risk-free interest rate r, i.e.:
Example 10.5. Why not use arbitrage arguments instead?
Solution 10.5.
Once p has been calculated, the same p can be used repeatedly to obtain the risk-neutral price
at every node in the tree.
Whereas, in order to use arbitrage arguments, the value of ∆ must be re-calculated at every
step, which would require up to twice as many calculations as the risk-neutral method.
Hence, calculating via the risk-neutral probability measure is more efficient and has become
extremely popular in simulations.
10.4 Approximating the Continuous Model

If our aim is to develop a discrete binomial model which approximates the continuous model, we
need to choose the values of the discrete model parameters – u, d, and q – such that the discrete
model matches the key properties of the continuous model. In particular the mean and variance
of the models should match.
Suppose that we fix the initial share price S0 and the time-step δt. Let’s compare the mean

share price E(S) and the variance of the return Var ∆S S , for both models.
Mean share price

• For the continuous model, at time δt we have E(S) = S0 eµδt , since µ is the expected
growth rate of the share price.
• For the discrete model, after one time-step we have E(S) = qS0 u + (1 − q)S0 d.
Example 10.6. Combine these two results to obtain the first equation needed to match the
continuous and discrete models.
Solution 10.6.
81
Set the two expressions for E(S) equal to each other:
qu + (1 − q)d = eµδt .
Variance of the return

• For the continuous model, at time δt we have

∆S
Var = σ 2 δt,
S
since σ is the volatility of the share price.
Example 10.7. Calculate the variance of the return after one time-step for the discrete model,
and hence obtain the second equation needed to match the models.
Solution 10.7.
For the discrete model, an upward movement in share price (probability q) gives a return of
∆S S0 u − S0
= = u − 1,
S S0
while a downward movement (probability 1 − q) gives a return of (d − 1).
We can calculate the variance as:
" 2 # 2
∆S ∆S ∆S
Var =E − E
S S S
2
= q(u − 1)2 + (1 − q)(d − 1)2 − [q(u − 1) + (1 − q)(d − 1)]
= q(1 − q)(u − 1)2 + q(1 − q)(d − 1)2 − 2q(1 − q)(u − 1)(d − 1)
= q(1 − q)(u − d)2 .
Setting the two expressions for the variance equal to each other, we have the second equation:
q(1 − q)(u − d)2 = σ 2 δt.
What is missing?
We now have two equations for three unknowns – u, d, and q – so what is missing? In fact, we
have a free choice for a third equation. One of the most popular binomial models is the CRR1
model which imposes u = d−1 as the third equation. This means that an upward movement
followed by a downward movement takes you back to where you started, i.e. S0 ud = S0 .
1
Cox, Ross, and Rubinstein 1979.
82
Can we solve these equations?

See example sheet 4.
83
Chapter 11
American Options and

Replicating Portfolios
American Option : an option that may be exercised at any time prior to expiry (t = T ).
The decision about when to exercise the option is up to the holder of the option. In order to
price the option, we need to determine when it is in the holder’s best interests to exercise. In
fact, it can be shown that this is not subjective; it can be determined in a systematic way.
Example 11.1. Consider the historical data for a particular share price S(t) in Figure 11.1.
Can you decide when it would have been best for the holder to exercise an American put option?
Solution 11.1.
• Based on the historical data, the share price was lowest at time t = t0 so this would have
been the best time to exercise the put option (highest payoff).
• But this is hindsight. The holder cannot see into the future, and the share price could
have gone down even further after t0 .
• The decision to exercise must take account of all future scenarios. In fact, the optimal
decision is to exercise earlier at t = t1 . But determining this involves solving a stochastic
PDE which we have not yet derived.
84
10
8
Stock Price S(t)
Share Price S(t)
4 t1 t0
0
0 0.25 0.5 0.75 1
Timet t
Time
10
8
Exercise value S(t) − E
Put option payoff E − S(t)
6 t1 t0
0
0 0.25 0.5 0.75 1
Timet t
Time
Figure 11.1: Above: a simulated share price S as a function of time t. Below : the
corresponding put option payoff E − S(t) (if exercised at time t), where E = 10.
85
The flexibility of the exercise time means that pricing American options is harder than pricing
European options. However, we can say something straight away about the value of an American
put option, based on the simple condition that
P (S, t) ≥ E − S(t), (11.1)
for all S and t. This condition says that the value of an American put option must always be
greater than or equal to its payoff function. Or in other words, the option cannot be worth less
than its payoff would be if it was exercised immediately.
Example 11.2. If 0 < P < E − S, show that there is an arbitrage opportunity. Hence deduce
that condition (11.1) must hold.
Solution 11.2.
Suppose that at some time (wlog the initial time t = 0) we have
0 < P0 < E − S0 . (?)
Then at that time, the put option costs less than the payoff E −S0 we would receive by exercising
it immediately. Therefore the put option is trading too low, and we should buy the option and
sell the right-hand side of (?). Our portfolio is
Π = P − (B − S) = P − B + S.
The initial cost of this portfolio is
Π0 = P0 − B0 + S0 ,
and we set Π0 = 0 by choosing

B0 = P0 + S0
and (?) tells us that
B0 < E. (??)
Now all we need to do to make an immediate profit is to exercise our put option immediately
(as soon as possible, say t = 0+ ) so that our payoff is
Π0+ = E − S0 − B0 + S0 = E − B0 .
By (??), the payoff is strictly positive, i.e. Π0+ > 0, so we have demonstrated that an arbitrage
opportunity exists. This contradicts the no-arbitrage principle, so our initial supposition (?) was
false, and we deduce condition (11.1):
P (S, t) ≥ E − S(t).
86
Condition (11.1) looks simple enough, but the valuation of an American put option is a
non-linear problem with no analytical solution in the continuous share price model.
However, we can derive solutions using discrete binomial models – which is one of the reasons
why binomial trees are popular with quantitative analysts.
11.1 American Option Tree

Let us denote by Pnm the value of a put option at the m-th time-step, t = mδt, if there have been
n upward movements in the share price (and m − n downward movements). Here 0 ≤ n ≤ m.
The recurrence relation we had for a European put option was

Pnm = e−rδt pPn+1
m+1
+ (1 − p)Pnm+1 ,
erδt −d
where p is the risk-neutral probability, p = u−d .
To adapt this for an American put option, we must include condition (11.1) in our calcula-
tions. At each node of the tree we must compare the valuation obtained by continuing to the
next time-step, and the valuation based on exercising the option at the current time-step:

Pnm = max e−rδt pPn+1 m+1
+ (1 − p)Pnm+1 , E − Snm , (11.2)
where Snm is the share price at the m-th time-step if there have been n upward movements.
The possible payoffs at the expiry time t = T = N δt are the same as before:

PnN = max E − SnN , 0 .
Example 11.3. Evaluating an American Put Option on a Two-Step Tree.
Suppose that over each of the next two years, the share price either moves up by 20% or
down by 20%. The risk-free interest rate is 5% per annum.
Find the initial price of a two-year American put option with exercise price $52, given that
the initial share price is $50.
Solution 11.3.
In our notation, we have N = 2, T = 2, δt = 1 (two time-steps of one year each), u = 1.2,

d = 0.8, r = 0.05, E = 52, S0 = 50.
The risk-neutral probability for our two-step tree is

erδt − d e0.05 − 0.8
p= = ≈ 0.628178.
u−d 1.2 − 0.8
87
The share price tree: The put option tree:
S0 u2 = 72 Puu = max (E − S0 u2 , 0)
t t

=0
S0 u =
60 Pu
t

P t
PP
PP
S0 = 50 PP S ud = 48 P0 t
P PP P = max (E − S ud, 0)
t
PPtt0 PPttud 0
PP
PP PP
PP = 4
S0 d =
40 Pd
t t
PP P PPt
Pt
P PP
PP PP
PP S d2 = 32, PP Pdd = max (E − S0 d2 , 0)
PPt0 Pt
= 20.
At the intermediate nodes, we must compare the value of continuing to the second time-step
(calculated by risk-neutral probability) and the value of exercising after the first time-step.
For Pu , the value of continuing to the second step is

−rδt −0.05×1
e (pPuu + (1 − p)Pud ) ≈ e 0 + (1 − 0.628178) × 4 ≈ 1.4148,
whereas the value of exercising after one step is
E − S0 u = 52 − 60 = −8.
Therefore
Pu = max (1.4148, −8) ≈ 1.4148.
Similarly, for Pd the value of continuing to the second step is

−rδt −0.05×1
e (pPud + (1 − p)Pdd ) ≈ e 0.628178 × 4 + (1 − 0.628178) × 20 ≈ 9.4639,
whereas the value of exercising after one step is
E − S0 d = 52 − 40 = 12,
so we have
Pd = max (9.4639, 12) = 12.
Now that we have calculated the American put option’s value at every future node, we can
finally calculate the option’s initial price P0 using risk-neutral valuation:

