Main

Mathematical Finance Lecture Notes 2024-25
Clare Wallace
2024-11-28
Contents
1 Introduction 2
1.1 What is Mathematical Finance? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Financial background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Discrete-time model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2 The one-period binomial model 12

2.1 The model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2 Portfolios and arbitrage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Contingent claims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4 The martingale measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 Risk neutral valuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.6 Some generalisations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3 The multi-period binomial model: Part 1 19

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Portfolios and the self-financing condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3 The martingale measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.4 Pricing European contingent claims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.5 Recombinant trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.6 The Cox–Ross–Rubinstein formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4 Filtrations, Conditional Expectation, and Martingales 30

4.1 The probability space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.2 Partial information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.3 Conditional expectation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.4 Martingales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5 Pricing options from around the world 37

5.1 An algorithm to price any derivative asset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.2 Popular options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.3 American Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6 The Black–Scholes formula 44

6.1 Non-binomial prices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1
§1 Mathematical Finance Michaelmas 2024
1 Introduction
These lecture notes are based on the content we will cover in Mathematical Finance in Michaelmas term, 2024 - 25.
They are based on previous sets of notes put together by Nic Georgiou, Andrew Wade, and others.
1.1 What is Mathematical Finance?

Mathematical Finance is the study of the mathematics used to model and analyse financial markets. These models
are constructed to try to better understand how markets behave in reality, and to inform decisions about investments.
In reality, these markets are incredibly complex, but under some simplifying assumptions, the mathematics becomes
quite elegant, and allows us to develop methods for pricing and valuing portfolios based on a wide range of financial
derivatives.
In Michaelmas term, we’ll focus on discrete-time versions of these models, where we assume that trades can only
happen at specified moments. We’ll build up a theory using probabilistic concepts like filtrations, conditional
expectation, and martingales. Because discrete-time models lead to countable probability spaces, we’ll be able to
do a lot of this work in very concrete settings, and calculate the prices of some quite complex financial products.
In Epiphany term, you’ll see the continuous-time versions of the models. Here, the continuous (uncountable!)
probability spaces mean that the theory becomes much more complex and subtle, and some of the generalisations
to continuous time require some pretty sophisticated measure theory. The understanding of concrete fundamental
concepts you build up in Michaelmas term will put you on a solid footing to start working with the more abstract
theory to come in Epiphany.
A note on financial models
The model of a financial market used in this course is essentially a Western one. Our departmental decolonisation
interns have written an introduction to another major financial system, Islamic finance, which you can find here.
For more information on research into Islamic Finance at Durham, see the work done at the Durham Centre for
Islamic Economics and Finance.
1.1.1 Finance or probability?

I consider myself a probabilist, and that’s the approach I’m bringing to this course. I see it as being about the
mathematics of finance, rather than the economics, and we’ll be coming at a lot of the material from a probability
perspective. That said, we’ll need to use some financial terminology to describe the concepts we’re using, and many
of the examples in the course will be from a financial context.
2
1.2 Financial background

The main focus of this course is on financial derivatives, and how they should be priced and can be hedged. In this
chapter, we set up the framework under which we work, including a lot of the definitions we’ll need. In Chapters 2
and 3 we’ll build up our theory using an example of a discrete-time market; then in Chapters 4 and 5 we’ll look at
extensions to this theory.
1.2.1 Underlying and derivative assets

The assets for sale in financial markets fall into two broad categories: underlying assets, and derivative assets.
Underlying assets have intrinsic value, such as currency, stocks, bonds, and commodities. In this course, we are
particularly interested in two types of underlying asset: bonds and shares.
A bond is a risk-free asset with a predictable price and future value. For example, it can take the form of a loan
between an investor and a borrower, with a fixed rate of return. We’ll discuss interest in more detail further on in
this chapter, but you can think of the bond price 𝐵𝑡 in terms of compound interest, so that 𝐵𝑡 = 𝐵0 (1 + 𝑟)𝑡 .
A share (or a single unit of a stock) is a risky asset, which future price and value are unpredictable. We will write
𝑆𝑡 for the price of a share at time 𝑡. Throughout this course, we’ll model the way the share price evolves using
different probabilistic models, in discrete or continuous time.
For some more information…