−rδt −0.05×1
P0 = e (pPu + (1 − p)Pd ) ≈ e 0.628178 × 1.4148 + (1 − 0.628178) × 12 ≈ 5.090.
Therefore the initial price of the American put option is approximately $5.09.
You can check that the corresponding European put option would have initial price $4.19.
Note: strictly speaking we should also verify that it is not more profitable simply to exercise the option immediately
at t = 0: the value of doing that would be E − S0 = 52 − 50 = 2, and this is less than 5.09.
88
11.2 Replicating Portfolios

This next part of the course asks whether we can replicate a portfolio of options by using a
trading strategy involving only shares and bonds. Being able to do this is extremely important
for the “big players” in the stock market (investment banks), as it allows them to lock in the
profit of a contract at the point of sale, rather than having to wait until the contracts are settled.
The term strategy here indicates that we will determine the number of shares we need according
to some objective, depending on what we are trying to replicate.
To consider a particular case, let us establish a portfolio Π consisting only of shares and
bonds, such that the payoff of a European call option is replicated. Our replicating portfolio is
Π = ∆S − N B,
where ∆ is the number of shares held long, and N is the number of bonds held short. The payoff
of Π replicates a European call option if
ΠT = CT = max (S − E, 0).
By the no-arbitrage principle, if two portfolios have equal payoff values then their values at any
earlier time should also be equal, so we have Πt = Ct for t ≤ T . In this sense, we say that the
portfolio Π replicates the option C.
The pair (∆, N ) is called the trading strategy. In general the values of ∆ and N may be
different for every value of S and t. The portfolio Π is called self-financing if any change in ∆
(buying/selling shares) is offset by a change in N (selling/buying bonds).
11.2.1 Replicating the One-Step Binomial Tree

Example 11.4. For the one-step binomial model, can we find (∆, N ) such that ΠT = CT ?
Solution 11.4.
As usual in the one-step tree, the share price starts at S0 and can either move up from to
S0 u or down to S0 d. At expiry time T , the European call option payoff is Cu if the share price
moves up, or Cd if the share price moves down.
Su Cu
u u

S0 C0
u u

P
PP
PP PP
PP
PP Sd PP
PPCd
PPu Pu
89
Our replicating portfolio is

Π = ∆S − N B,
so at time T its value is

(
∆S0 u − N B0 erT (upward share movement)
ΠT =
∆S0 d − N B0 erT (downward share movement)
We need ΠT = CT in order for our portfolio to replicate the call option payoff, so we have
the following two equations:
∆S0 u − N B0 erT = Cu ,
∆S0 d − N B0 erT = Cd .
Therefore we have two simultaneous equations which can be solved for two variables ∆ and N .
See Example Sheet 5 for the conclusion.
90
Chapter 12
The Black-Scholes Model
Myron Scholes and Fischer Black
The Black-Scholes model for option pricing was developed by Fischer Black and Myron
Scholes in the early 1970s, and was published in 1973.1 This model was one of the most important
developments in financial mathematics. Previously, traders and banks used hedging techniques
to minimize risk but did not do so in a way that was consistent with the market and pricing
framework. Black and Scholes were the first to show how risk could be eliminated – this was
the most important step forward.
The Black-Scholes model can be used to calculate an option’s value using a small set of
parameters: these include the exercise price E, the expiry time T , the current share price S0 ,
the volatility σ, and the risk-free interest rate r. An important parameter which is omitted
(intentionally) from that set is µ, the expected growth rate of the share price. One of the
benefits of the Black-Scholes model is that it not only eliminates risk, but it also eliminates the
need to estimate the growth rate µ.
1
“The Pricing of Options and Corporate Liabilities,” Journal of Political Economy, Vol. 81,
No. 3 (1973), pp. 637–654.
91
12.1 Model Assumptions

Example 12.1. Can the Black-Scholes model be used to price any financial contract?
Solution 12.1.
Not exactly. In order for Black & Scholes to set up a framework in which risk could be eliminated,
they had to make restrictive assumptions. Subsequently, other academics and practitioners tried
to relax these assumptions to use the model in situations for which it was not designed.
The assumptions are:
• money can be borrowed and lent at a constant risk-free interest rate r;
• the share price follows a Geometric Brownian Motion (see section 3.3) with constant
expected growth rate µ and volatility σ;
• there are no transaction costs;
• the share does not pay dividends;
• financial contracts are perfectly divisible (e.g. we can buy/sell any fraction of a share);
• there are no restrictions on short selling.
12.2 Deriving the Black-Scholes Equation
In the following we denote by V (S, t) the value of an option. As before, we use the notation
C(S, t) for a call option, and P (S, t) for a put option, whenever the distinction is important.
First, we set up a portfolio which is ‘long’ one option (+V ) and ‘short’ ∆ shares (−∆S).
The value of this portfolio is
Π = V − ∆S. (12.1)
Next, we find the number ∆ of shares that makes this portfolio risk-free over a time period dt.
To do so, we need to revisit Itô’s Lemma (3.1) to find out how the value of this portfolio Π
changes over time.
92
Example 12.2. Use Itô’s Lemma to show that the change in the portfolio’s value is

∂V 1 2 2 ∂2V ∂V ∂V
dΠ = + σ S + µS −∆ dt + σS − ∆ dW. (12.2)
∂t 2 ∂S 2 ∂S ∂S
Solution 12.2.
Over time dt, we assume that our strategy ∆ is constant. Then we have that change in the value
of the portfolio Π over the time interval dt is
dΠ = dV − ∆dS. (?)
We are assuming Geometric Brownian Motion for the share price, so we have
dS = µSdt + σSdW.
For the change in option value dV , we use Itô’s Lemma with f = V which gives

∂V ∂V 1 ∂2V ∂V
dV = + µS + σ 2 S 2 2 dt + σS dW.
∂t ∂S 2 ∂S ∂S
Substituting these two expressions for dS and dV into (?), we obtain

∂V ∂V 1 2 2 ∂2V ∂V
dΠ = + µS + σ S − ∆µS dt + σS − ∆σS dW.
∂t ∂S 2 ∂S 2 ∂S
Gathering terms with coefficients µ and σ, we find

∂V 1 2 2 ∂2V ∂V ∂V
dΠ = + σ S + µS −∆ dt + σS − ∆ dW.
∂t 2 ∂S 2 ∂S ∂S
∂V
Can we eliminate risk? Yes, we can choose ∆ = ∂S to eliminate the dW terms.
∂V
Example 12.3. Put ∆ = into (12.2).
∂S
Solution 12.3.
∂V
Putting ∆ = , our portfolio is
∂S
∂V
Π=V − S,
∂S
and (12.2) becomes
∂V 1 ∂2V
dΠ = + σ2 S 2 2 dt.
∂t 2 ∂S
Since there is no dW term now and the change in the value of portfolio only depends on the
current stock price S, we have that Π is a risk-free portfolio over the interval dt. Therefore,
investing in this portfolio over the short period dt is equivalent to investing in a bond.
93
Risk-free Return As discussed in section 6.2.4, a consequence of the no-arbitrage principle

is that the return on a risk-free portfolio over an interval dt is rdt, where r is the risk-free interest
rate. Since our portfolio Π is risk-free over the interval dt, its return is
dΠ
= rdt. (12.3)
Π
Example 12.4. Using equation (12.3) and Solution 12.3, derive the Black-Scholes equation:
∂V 1 ∂2V ∂V
+ σ 2 S 2 2 + rS − rV = 0. (12.4)
∂t 2 ∂S ∂S
Solution 12.4.
Solution 12.3 gives us expressions for Π and dΠ, so we substitute these into equation (12.3):

∂V 1 ∂2V
+ σ 2 S 2 2 dt
∂t 2 ∂S
= rdt.
∂V
V − S
∂S
Dividing by dt and multiplying by the denominator, we have

∂V 1 2 2 ∂2V ∂V
+ σ S =r V − S ,
∂t 2 ∂S 2 ∂S
and putting everything on the left-hand side, we obtain the Black-Scholes equation:
∂V 1 ∂2V ∂V
+ σ 2 S 2 2 + rS − rV = 0.
∂t 2 ∂S ∂S
Myron Scholes received the Nobel Prize for Economics in 1997 together
with Robert Merton, who was also instrumental in the derivation of
the model. Fischer Black had died in 1995 – Nobel Prizes are not
awarded posthumously.
Merton and Scholes
Example 12.5. Can we solve the Black-Scholes equation for the price of an option?
Solution 12.5.
In principle, the Black-Scholes equation can be solved to find the price of an option (or any
financial contract satisfying the assumptions in Section 12.1). The time boundary condition is
given by the payoff at t = T .
94
The equation itself is a backward parabolic partial differential equation, as it must be solved
backwards in time since the time boundary condition is the payoff at the expiry time.
There are analytical techniques to solve for the prices of some simple contracts, but many
cases require numerical methods.
For a call option, the time boundary condition is given by the payoff
C(S, t = T ) = max (S − E, 0).
An example solution is shown in figure 12.1.
C(S, t = 0)
C(S, t = T )
0
E 2E S
Figure 12.1: Plot of a European call option’s value against share price.
95
Chapter 13
Exact Solution to the