…about how bonds and shares work in the “real world”, check out the BBC radio “understanding the economy”
podcast: link to the whole show and to the episode on bonds and shares.
Example 1.1
A cafe in Durham wants to order enough coffee beans to keep its cafes running for Epiphany term (hopefully,
we already have enough coffee to keep us running for Michaelmas; otherwise, we’re in trouble).
It orders 10,000kg of Arabica coffee beans, at about 30 Brazilian Real per kg, for a total of 300,000 BRL.
The exchange rate between BRL and GBP is currently 1:6.02, so three hundred thousand Real is worth about
fifty thousand pounds, which the cafe will have to pay on delivery of the coffee, in January.
The issue here is the uncertainty associated with this transaction. Since we can’t know what the exchange
rate will be in January, it is impossible to know today what the price in GBP will be. If the exchange rate
is still close to 1:6, then the cafe will have to pay £50k, but if the rate decreases to, say, 1:5, then the coffee
will cost more like £60k. This introduces a currency risk which the cafe would presumably rather avoid.
They can avoid this currency risk in several different ways, including:
1. They could buy 300,000 BRL £1,000,000 today, at a price of 50,000 GBP, and keep this money in a
bank account until January.
• Pros: The currency risk is completely eliminated.
• Cons: This is essentially paying for the coffee months before it’s delivered. It’s a lot of money to
tie up, and the cafe might not have it on hand right now.
2. The cafe could “buy” a forward contract for 300,000 BRL with delivery three months from now. This
is an agreement with (e.g.) a bank that the cafe will buy the Real from them, at an exchange rate which
is agreed at time 𝑡 = 0. The rate is called the forward price, and usually denoted 𝐾; here we might have
𝐾 = 5.5. Typically forward contracts involve no upfront cost.
• Pros: There is nothing to pay now, and the currency risk is completely eliminated; the cafe know
exactly how much they will have to pay. If the exchange rate at 𝑡 = 𝑇 is higher than 𝐾, then they
will have saved some money in buying at the lower rate of 1 GBP : 𝐾 BRL - but…
• Cons: …if the exchange rate at 𝑡 = 𝑇 is lower than 𝐾, then the cafe will still have committed to
pay the higher rate, and will lose out compared to paying the market rate for their 300,000 BRL.
3
3. What the cafe really want is a contract allowing them the option to buy at an agreed price, but not the
obligation. This is called a European call option: they agree on a strike price of 𝐾 BRL : 1 GBP
and an expiry time 𝑇 with the broker, and pay a (hopefully fair!) upfront cost for the agreement. If the
exchange rate at time 𝑇 is higher than 𝐾, then they can exercise the option and pay the lower price; if
the exchange rate at time 𝑇 is lower than 𝐾, they can ignore the option and simply buy 300,000BRL
at the market price.
• Pros: the currency risk is completely eliminated; the cafe will either pay 𝐾 BRL : 1 GPB or the
exchange rate at time 𝑇 , whichever is lower.
• Cons: there is an upfront cost for the contract; how much should the cafe be willing to pay?
Both the forward contract and the European call option in this example are derivative assets. They do not have
intrinsic value like the coffee beans or the currency; instead, their value derives from that of the underlying assets.
We might also call these assets contingent claims, as their value is contingent on that of something else.
Derivative assets come in two types: locks and options. Lock products, such as the forward contract, are an
agreement that the holder will buy (or sell) something from the writer at an agreed future date. Options, on the
other hand, are an agreement that the holder can buy or sell something at the future date – but is not obliged to
do so.
There are two main categories of options: call and put. Call options give the holder the right, but not the obligation,
to buy the underlying asset at an agreed price; put options give the holder the right to sell it. The prefix European
means that the option can only be exercised at exactly time 𝑇 ; American options can be exercised at any time
𝑡 ∈ [0, 𝑇 ].
1.2.2 Payoffs
The payoff of an asset is the amount of money it is worth to us at time 𝑇 . We usually denote the payoff as Φ or,
sometimes, Φ𝑇 . For example, if we buy £𝑃 worth of bonds with interest rate 𝑟 at time 0, the payoff at time 𝑇 will
be Φ𝑇 = 𝑃 (1 + 𝑟)𝑇 . Similarly, the payoff of one stock at time 𝑇 is Φ𝑇 = 𝑆𝑇 : its value is the same as its price.
In entering a forward contract, we agree that at time 𝑇 we will buy an asset at a forward price 𝐾. The payoff for
this contract is Φ𝑇 = 𝑆𝑇 − 𝐾: we are receiving assets worth 𝑆𝑇 (the share) and giving away assets worth 𝐾 (in
cash). Depending on the value of 𝑆𝑇 , this contract could have a negative payoff: if the share price is lower than 𝐾,
buying one share for £𝐾 equates to losing money.
On the other hand, writing a forward contract has payoff of Φ𝑇 = 𝐾 − 𝑆𝑇 : we sell one share worth 𝑆𝑇 and receive
𝐾 in cash.
To calculate the payoff of a European call option, we need to consider two cases. If the asset price is higher than
the strike price, 𝑆𝑇 > 𝐾, then the holder can exercise the contract, buying the asset and handing over 𝐾 in cash,
for a payoff of 𝑆𝑇 − 𝐾. On the other hand, if the asset price is lower than the strike price, 𝑆𝑇 < 𝐾, then there is
no reason to exercise the option and lose money; instead, the holder will do nothing.
As a result, the payoff of the option at time 𝑇 is
𝑆 −𝐾 if 𝑆𝑇 > 𝐾
{ 𝑇
0 if 𝑆𝑇 ≤ 𝐾.
We can also view this as max(𝑆𝑇 − 𝐾, 0), which we write as (𝑆𝑇 − 𝐾)+ (as it’s the positive part of this value).
By the same argument, the return at time T of a put option is max(𝐾 − 𝑆𝑇 , 0) = (𝐾 − 𝑆𝑇 )+ .
Remark. Notice that selling call options is not the same thing as buying put options! Mathematically, this is because
(𝐾 − 𝑆𝑇 )+ ≠ −(𝑆𝑇 − 𝐾)+
(one side is always positive or zero, and the other is always negative or zero). In terms of the transactions, the
difference arises from the fact that the choice always lies with the buyer of the option; the seller is obligated to go
along with the buyer’s decision).
4
Figure 1: Payoff graphs for European call (left) and put (right) options with strike price 𝐾.
Exercise 1.1
Draw the graphs of the payoffs of the following combinations of options, as a function of the asset price:
• buying one call plus one put option with the same strike price 𝐾 (this is known as a straddle)
• buying one call option and selling one put option with the same strike price 𝐾
• buying one call option with strike price 𝐾1 , and selling another call option with strike price 𝐾2 (this is
known as a bull spread; you’ll want 𝐾2 > 𝐾1 )
1.2.3 Risk
Investors in financial markets are hugely concerned with risk. Given two investments with the same expected return,
investors will generally prefer the one with smaller variation. For instance, imagine you have £10,000 to invest, and
you are choosing between the following strategies:
• Strategy A will return £0 with probability 0.45, or £20,000 with probability 0.55;
• Strategy B will return £10,000 with probability 0.5, or £12,000 with probability 0.5.
In both cases, the expected return is £11,000; but the variability (and hence the risk) is much higher with Strategy
A. Nearly all investors would prefer Strategy B, as the potential losses are much smaller.
On the other hand, we could consider a third strategy: placing all of our money in a risk-free account, where it will
sit and quietly earn interest.
• Strategy C will return £11,000 with probability 1.
Would you prefer Strategy C over Strategy B? For many investors, the low-risk (but not risk-free) option is most
attractive.
In every market we study, we assume that there are opportunities for risk-free investment in the form of bonds.
These bonds pay interest at a known, constant rate 𝑟.
Context: nominal and effective interest
If the interest is compounded once per year, then after one year an initial deposit of 𝐵(0) will be worth 𝐵(1) =
𝐵(0)(1 + 𝑟), and after 𝑡 years, 𝐵(𝑡) = 𝐵(0)(1 + 𝑟)𝑡 .