Black-Scholes Equation
13.1 Boundary Conditions

Boundary conditions are important when solving partial differential equations (PDEs) such as
the Black-Scholes equation (12.4), especially if you need to solve them numerically. In fluid
dynamics there is a variety of techniques to describe how a solution behaves near a boundary.
There are different types of boundaries, but most in finance are quite ‘passive’ – which is to say
that nothing dramatic happens near the boundaries.
European Call Option

Example 13.1. If the share price is extremely low, what is the likelihood that a call option will
be exercised? Deduce that the value of the call option when S = 0 is
C(S = 0, t) = 0. (13.1)
Solution 13.1.
In the extreme case, if S0 = 0 then dS = 0 in our continuous share price model, so St = 0

for all times t ≥ 0. Therefore the call option will not be exercised, so its value is zero at all times
t ≥ 0, i.e.
C(S = 0, t) = 0.
96
Example 13.2. If the share price is extremely high, what is the likelihood that a call option
will be exercised? Deduce that the value of the call option as S → ∞ is
C(S, t) ∼ S as S → ∞. (13.2)
Solution 13.2.
At expiry time T , a call option will be exercised if ST > E. Suppose that the share price at
some earlier time t is extremely high, St E, then we can say that it is extremely likely that we
will have ST > E later, so it is extremely likely that the call option will be exercised at time T .
In the limiting case, as S → ∞ the call option becomes certain to be exercised at time T , so
C(S, T ) = S − E ∼ S as S → ∞,
since E is negligible in comparison to S in this limit. Therefore the call option’s payoff is
asymptotically equivalent to a share, and hence the value of the option at earlier times is
C(S, t) ∼ S as S → ∞.
European Put Option

Example 13.3. In a similar way, show that the analogous conditions for a put option are
P (S = 0, t) = Ee−r(T −t) , (13.3)
and
P (S, t) → 0 as S → ∞. (13.4)
Solution 13.3.
As before, if S0 = 0 then dS = 0 in the continuous model, so St = 0 for all times t ≥ 0.

Therefore the put option will be exercised at time T with payoff P (S = 0, T ) = E − S = E.
Discounting backwards to time t, a payoff of E at time T is worth
P (S = 0, t) = Ee−r(T −t) ,
if the interest rate r is constant.
In the other extreme limit, as St → ∞ we can say that the put option is extremely unlikely
to be exercised at time T , resulting in an almost certain payoff of zero. Thus the value of the
put option at any earlier time t is zero in this limit:
P (S, t) → 0 as S → ∞.
97
13.2 Solution for a Call Option

For a call option C(S, t), the Black-Scholes equation is
∂C 1 ∂2C ∂C
+ σ 2 S 2 2 + rS − rC = 0,
∂t 2 ∂S ∂S
with time boundary condition
C(S, t = T ) = max (S − E, 0),
and share boundary conditions (13.1) and (13.2).
The solution is
C(S, t) = SN (d1 ) − Ee−r(T −t) N (d2 ), (13.5)
where N is the cumulative distribution function (CDF) of the standard normal distribution
Z x
1 2
N (x) = √ e−y /2 dy,
2π −∞
and where
ln (S/E) + (r + σ 2 /2)(T − t) √
d1 = √ , and d2 = d1 − σ T − t.
σ T −t
In this course we simply state this solution without derivation – although some aspects of it will
be justified. The CDF of the standard normal distribution has the following properties:
N (∞) = 1 (13.6)
N (−∞) = 0 (13.7)
N (0) = 1/2 (13.8)
N (−x) = 1 − N (x) (13.9)
These properties are important when we wish to investigate how the solution behaves when
certain parameters in the problem are varied. See Figure 13.1 for a plot of this solution (13.5).
98
E Maximum Value

Actual Value Speculative Value



Intrinsic Value



0
0 E S
Figure 13.1: A plot of the call option solution (13.5) (actual value). The difference
between the option’s actual value and its intrinsic value is called its speculative
value, also known as its time value.
Example 13.4. Why does the CDF of the standard normal distribution appear in the solution?
Solution 13.4.
By definition, the CDF of the standard normal distribution is
N (x) = P(X ≤ x) where X ∼ N (0, 1).
Using the given expressions for d1 and d2 , it can be shown that N (−d2 ) represents the risk-neutral
probability P(ST ≤ E) based on the log-normal share price distribution

ST
ln ∼ N (r − σ 2 /2)T, σ 2 T
S0
[Recall the log-normal distribution from Section 3.2. Here r replaces µ because we use risk-neutral valuation.]
Using property (13.9), N (d2 ) = 1 − N (−d2 ) represents the risk-neutral probability P(ST > E).
Therefore the second term in solution (13.5) may be interpreted as
−Ee−r(T −t) N (d2 ) = −Ee−r(T −t) × N (d2 ).

| {z } | {z }
exercise price discounted risk-neutral probability
backward to time t of exercising the option
The first term in solution (13.5) is bit harder to justify but the reasoning is essentially the same.
99
Example 13.5. The current share price is £21.60. Calculate the current price of a European
call option with exercise price £25 and expiry date in three months’ time. The volatility is 35%
and the risk-free interest rate is 1% p.a.
Solution 13.5.
Setting t = 0 as the current time, we have S0 = 21.6, E = 25, T = 0.25, σ = 0.35 and r = 0.01.
We calculate the current value of the call option, C(S = S0 , t = 0), using solution (13.5):
C0 = C(S = S0 , t = 0) = S0 N (d1 ) − Ee−rT N (d2 ). (?)
First we compute the values of d1 and d2 :
ln (S0 /E) + (r + σ 2 /2)T

d1 = √
σ T

ln (21.6/25) + 0.01 + (0.35)2 /2 × 0.25
= √ ≈ −0.733543,
0.35 × 0.25
√ √
=⇒ d2 = d1 − σ T ≈ −0.733543 − 0.35 × 0.25 ≈ −0.908543.
Next we use a calculator or a spreadsheet to evaluate the CDF:
N (d1 ) ≈ N (−0.733543) ≈ 0.231614, N (d2 ) ≈ N (−0.908543) ≈ 0.181796.
[For example, in Microsoft Excel N (x) can be evaluated via the in-built function =NORMSDIST(x).]
Altogether, (?) gives us
C0 ≈ 21.6 × 0.231614 − 25 × e−0.01×0.25 × 0.181796

≈ 0.4693.
Thus the current price of the call option is approximately £0.47.
Example 13.6. Find the high volatility limit of a call option’s value
lim C(S, t)
σ→∞
at any time t prior to expiry.
Solution 13.6.
From solution (13.5), we have
C(S, t) = SN (d1 ) − Ee−r(T −t) N (d2 ). (?)
100
Now, for any fixed t < T , consider the behaviour of d1 as σ → ∞:
ln (S/E) + (r + σ 2 /2)(T − t)
d1 = √
σ T −t
√ √
ln (S/E) r T − t σ T − t
= √ + +
σ T −t σ 2
√
σ T −t
=⇒ d1 ∼ → ∞.
2
Similarly
√
d2 = d1 − σ T − t
√
σ T −t
=⇒ d2 ∼ − → −∞.
2
Thus lim N (d1 ) = N (∞) = 1, and lim N (d2 ) = N (−∞) = 0, using properties (13.6) & (13.7).
σ→∞ σ→∞
Therefore, taking the limit in (?), we have
lim C(S, t) = S.
σ→∞
This says that, in the high volatility limit, the call option approaches its maximum possible
value: the upper bound C ≤ S.
101
Chapter 14
The “Greeks”
In practice, options are traded on the market – if you just want to know the current price of a
particular option, you don’t need to solve a partial differential equation... you can just check the
market to see how much the option is selling for. So why do we need the Black-Scholes model?
Recall the derivation of the Black-Scholes equation: the first step was to set up a portfolio
Π = V − ∆S,
and then choose ∆ appropriately in order to make Π risk-free. But how do we choose the
appropriate value of ∆ in the real world? We calculate what it should be by using the Black-
Scholes model.
102
Example 14.1. How might a trader use the Black-Scholes model in a real stock market?
Solution 14.1.
Observe data on the market
Calibrate model
parameters r and σ
Buy Option Create new option to sell
Calculate V using Black-

Scholes and sell option
Identify option
and buy/sell
Calculate ∆ using
Black-Scholes
Hedge
Profit
103
14.1 Calculating Delta

In order to trade options on the market, it is necessary to know how to calculate the appropriate
value of ∆. For a European call option, ∆ can be calculated by differentiating the exact solution
for C(S, t), given by (13.5), with respect to S.
Example 14.2. Differentiate (13.5) with respect to S to obtain
∂C ∂d
1
∆= = N (d1 ) + SN 0 (d1 ) − Ee−r(T −t) N 0 (d2 ) = N (d1 ). (14.1)
∂S ∂S
Solution 14.2.
Equation (13.5) gives the solution of the Black-Scholes equation for a European call option:
C(S, t) = SN (d1 ) − Ee−r(T −t) N (d2 ).
Differentiating with respect to S, we have

∂C ∂ −r(T −t)
∆= = SN (d1 ) − Ee N (d2 ) .
∂S ∂S
Recall that d1 and d2 are themselves functions of S, so we need to use the chain rule to differ-
entiate N (d1 ) and N (d2 ):
∂d1 ∂d2
∆ = N (d1 ) + SN 0 (d1 )− Ee−r(T −t) N 0 (d2 ) .
∂S ∂S
√
Now we note that, by definition, d2 = d1 − σ T − t so we have
∂d2 ∂d1
= .
∂S ∂S
Applying this gives us the required expression for ∆:
∂d
1
∆ = N (d1 ) + SN 0 (d1 ) − Ee−r(T −t) N 0 (d2 ) .
∂S
Thus we have obtained the first part of (14.1).
To obtain the second part of (14.1), it remains to show that
SN 0 (d1 ) − Ee−r(T −t) N 0 (d2 ) = 0,
and this is left as an exercise on Example Sheet 7. From this, we conclude that
∆ = N (d1 ).
104
Example 14.3. Calculate ∆ for a six-month European call option with exercise price equal to
the current share price. The risk-free interest rate is 6% p.a. and the volatility is 16%.
Solution 14.3.
Previously we showed that, for a European call option, ∆ is given by (14.1), i.e.
∆ = N (d1 ).
By definition,
ln (S/E) + r + σ 2 /2 (T − t)
d1 = √ .
σ T −t
Taking t = 0 as the current time (or initial time), the given information is T = 0.5 (six
months), E = S0 , r = 0.06, and σ = 0.16, so we have

ln (1) + 0.06 + (0.16)2 /2 × 0.5
d1 = √ ≈ 0.32173.
0.16 × 0.5
To calculate ∆ we evaluate N (d1 ) via computation (e.g. Excel, Wolfram Alpha) to find
∆ ≈ N (0.32173) ≈ 0.6262.
Note that the answer found above is the instantaneous value of ∆, appropriate at time t = 0.
As time progresses, ∆ must change accordingly because d1 is a function of t. The process of
adjusting ∆ continually with time is called delta-hedging.
Example 14.4. Find ∆ for a European put option by using the put-call parity relationship (6.6):
S + P − C = Ee−r(T −t) .
Solution 14.4.
Differentiating put-call parity (6.6) with respect to S, we find

∂P ∂C
1+ − = 0.
∂S ∂S
For a European put option, ∆ = ∂P/∂S, so rearranging, we have
∂P ∂C
∆= = − 1 = N (d1 ) − 1 = −N (−d1 ).
∂S ∂S
105
Example 14.5. Sketch the value of the call option and its corresponding ∆.
Solution 14.5.
C(S,t)
C(S,t) = SN(d1) - Eexp(-r(T-t))N(d2)
S - Eexp(-r(T-t)) S
� = N(d1)
1
∂C
Figure 14.1: Plots of the call option’s value C(S, t) and its ∆ = ∂S
.
106
Example 14.6. Sketch the value of the put option and its corresponding ∆.
Solution 14.6.
P(S,t)
P(S,t) = E exp(-r(T-t)) N(-d2) - SN(-d1)
E exp(-r(T-t)) - S S
� = N(d1) - 1
0
S
-1
∂P
Figure 14.2: Plots of the put option’s value P (S, t) and its ∆ = ∂S
.
107
14.2 “Greeks”
To highlight the dependence of an option’s value upon the variables and parameters in the
Black-Scholes model, we can use the notation
V = V ( S, t ; σ, r, T ).
|{z} | {z }
variables parameters
All of these variables and parameters may vary with time, so how do traders manage this risk? In
fact, in the same way that Black and Scholes eliminated the risk of the portfolio Π with respect
∂V
to changes in the share price S by hedging with ∆ = ∂S , the same can be done with other
types of risk by hedging with the partial derivative of V with respect to the relevant variable or
parameter.
As a consequence, the partial derivatives of the value function V with respect to the differ-
ent parameters have become extremely important to traders. Because the symbols associated
with these partial derivatives are (mostly) Greek letters, they became known collectively as
“the Greeks.”
Each “Greek” quantifies the sensitivity of an option’s current value V (or some function
of V ) with respect to changes in one of the variables or parameters upon which V depends.
• Delta: ∆ = ∂V
∂S measures the rate of change of the option’s value with respect to changes
in the underlying share price S.
• Theta: Θ = ∂V
∂t measures the rate of change of the option’s value with respect to time t.
∂2V
• Gamma: Γ = ∂S 2 = ∂∆
∂S measures the rate of change of ∆ with respect to changes in the
N 0 (d1 )
underlying share price S. On Example Sheet 7 we show that Γ = √
Sσ T −t
for a European
call option.
• Vega: V = ∂V
measures the sensitivity of V to the volatility σ. It can be shown that
∂σ
0
√
V = SN (d1 ) T − t for a European call option.
• Rho: ρ = ∂V
∂r measures the sensitivity of V to the risk-free interest rate r. It is straight-
forward to show that ρ = E(T − t)e−r(T −t) N (d2 ) for a European call option.
108
Chapter 15
More on Replicating Portfolios
15.1 Replicating Portfolios

If the payoff of a financial contract can be replicated by a portfolio of other financial products
that are available on the market, then we can find the value of that contract at an earlier time.
A replicating portfolio also allows banks to “lock in” risk-free profit at the point of sale, rather
than needing to wait until expiry. We will see an example of this later in this chapter, but first
let us consider how to set up a replicating portfolio strategy.
Example 15.1. What does it mean if a portfolio replicates a financial contract?
Solution 15.1.
Suppose a financial contract V and a portfolio Π have the same payoff at expiry time T , i.e.
VT = ΠT .
Then, by the no-arbitrage principle, they have the same value at any earlier time:
Vt = Πt for all t ≤ T.
Conversely, suppose the two investments have the same value initially (at time t = 0), i.e.
V0 = Π0 ,
and suppose further that we can make their increments equal as time progresses: dV = dΠ.
Then they again have the same value at all times:
Vt = Πt for all t ≤ T.
109
Our aim is to find a replicating portfolio, consisting only of shares and bonds, that can
reproduce the Black-Scholes equation.
Given some initial investment, we can set up a portfolio Π with ∆ shares held ‘long’ and
N bonds (of a given face value) held ‘short’ :
Π = ∆S − N B.
Recall that a pair (∆, N ) is called a trading strategy.
Now, recall the expressions for the continuous evolution of the share price and the bond value:
• SDE for share price S(t): dS = µSdt + σSdW ;
• Equation for bond value B(t): dB = rBdt.
Example 15.2. Let V (S, t) be the value at time t of a contract with underlying asset S(t). Use
a self-financing portfolio to replicate the contract and deduce the Black-Scholes equation
∂V 1 ∂2V ∂V
+ σ 2 S 2 2 + rS − rV = 0. (15.1)
∂t 2 ∂S ∂S
Solution 15.2.
Applying Itô’s Lemma to the contract V , the change in the contract’s value over time dt is

∂V ∂V 1 2 2 ∂2V ∂V
dV = + µS + σ S 2
dt + σS dW. (?)
∂t ∂S 2 ∂S ∂S
We define the portfolio Π = ∆S − N B as above; this is self-financing if no new money is put into
or withdrawn from the portfolio, so that any purchases or sales of S are financed via borrowing B.
Over time dt, we assume that the trading strategy (∆, N ) is held constant. Then we have
dΠ = ∆dS − N dB.
Now, using the SDE for share price and the equation for bond value, we find
dΠ = ∆(µSdt + σSdW ) − N rB dt = (∆µS − rN B)dt + ∆σSdW. (??)
For Π to replicate V we need dΠ = dV so, equating terms in dW and dt in (?) & (??), we obtain
∂V ∂V 1 ∂2V
∆= , and − rN B = + σ2 S 2 2 .
∂S ∂t 2 ∂S
Substituting these into the portfolio Π, we have

∂V 1 ∂V 1 ∂2V
Π = ∆S − N B = S + + σ2 S 2 2 .
∂S r ∂t 2 ∂S
Multiplying by r and noting that Π = V (by design), we deduce the Black-Scholes equation
∂V 1 ∂2V ∂V
+ σ 2 S 2 2 + rS − rV = 0.
∂t 2 ∂S ∂S
110
Example 15.3. Suppose that the risk-neutral valuation of a call option is C0 = £10, but a
trader is able to sell the same option for £11. How can the trader use a replicating portfolio to
“lock in” a guaranteed profit now, rather than waiting until expiry?
Solution 15.3.
At time t = 0, the trader sells the option for £11 and uses £10 of this money to set up the
replicating portfolio, with £1 left over. The initial value of the trader’s overall portfolio is
Π̃0 = −C0 + ∆S − N B0 + 1,
|{z} | 0 {z } |{z}
sell option replicating profit
portfolio
as the profit at time t = 0 is £1.
If the trading strategy (∆, N ) is correct, then ∆S − N B replicates the option’s value C, so
at any time t the value of the trader’s portfolio is
Π̃ = −C
+ ∆S
( (−
(N B
(( + ert .
This means that if the trader manages to sell the option today, he/she can receive the profit
straight away without having to wait months or years before the option’s expiry date.
15.2 Static Hedging