If it is compounded 𝑛 times per year (common choices for 𝑛 include 2, 4, 12, 52, and 365) at a nominal rate
𝑟, then at each of 𝑛 equally-spaced intervals, the value of the investment increases by a factor (1 + 𝑟/𝑛), so that
𝐵(1/𝑛) = 𝐵(0)(1 + 𝑟/𝑛) and, more generally, 𝐵(𝑚/𝑛) = 𝐵(0)(1 + 𝑟/𝑛)𝑚 .
If we deposit £𝐵 in bonds at the start of the year, and interest is compounded 𝑛 times at nominal rate 𝑟, then
after one year our investment will be worth £𝐵(1 + 𝑟/𝑛)𝑛 . The effective interest rate is the value 𝑟∗ such that
1 + 𝑟∗ = (1 + 𝑟/𝑛)𝑛 .
If you take out a loan or a mortgage, the APR you’ll see advertised is the effective interest rate, expressed as a
percentage.
5
Present-value analysis
Because we can always invest money into bonds for a guaranteed interest rate, a deposit of £1000 today is “worth
more” than a deposit of £1000 in a year’s time. To account for this, we calculate the present value of a cash flow.
To do this, we multiply each deposit by the discount factor
1
𝛼= ∶
1+𝑟
a deposit of £𝑥 at time 𝑡 is worth 𝑥𝛼𝑡 “in today’s money”.
Example 1.2
Which of these cash flows has the highest (and the lowest) present value?
1. 𝑥𝑖 = 100, 𝑡𝑖 ∈ {1, 3, 5, 7, 9}
2. 𝑥𝑖 = 50, 𝑡𝑖 ∈ {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
3. 𝑥𝑖 = 200, 𝑡𝑖 ∈ {1, 3, 5, 7, 9} and 𝑥𝑖 = −100, 𝑡𝑖 ∈ {2, 4, 6, 8, 10}
4. Wildcard: construct your own cash flow.
Example 1.3
You have a bank account in which the annual interest rate is 5%, compounded monthly You plan to pay
in £𝐷 every month for thirty years (360 months), and then withdraw £1000 every month for the following
twenty years (months 361 - 600). What is the minimum deposit 𝐷 you should be making, to ensure you have
enough in your account?
Here 𝑟 = 0.05/12 = 1/240, so 𝛼 = 240/241. The present value of all the deposits is
359
1 − 𝛼360
𝐷 + 𝐷𝛼 + 𝐷𝛼2 + ⋯ + 𝐷𝛼359 = 𝐷 ∑ 𝛼𝑖 = 𝐷 .
𝑖=0
1−𝛼
Next, the present value of all the withdrawals is
1 − 𝛼240
1000𝛼360 (1 + 𝛼 + 𝛼2 + ⋯ + 𝛼239 ) = 1000𝛼360 .
1−𝛼
These are equal when
1 − 𝛼240 1 − 𝛼 360 1 − 𝛼
240
𝐷 = 1000𝛼360 = 1000𝛼 ≈ 182.065.
1 − 𝛼 1 − 𝛼360 1 − 𝛼360
1.2.4 Portfolios
In general, a portfolio is a description of the shares, options, and cash we hold at any given time. The values of
any of these things can be negative if we engage in short selling: borrowing units of an asset at one time, to be
returned later. In this course, it is always possible to short sell anything we need to: there is always someone willing
to lend it to us. This is part of an assumption of liquidity which we will discuss later.
We can describe a portfolio in one of two different ways: in terms of the amount of each asset we hold at every time
𝑡, or in terms of the amount of each asset we buy or sell.
The trades defining a portfolio need not happen at deterministic (pre-determined) times. For example, we may
wish to create a portfolio in which we buy shares of a stock at time 0, and then sell them as soon as the share
price doubles. In that case, the time of our trade is inf {𝑡 ∶ 𝑆𝑡 ≥ 2𝑆0 }. This is a permissible trade, because we will
recognise the moment when it happens; on the other hand, we can’t decide to sell “whenever the price is highest
between time 0 and time 𝑇 ”, because we can’t know whether or not that moment has arrived. If you took Markov
Chains last year, you might recognise that we’re talking about stopping times.
6
We say that a portfolio is self-financing if each of the trades involved produces no cash flow – that is, if the total
value of what we buy is equal to the total value of what we sell. In practice, this means that we use the principle
that, aside from some initial investment, the trading strategy should take care of itself.
When it is possible to assemble a portfolio that is certain to produce a profit, we say that there is an arbitrage
opportunity in the market. We can think of arbitrage as arising from two different possible sources. Firstly, if
an underlying asset is guaranteed to out-perform the other assets in the market (including the risk-free investment
in bonds), then there is arbitrage inherent in the market. In some sense, the whole market is rigged; there’s a