Recall the put-call parity relationship (6.6). When we derived this relationship, we set up a
portfolio Π which was long one share, long one put option, and short one call option, where both
options had the same expiry date and exercise price:
Π = S + P − C.
Example 15.4.
(a) What is the payoff of this portfolio at expiry?
(b) What is the risk associated with holding this portfolio?
Solution 15.4.
111
(a) The payoff of this portfolio at expiry time T is
ΠT = ST + PT − CT
= ST + max (E − ST , 0) − max (ST − E, 0)
(
ST + (E − ST ) if ST < E
=
ST − (ST − E) if ST ≥ E
=E ∀ ST .
(b) The payoff is the same regardless of the share price ST at t = T . Therefore the variance
of the payoff is zero, and hence the risk is zero for this portfolio (i.e. it is risk-free). This
is an example of complete risk elimination.
[To finish off the derivation of put-call parity as in Example 6.6:]

By the no-arbitrage principle, since Π is risk-free it accumulates value at rate r, so dΠ = rΠdt, and
therefore at any earlier time t its value is Πt = Ee−r(T −t) , which gives us put-call parity
St + Pt − Ct = Ee−r(T −t) .
The derivation of the put-call parity relationship is an example of complete risk elimination
via static hedging, where we set up the portfolio initially and it does not need to be adjusted
subsequently as time progresses. All financial transactions in the portfolio are made at the initial
time t = 0.
15.3 Dynamic Hedging

Now we consider a dynamic risk elimination procedure. Suppose we set up a portfolio which
is long one call option and short ∆ shares:
Π = C − ∆S.
As before, we can eliminate the randomness in dΠ (and hence the risk in Π) by choosing
∂C
∆= .
∂S
This is a delta-hedging strategy – it requires continuous re-balancing of the number of shares (∆)
in the portfolio as time progresses. Adjusting ∆ with time is an example of dynamic hedging.
Example 15.5. ∆ is a function of S and t, so how does this work in the real world?
Solution 15.5.
112
In theory, ∆ must be adjusted continuously in time for our portfolio to remain precisely
risk-free at all times. This would require financial transactions to be made infinitely often.
In practice we have to adjust ∆ discretely in time by trading at discrete time-steps, and so

our portfolio is only approximately risk-free at each time.
As a particular example, suppose we have a portfolio Π̃ in which we sell a call option and
hedge dynamically via a replicating portfolio ∆S − N B, so we have
Π̃ = −C + ∆S − N B. (?)
Example Sheet 7 includes a simulated scenario for future share and option prices at discrete
time-steps. Here we simplify the dynamic hedging procedure by imposing r = 0 so that the bond
value B is constant in time, wlog B = 1.
The initial data (at t = 0) are: S0 = 1, C0 = 0.07965, and ∆0 = 0.5398. If our portfolio Π̃
is to be self-financing then Π̃0 = 0, so (?) gives us
N0 = −C0 + ∆0 S0 = −0.07965 + 0.5398 × 1 = 0.46015,
i.e. we borrow £0.46015 to buy ∆0 shares so the portfolio is approximately risk-free at t = 0.
After the first time-step t = t1 , the share price changes to S(t1 ) = S1 = 1.138 according
to the given data. The value of the call option also changes to C1 = 0.1631 but we have not
adjusted (∆, N ) over the time interval t ∈ [0, t1 ]. Thus, at time t1 the value of our portfolio is
Π̃1 = −C1 + ∆0 S1 − N0 B1 = −0.1631 + 0.5398 × 1.138 − 0.46015 × 1 = −0.0089576,
so there has been a slight reduction in value (Π̃1 < Π̃0 = 0).
We now adjust (∆, N ) i.e. we buy/sell shares (S) and borrow/invest (B) accordingly. Our
portfolio is self-financing so Π̃1 is the same before and after the transactions at time t1 , i.e.
Π̃1 = −C1 + ∆0 S1 − N0 B1 = −C1 + ∆1 S1 − N1 B1 .
Rearranging this provides a formula to calculate N1 given the data for ∆1 :
N1 = N0 + (∆1 − ∆0 )S1 = 0.46015 + (0.66 − 0.5398) × 1.138 ≈ 0.5969.
After the second time-step t = t2 , S and C have changed again so we have at t = t2
Π̃2 = −C2 + ∆1 S2 − N1 B2 = . . .
and we again adjust (∆, N ) using the self-financing constraint to provide the formula
N2 = N1 + (∆2 − ∆1 )S2 = . . .
We continue this dynamic hedging process iteratively at each time-step (see Example Sheet 7).
113
Chapter 16
Extensions to the Black-Scholes

Model
16.1 Dividends
Dividends are payments made by a company to its shareholders at regular intervals (e.g.
annually), the amount per share being a proportion of the company’s profits.
In this chapter we consider how to represent dividends within a share price model, and how
to extend the Black-Scholes model to incorporate dividends.
Example 16.1. Suppose that a particular company is about to pay dividends. Each shareholder
will receive £10 per share held. What happens to the share price when the dividends are paid?
Solution 16.1.
Immediately after the dividends are paid, the share price will drop by £10.
To prove this, let the share price be S0 immediately before the dividends are paid out, and S0+
immediately afterwards. Suppose that we buy a share just before the payout, so our portfolio is
Π = S − B,
where B0 = S0 is the amount borrowed to buy the share (so Π0 = 0). We then receive our
dividend payment of £10, so just afterwards our portfolio is worth
Π0+ = S0+ − B0+ + 10.
114
To avoid an instantaneous arbitrage opportunity, we must have Π0+ = Π0 which means
S0+ − B0+ + 10 = S0 − B0 ,
and from this we obtain the change in the share price over this instant in time:
S0+ − S0 = B0+ − B0 − 10 = −10.
Thus the share price decreases by £10 over this instant in time.
What the previous example shows is that there is a distinction between the change in the
share price on the market (which drops by £10) and the value of actually holding the share in a
portfolio (which doesn’t change because the holder receives the dividend).
In reality, dividends are paid at regular intervals, but we will simplify things mathematically
by modelling dividends as continuous-time rather than discrete-time payments. We suppose that
each share pays its holder a continuous dividend at a rate proportional to the share price S, so
that over a time increment dt the dividend is
D0 S dt,
where D0 is a constant called the dividend yield.
Example 16.2. Suppose that we have a portfolio consisting of a single share
Π0 = S0 .
If the share pays a continuous dividend, how does the value of this portfolio change in time?
Solution 16.2.
Over a time increment dt, the change in the value of this portfolio is
dΠ = dS
|{z} + cash .
|{z}
change in dividend
market price payment
The dividend received (as cash) over time dt is D0 S dt where S is the current share price. If we
use this cash to buy more shares (rather than investing it in the bank), we have
dΠ = dS + D0 S dt.
The key point here is that the value of holding a share in a portfolio grows slightly faster than
the share price grows, due to the dividend.
115
To incorporate this effect in our share price model, we can write
dS = µ̂Sdt + σSdW,
where µ̂ is the expected growth rate of the share price including dividends. Again considering
the portfolio Π0 = S0 , over a time increment dt the change in the value of Π is
dΠ = µ̂Sdt + σSdW + D0 Sdt,
and rearranging this we have

dΠ = (µ̂ + D0 )Sdt + σSdW.
In order to compare investments in shares which don’t pay dividends with investments in
shares which pay dividends (assuming the dividend payments are re-invested in shares), we set
µ̂ = µ − D0 ,
so that the dividend payments are incorporated, and then µ represents the real growth rate.
The term adjusted price is used on the stock market to refer to the share price taking
dividends into account, along with other practical aspects∗ that we do not consider here.
[∗ : for example, a share split which is an increase in the total number of shares in a particular company, causing
a reduction in the price of individual shares.]
16.2 Options on Dividend Paying Shares

In this section we derive the modified Black-Scholes equation for the value V (S, t) of an option
written on a dividend paying share. The steps are analogous to the derivation of the Black-Scholes
equation given in chapter 12.
First, we set up a portfolio Π consisting of a long position in the option V and a short
position in ∆ shares:
Π = V − ∆S.
Example 16.3. Show that the change in value of this portfolio in the time interval dt is
dΠ = dV − ∆dS − ∆D0 Sdt. (16.1)
Solution 16.3.
Following the ideas in Example 16.2, the change in value of the portfolio Π = V − ∆S is
dΠ = dV − ∆ (dS + D0 S dt) ,
116
except that here the shares are short in our portfolio, so the dividend payment represents cash
that we owe, and we finance this by selling more shares, whereas in Example 16.2 we used the
dividend cash to buy more shares. Thus the change in value of Π over time dt is
dΠ = dV − ∆dS − ∆D0 Sdt.
Example 16.4. Using Itô’s Lemma written in the form:

∂V 1 2 2 ∂2V ∂V
dV = + σ S dt + dS, (16.2)
∂t 2 ∂S 2 ∂S
derive the modified Black-Scholes equation for the option’s value V (S, t):
∂V 1 ∂2V ∂V
+ σ 2 S 2 2 + (r − D0 )S − rV = 0. (16.3)
∂t 2 ∂S ∂S
Solution 16.4.
Substituting (16.2) into (16.1) and gathering terms in dt and dS, we obtain

∂V 1 ∂2V ∂V
dΠ = dV − ∆dS − ∆D0 Sdt = + σ 2 S 2 2 − ∆D0 S dt + − ∆ dS.
∂t 2 ∂S ∂S
As in chapter 12, to eliminate the random component (given by dW within dS) we choose
∂V
∆= .
∂S
∂V
This choice of ∆ results in a risk-free portfolio Π = V − S
with increment
∂S

∂V 1 ∂2V ∂V
dΠ = + σ 2 S 2 2 − D0 S dt.
∂t 2 ∂S ∂S
By the no-arbitrage principle, the return from a risk-free portfolio must be rdt so we have
dΠ
= rdt
Π
and from this we obtain the equation

∂V 1 ∂2V ∂V ∂V
+ σ 2 S 2 2 − D0 S =r V −S .
∂t 2 ∂S ∂S ∂S
Finally, we rearrange this to find the modified Black-Scholes equation (16.3):
∂V 1 ∂2V ∂V
+ σ 2 S 2 2 + (r − D0 )S − rV = 0.
∂t 2 ∂S ∂S
117
How can we adapt the solutions we have already to satisfy this modified Black-Scholes
equation? The modified equation (16.3) differs from the original equation (12.4) only in the
sense that r has been replaced by (r − D0 ) in one of the two terms in which it was a coefficient in
the original equation. This suggests that an appropriate integrating factor could equalize these
two coefficients.
To confirm this, we can write the modified Black-Scholes equation (16.3) as

∂V 1 ∂2V ∂V
− D0 V + σ 2 S 2 2 + (r − D0 )S − (r − D0 )V = 0,
∂t 2 ∂S ∂S
and now we combine the two terms in square brackets into a single time derivative by multiplying
throughout by the integrating factor ke−D0 t where k is an arbitrary constant:

∂ V ke−D0 t 1 2 2 ∂ 2 V ke−D0 t ∂ V ke−D0 t
+ σ S + (r − D0 )S − (r − D0 ) V ke−D0 t = 0.
∂t 2 ∂S 2 ∂S
This is now in the form of the original Black-Scholes equation with r replaced by (r − D0 ). Thus
if V1 (S, t; r) satisfies the original Black-Scholes equation then V (S, t; r) = k −1 eD0 t V1 (S, t; r − D0 )
satisfies the modified Black-Scholes equation (16.3). Further, if we choose k = eD0 T then V and
V1 are equal at expiry time T so the time boundary condition does not need to be adapted.
In summary, if V1 (S, t; r) satisfies the original Black-Scholes equation (12.4) then
V (S, t; r) = e−D0 (T −t) V1 (S, t; r − D0 )
satisfies the modified Black-Scholes equation (16.3) with the same boundary conditions.
In particular (see Example Sheet 8), the modified solution for a European call option is
C(S, t) = Se−D0 (T −t) N (d10 ) − Ee−r(T −t) N (d20 ), (16.4)
where
ln (S/E) + r − D0 + σ 2 /2 (T − t)
d10 = √ ,
σ T −t
and
√
d20 = d10 − σ T − t .
118
16.3 Early Exercise

Recall that an American option is an option that may be exercised at any time prior to
expiry. In chapter 11 we discussed the implications of this in the context of a discrete-time
binomial model, but what happens in the continuous time model?
The value of an American call option C(S, t) is the same as the value of a European call
option with the same exercise price and expiry date, as shown in question 4 of Example Sheet 5.
However, for an American put option P (S, t), there is no analytical solution although we still
have condition (11.1):
P ≥ E − S.
For more general derivatives, in the continuous model the problem of option valuation can
be reformulated and solved in several different ways:
• Optimal Stopping Problem: this formulation is popular with researchers in probability

theory. The problem is stated as calculating the optimal time at which to exercise the
option (and hence to stop holding the option). To obtain a numerical value, however, one
of the following two methods must be used.
• Variational Inequalities: this formulation is the most robust way to express the prob-
lem, and there are many numerical techniques (but very few analytical ones) available to
solve problems of this type.
• Free Boundary: this formulation is popular with applied mathematicians, as it means

that analytical solutions can be derived in some special cases, and that many techniques
from other fields (such as fluid mechanics) can be used. However, this formulation is not as
robust, since assumptions have to be made about the existence of the ‘exercise boundary’
and numerical solutions can be difficult to code, although they are very accurate.
In summary, formulating the problem and then solving for the value of a contract can be
very difficult as there are often no explicit analytical solutions.
Example 16.5. Reformulate the value of an American put option as a free boundary problem.
Solution 16.5.
The idea is to partition the share price axis S into two disjoint intervals:
0 ≤ S ≤ Sf (t) and Sf (t) < S < ∞,
where Sf (t) is the exercise boundary. Note that we do not know a priori what Sf (t) is.
119
• When S < Sf (t), early exercise of the option is optimal and so
P (S, t) = E − S.
• When S > Sf (t), early exercise is not optimal so P satisfies the Black-Scholes equation
∂P 1 ∂2P ∂P
+ σ 2 S 2 2 + rS − rP = 0.
∂t 2 ∂S ∂S
Therefore the problem involves solving the Black-Scholes equation with boundary conditions
P (Sf (t), t) = E − Sf (t) and P → 0 as S → ∞.
Further, by no-arbitrage arguments we can also establish what is often referred to as the “smooth
pasting condition,” which is that ∆ = ∂P/∂S should be continuous across the exercise boundary,
i.e.
∂P ∂
lim = lim (E − S) = −1.
S&Sf ∂S S%Sf ∂S
120
Chapter 17
Bonds and Interest Rates
17.1 Time Varying Interest Rates

Bond : a contract which entitles the holder to receive a known amount (the face value, also
called nominal, principal, or redemption value) of money from the issuer at the maturity date
t = T (also called the redemption date). The holder may also be entitled to receive coupons
(interest payments) at fixed times, e.g. annually. If there is no coupon payment, then it is called
a zero-coupon bond.
A bond with face value $20, maturity date 15th November 1987, and coupon rate 5.25% p.a.
We will use the following notation:
• V (t) is the value of the bond at time t;
• r(t) is the interest rate, which is now a function of time t;
• K(t) is the coupon rate, modelling the coupon payments as continuous in time.
121
Example 17.1. Show that the value of a bond with coupon rate K(t), maturity date T , and
face value F satisfies the differential equation
dV
= r(t)V − K(t), (17.1)
dt
with the boundary condition
V (T ) = F.
Solution 17.1.
Over a small time increment dt, we can say:
(change in bond value) = (risk-free interest on current value) − (coupon paid out).
In our notation this corresponds to the differential equation
dV
|{z} = r(t)V (t)dt − K(t)dt,
| {z } | {z }
change in value interest coupon payment
which we can write as

dV
= r(t)V − K(t).
dt
The boundary condition V (T ) = F follows because the holder receives the face value F at the
maturity date T .
Note that (just as in the case of a share with dividend payments), the coupon payments lead
to a decrease in the value of the bond on the market as time goes on. However, a bond which
pays coupons will be more valuable than a zero-coupon bond of the same face value.
Example 17.2. Sketch the value of a bond as a function of time in the following cases:
1. r(t) > 0 and K(t) = 0 (zero-coupon bond);
2. r(t) = r0 > 0 and K(t) = K0 > r0 F , where r0 and K0 are constants;
3. r(t) = r0 > 0 and K(t) represents a single payment of K0 at time t = T /2.
Solution 17.2.
122
1. r(t) > 0 and K(t) = 0 (zero-coupon bond);

In this case the bond’s value increases with time up to maturity when V (T ) = F . If r is
constant then the growth is exponential.
V (t)
F
0 t
T
2. r(t) = r0 > 0 and K(t) = K0 > r0 F , where r0 and K0 are constants;

In this case dV /dt < 0 at time T , so the bond’s value decreases with time.
V (t)
0 t
T
3. r(t) = r0 > 0 and K(t) represents a single payment of K0 at time t = T /2.

Except at t = T /2, the bond’s value behaves as in case 1 above, increasing exponentially
with time. At t = T /2 there is a jump of size K0 in the bond’s value.
V (t)
F
0 t
T /2 T
123
17.2 Zero-Coupon Bonds

Let us now consider the case when the coupon rate is zero, i.e. K(t) = 0, and the interest rate r(t)
varies with time. Then the bond value equation (17.1) becomes
dV
= r(t)V. (17.2)
dt
Example 17.3. Solve (17.2) with boundary condition V (T ) = F to find
Z T !
V (t) = F exp − r(s)ds . (17.3)
t
Solution 17.3.
Equation (17.2) is separable. Using a dummy variable s, we write

dV
= r(s)ds,
V
and then we can integrate directly with respect to s with limits t and T :
T Z T
ln V (s) = r(s)ds.
t t
Applying the boundary condition, we have

Z T
ln V (t) = ln F − r(s)ds,
t
and taking the exponential gives us (17.3)

Z !
T
V (t) = F exp − r(s)ds .
t
Example 17.4. A particular zero-coupon bond has face value F = 1 and maturity date T = 1.
Find the bond’s price V (t) at time t < 1 and hence find V (0), if the interest rate is r(t) = t2 .
Solution 17.4.
Applying equation (17.3) with T = 1, F = 1, and r(s) = s2 , we have

Z 1
1
V (t) = exp − s2 ds = exp − (1 − t3 ) .
t 3
Therefore
V (0) = exp (−1/3) ≈ 0.7165.
124
17.3 Coupon Bonds

We now consider the case when the coupon rate K(t) is non-zero. In this case, we can solve (17.1)
using the integrating factor
Rt
e− r(s)ds
. (17.4)
Note that we use the dummy variable s inside the integral, since t appears in the limits.
Example 17.5. Using the integrating factor (17.4), solve the bond value equation (17.1) with
boundary condition V (T ) = F to find
RT
Z T Rs
V (t) = F e− t
r(u)du
+ e− t
r(u)du
K(s)ds (17.5)
t
Solution 17.5.
First we multiply equation (17.1) by the integrating factor (17.4):
r(s)ds dV
Rt Rt Rt
e− − e− r(s)ds
r(t)V = −e− r(s)ds
K(t),
dt
and we group the left-hand side into a single derivative (by the product rule)
d n − R t r(s)ds o Rt
e V (t) = −e− r(s)ds K(t).
dt
Next we will integrate, but as we want t in the limits we should use dummy variables throughout:
d n − R s r(u)du o Rs
e V (s) = −e− r(u)du K(s).
ds
We now integrate with respect to s with limits t and T to obtain
R T Z T
s Rs
e− r(u)du V (s) = − e− r(u)du
K(s)ds.
t t
Applying the boundary condition V (T ) = F gives us

RT Rt
Z T Rs
− −
Fe r(u)du
−e r(u)du
V (t) = − e− r(u)du
K(s)ds,
t
and then we rearrange and combine u-integrals to find

RT
Z T Rs
−
V (t) = F e t
r(u)du
+ e− t
r(u)du
K(s)ds.
t
125
Chapter 18
Term Structure of Interest Rates
18.1 Yield
Suppose that there are several risk-free zero-coupon bonds on sale with different maturity dates
and face values, and they are priced as follows:
Current Time t Maturity Date T Face Value F = V (T, T ) Current price V (t, T )
0 1 £100 £98.24
0 2 £250 £244.47
0 3 £100 £96.43
0 5 £1,000 £939.00
0 10 £500 £448.95
Here we use the notation V (t, T ) for the price at time t of the bond with maturity date T . The
table shows the bond prices at the current time t = 0 for different values of T . If we update
the same table in six months’ time, the t-column will change to 0.5 and the bond prices will all
change, but the maturity dates T are fixed.
Investors require a convenient way to compare different bonds and hence to compare the
returns on their investments. Since it is not the actual price of a bond that is important, but
rather the expected return on an investment, investors will often compare the yield.
Yield to Maturity: for zero-coupon bonds, the yield to maturity is defined as

ln V (t, T ) − ln V (T, T )
Y (t, T ) = − , (18.1)
T −t
and it gives a measure of the future interest rate, where V (t, T ) can be taken from market data.
For a bond with coupons the yield is more complicated and we will not study it in this course.
126
Example 18.1. Calculate the yields Y (0, 2), Y (0, 3), and Y (0, 5) from the previous table.
Solution 18.1.
ln (244.47) − ln (250)
Y (0, 2) = − ≈ 0.011184 ≈ 1.12%,
2
ln (96.43) − ln (100)
Y (0, 3) = − ≈ 0.012118 ≈ 1.21%,
3
ln (939) − ln (1000)
Y (0, 5) = − ≈ 0.012588 ≈ 1.26%.
5
By formula (17.3) for the price of a zero-coupon bond in terms of the interest rate r(t), we have
n R o
T
ln F exp − t r(s)ds − ln F
Y (t, T ) = − ,
T −t
and combining the logarithms we see that
Z T
1
Y (t, T ) = r(s)ds. (18.2)
T −t t
This makes it clear that Y (t, T ) represents the average interest rate over the time interval [t, T ].
The bond price can also be written as
V (t, T ) = F e−Y (t,T ) (T −t) . (18.3)
Example 18.2. A zero-coupon bond with face value £1,000 matures in five years’ time. Using
the table, calculate its value in three years’ time and hence calculate the future yield Y (3, 5).
Solution 18.2.
We are asked for V (3, 5) but the table only gives us V (0, 3) and V (0, 5). By the no-arbitrage
principle, the interest accumulated on a risk-free investment from t = 0 to t = 5 must be the
same as that on a risk-free investment from t = 0 to t = 3 followed by another from t = 3 to
t = 5. Thus we have
V (5, 5) V (3, 3) V (5, 5)
= × ,
V (0, 5) V (0, 3) V (3, 5)
and we can rearrange to find
V (0, 5) × V (3, 3) 939 × 100
V (3, 5) = = ≈ 973.76.
V (0, 3) 96.43
127
[Note that the table does not give us enough significant figures to be confident about the nearest penny here.]
From the definition of Y (t, T ) we have
ln V (3, 5) − ln V (5, 5) ln (973.76) − ln (1000)

Y (3, 5) = − =− ≈ 0.0133 ≈ 1.33%.
5−3 2
From the above example we see that the future yield can be calculated from the current
bond prices, and this gives us a measure of the average interest rate over the next few years.
But what about the interest rate at a particular time in the future?
Suppose that the market data include bond prices for all maturity dates T , so that V (t, T )
is known for all T ∈ (t, ∞). What is the relationship between the quoted bond prices and the
future interest rate?
Example 18.3. By differentiating equation (17.3), show that the forward interest rate is
given by
1 ∂V (t, T )
r(T ) = − . (18.4)
V (t, T ) ∂T
Solution 18.3.
Starting from equation (17.3)

Z !
T
V (t, T ) = F exp − r(s)ds ,
t
we differentiate with respect to T :

Z !
T
∂V (t, T )
= F exp − r(s)ds × −r(T ) = −V (t, T ) r(T ).
∂T t
Therefore we have
1 ∂V (t, T )
r(T ) = − .
V (t, T ) ∂T
This is the interest rate at the future date T , i.e. the forward rate.
128
18.2 Term Structure of Interest Rates

We define the term structure of interest rates (also referred to as the yield curve) as the
graph of the time-averaged interest rate over the future time interval [0, T ] as a function of T .
From (18.2) with t = 0, this average is given by the yield
Z
1 T
Y (0, T ) = r(s)ds,
T 0
and we plot Y (0, T ) against T .
A typical application (in this course) is to assume that we have already modelled the future
interest rate itself as a function of time r(t), so that we can calculate bond prices, yields, and
plot the term structure of interest rates. However in practice it often works vice versa: we obtain
a set of prices V (t, T ) from market data, and we must find a suitable function r(t) which models
the corresponding time-dependent interest rate.
Example 18.4. Suppose that the interest rate at time t is given by
r(t) = r0 + at,
where r0 and a are positive constants. Calculate the value at time t of a zero-coupon bond with
maturity date T and face value F , and sketch the term structure of interest rates.
Solution 18.4.
By (17.3) the bond value formula is

Z ! Z !
T T
2
−t2 )
= F e−r0 (T −t)− 2 (T
a
V (t, T ) = F exp − r(s)ds = F exp − (r0 + as)ds .
t t
The yield to maturity is

Z T
1 1 aT 2 aT
Y (0, T ) = r(s)ds = r0 T + = r0 + ,
T 0 T 2 2
so the term structure of interest rate is thus a straight line with gradient a/2.
Y (0, T )
r0
0 T
129
18.3 Default Risk

So far we have referred to bonds as risk-free investments. In practice, there is always some
element of risk (usually very small) that the issuer cannot pay the holder the face value at
maturity as promised in the contract. This is called risk of default.
If the risk of default is non-negligible, then the holder should expect to pay less for the bond
and therefore the yield should be higher. This increase in the yield is called the yield spread.
We can show how the yield spread is related to probability of default.
Example 18.5. Suppose that the interest rate r is constant. Consider a 1-year bond of face
value F and current value V (0, 1), and let p be the probability of default. Show that the yield
spread is approximately equal to p.
Solution 18.5.
Here we have a simple binomial situation where the bond’s value at maturity (T = 1) is either
the face value F (with probability 1 − p) or zero (with probability p).
Bond value tree: If the bond was risk-free, the yield would be
Z
u0 1 1
p r(s)ds = r,
1 0

u
V (0, 1) H

H using (18.2). The yield spread s is the increase in the yield
HH
1 − p HHu due to risk of default, i.e.
F
s = Y (0, 1) − r.
Based on the tree, the bond’s expected value at maturity is (1 − p)F , so its current value is
V (0, 1) = e−r (1 − p)F,
and therefore by (18.1) if there is no default then the yield is
ln (e−r (1 − p)F ) − ln F
Y (0, 1) = − = r − ln (1 − p).
1
Hence the yield spread is
1
s = − ln (1 − p) = p + p2 + · · · ≈ p,
2
where we have used a Taylor series for small p.
130
Chapter 19
Exotic Options
Example 19.1. Given that this is a course on financial mathematics, is there anything I have
learned that will make me rich?
Solution 19.1.
What you have learned in this course will not necessarily make you rich. Apologies for waiting
until Chapter 19 to make this clear.
• European and American options are traded in very efficient markets.
• To make money, banks will work with high volume and low margin.
• The main assumption is that there is no arbitrage, and therefore no way to guarantee to
make money from nothing. Hence risks must be taken.
How can we become rich? – ownership, good luck (gambling)∗ , good timing, being
ahead of the competition (e.g. inside/extra information) . . .
∗ : this is also how to become poor.
European and American call and put options are often referred to as “plain vanilla options”
Other types of option are usually called exotic options. An exotic option might include special
clauses in its contract, or have a more complicated payoff that makes it difficult to calculate its
current value. Whereas call and put options are traded in efficient markets, exotic options are
often traded over the counter (OTC) which means that two participants must agree their own
price for a trade to take place. The participants must calculate the price themselves, and the
price is not necessarily made public.
131
Example 19.2. Why might exotic options present an opportunity to make money?
Solution 19.2.
By selling an exotic option, we might be able to:
• attract new business;
• insert clauses to our advantage;
• dictate the price.
The buyer of an exotic option may have a very different idea of what the current value (and
hence price) should be, or perhaps they might not even fully understand what they are buying.
Some examples of exotic options are:
• Asian – exercise price based on the average share price over a given time interval;
• Lookback – exercise price based on the maximum/minimum over a given time interval;
• Bermudan – can be exercised on particular dates only;
• Barrier – expires if the share price goes above/below a threshold value;
• Parisian – payoff depends how long the share has been above/below the exercise price;
• ParAsian – cross between Parisian and Asian;
• Compound – an option on another option (hence two expiry dates).
Often these options can be combined or included in other financial contracts.
Pricing exotic options is difficult in general, and further adaptations or extensions must be
made to the models we have considered. In some cases the payoff depends on historical data and
not only on the current share price – this is handled by the introduction of new variables.
19.1 Asian Options

Asian Option: a contract giving the holder the right to buy or sell an underlying asset (e.g.
a share) for its average price over a prescribed time interval.
Example 19.3. What is the average of the share price S over the time interval [0, T ]?
132
Solution 19.3.
If the share price at time t is S(t) then the average price over the period t ∈ [0, T ] is
Z T
1
A(T ) = S(t)dt.
T 0
Note that this integral converges with probability one by the continuity of the Wiener process
(one of the properties mentioned in Section 2.2).
If we have only discrete data available for S (say if we are looking back at historical data to
calculate the average), we could approximate the integral numerically via the trapezium rule or
Simpson’s rule etc.
A floating strike Asian option uses the share price S as the underlying asset and the average
share price A(T ) as the exercise price (recall that strike price is another term for exercise price).
For example, a floating strike Asian put option has payoff
Z !
T
1
V (S, T ) = max (A(T ) − S, 0) = max S(t)dt − S, 0 .
T 0
On the other hand, a fixed strike Asian option uses A(T ) as the underlying and E as the exercise
price, where E is agreed in advance.
Example 19.4. What is the payoff for a fixed strike Asian call option?
Solution 19.4.
Z !
T
1
V (S, T ) = max (A(T ) − E, 0) = max S(t)dt − E, 0 .
T 0
19.2 Black-Scholes for Asian Options

In order to adapt the Black-Scholes model to handle Asian options, we introduce a new variable
Z t
dI
I(t) = S(u)du such that = S(t).
0 dt
133
Note that at payoff (t = T ) we have I = T ×A(T ), so the boundary condition for the option’s
value at payoff can be expressed in terms of S, I, and T . For example, for a floating strike Asian
put option the payoff condition is

I
V (S, I, T ) = max − S, 0 .
T
We can now adapt the Black-Scholes model, treating the option’s value V as a function of
three variables V (S, I, t).
Example 19.5. Using an adapted version of Itô’s Lemma:

∂V 1 2 2 ∂2V ∂V ∂V
dV = + σ S 2
dt + dS + dI, (19.1)
∂t 2 ∂S ∂S ∂I
set up a hedging portfolio to eliminate risk and hence obtain the modified Black-Scholes equation
for an Asian option
∂V 1 ∂2V ∂V ∂V
+ σ 2 S 2 2 + rS − rV + S = 0. (19.2)
∂t 2 ∂S ∂S ∂I
Solution 19.5.
We start with the same Black-Scholes hedging portfolio as previously
Π = V − ∆S.
By (19.1), the change in the value of Π over the time increment dt is

∂V 1 2 2 ∂2V ∂V ∂V
dΠ = + σ S 2
dt + − ∆ dS + dI.
∂t 2 ∂S ∂S ∂I
The risk in the portfolio comes from the random element dW in the share price increment dS,
so we can eliminate risk by choosing
∂V
∆= .
∂S
As before, a risk-free portfolio grows in value at rate r so we have dΠ = rΠdt, which gives us

∂V 1 2 2 ∂2V ∂V ∂V
+ σ S dt + dI = r V − S dt.
∂t 2 ∂S 2 ∂I ∂S
Rearranging, and noting that dI = Sdt, we obtain the modified Black-Scholes equation for the
value of an Asian option:
∂V 1 ∂2V ∂V ∂V
+ σ 2 S 2 2 + rS − rV + S = 0.
∂t 2 ∂S ∂S ∂I
Although this equation looks very similar to the Black-Scholes equation (12.4), it is much
harder to solve analytically. The value of an Asian option must be calculated numerically.
134

Math20912 Notes

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Math20912 Notes

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Math20912 Notes

Uploaded by

Copyright:

Available Formats

Dr Peter Johnson

MATH20912 Introduction to Financial Mathematics

• The mid-term test will consist of an online test.

• Final exam: contributes 80% of the total mark.

• The final exam format will be a traditional invigilated in person exam.

• Asynchronous lectures will be made available the week before on Blackboard, to be

• P. Wilmott, S. Howison, and J. Dewynne, “The Mathematics of Financial Derivatives: A

1 Introduction to the Course 1

2 The Wiener process and Modelling Stock Price 9

5 Trading with Portfolios 39

5.2.3 Bull Spread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

6 Risk and No-Arbitrage 50

9 Two-Step Binomial Trees 71

10 The Multi-step Binomial Model 77

11 American Options and Replicating Portfolios 84

12 The Black-Scholes Model 91

13 Exact Solution to the Black-Scholes Equation 96

14 The “Greeks” 102

15 More on Replicating Portfolios 109

16 Extensions to the Black-Scholes Model 114

17 Bonds and Interest Rates 121

18 Term Structure of Interest Rates 126

19 Exotic Options 131

Introduction to the Course

1.1 Background Economics

See figure 1.1.

Stock Exchange is where stocks (shares) are bought and sold.

Value is a measure of worth to an individual. In almost any context it is extremely subjective,

Boris Johnson will underlying

Figure 1.1: An example financial contract.

£10 now > a promise of £10 in the future

• Continuously Compounded Interest Rate For many reasons (mainly practical) it is

• a simple interest rate with r = 0.05 paid at t = 2

• a compound interest rate with r = 0.05 in two payments per year

• a continuously compounded interest rate with r = 0.05

V (t) = P × (1 + rt) = 100 × (1 + 0.05 × 2) = 110 .

• Continuously Compounded Interest Rate P = 100 is the initial investment, r = 0.05

V (t) = 100 e0.05×2 = 110.52 .

1.3 Modelling the Share Price

A Stochastic Model for Shares

In words, our model (1.1) can be written as

• Normally distributed, over a long run.

The Wiener process and

• ∆t is a small but finite change in time;

Adding normal distributions together

then if we define Z as their sum, i.e.

2.1 Properties of SDEs

2.1.1 Random Walk

See figure 2.1.

Now if we write the discrete random walk (2.3) as

it is trivial to show that

W (t) ∼ N (0, t).

Figure 2.1: Constructing a discrete random walk.

2.2 Properties of the Wiener Process

• W (s + t) − W (s) ∼ N (0, t);

• W (t) has continuous paths, i.e. W (t) is continuous in t.

Example 2.2. Show that the following results hold:

and since W (0) = 0 we get

and its variance is

From (3.2), we can deduce that the

From (3.2) we have