guaranteed win somewhere in there. Secondly, if an asset (usually a derivative asset) is mis-priced, then investors
can create arbitrage by buying (or selling) the mispriced asset in large quantities, as we’ll see in the next example.
Example 1.4
A company’s shares are floated on the stock market. The value of one share is £100 today (at 𝑡 = 0), and
we know that tomorrow (at 𝑡 = 1) the value will either increase to £200, with some unknown probability
0 < 𝑝 < 1, or decrease to £50 with probability 1 − 𝑝.
We are a broker trading in stocks and options, and we’re considering the following (European call) option:
the holder has the right (but not the obligation) to buy one share for £150, at time 𝑡 = 1. At time 𝑡 = 0,
what is the fair price 𝐶 for this option? We have to be willing to both buy and sell options at this price, and
we will assume that there’s a wealthy trader “Agent A” who will take advantage and run us out of business
if we misprice the option and create an arbitrage.
One approach is to think about the value of the option at time 𝑡 = 1: if the share price is £50, then the option
is worthless and we throw it away. If the share price is £200, then by exercising the option and immediately
selling the share, the holder will make a profit of £50. So the expected value of the option is 50𝑝 – to decide
what 𝐶 should be, we’ll need an estimate for 𝑝. Let’s see what happens if we estimate that 𝑝 = 0.2, so that
𝐶 = 10.
Agent A comes along, and asks to sell us a share and buy 3 options at time 𝑡 = 0. With this transaction, we
initially hand over
100 − 3 × 10 = 70.
An important note here is that Agent A can sell us as many shares as he likes, through short selling. He
borrows the shares from some third party at time 0, and he will have to return them at time 1. In practice,
this means that for every share he sells us now, he will have to buy one back later.
At time 1, if the share price has gone down, the option is worthless. In this case, he can buy one share of the
stock for £50, return it to the third party, and leave with
70 − 50 = 20
in profit.
On the other hand, if the share price has gone up, each option represents a profit of £50. In this case, Agent
A will spend £450 to exercise the options; return one share to the third party; and sell the two remaining
shares for £400. This time, he leaves with
70 − 450 + 400 = 20
in profit.
In either case, Agent A has returned his borrowed share and made a £20 profit from us: there is no risk! This
is an example of an arbitrage opportunity. Since Agent A has effectively limitless money, he can instead sell
a million shares, buy three million options, and make £20 million profit - or even more - and ruin us.
In order to calculate the fair price for the option, we should think about Agent A’s profit in general terms.
Let’s consider a portfolio consisting of 𝑥 units of the stock and 𝑦 units of the option; here, negative values for
𝑥 and 𝑦 represent short selling. The initial cost to Agent A is
100𝑥 + 𝐶𝑦.
7
At time 1, this portfolio is either worth 50𝑥 (if the share price goes down), or 200𝑥 + 50𝑦 (if the share price
goes up). To eliminate the risk and make the two values equal to each other, we set 𝑦 = −3𝑥.
Overall, this portfolio has a risk-free profit of
50𝑥 − (100𝑥 − 3𝐶𝑥) = (3𝐶 − 50)𝑥.
If this is non-zero, Agent A can ruin us! If it’s positive, he takes a hugely positive value for 𝑥 (this is buying the
stock and short selling the option). If it’s negative, he can take a hugely negative value for 𝑥 (this corresponds
to short selling the stock and buying the option, as in the first part of the example). The only way to avoid
arbitrage is to set 𝐶 = 50/3.
A basic underlying principle of derivative pricing is:
No market should inherently contain arbitrage opportunities, and no financial derivative should be priced in
a way that creates an arbitrage opportunity.
In a liquid market, arbitrage opportunities are not stable and never exist for long: investors taking advantage of
the opportunity (or arbitrageurs) will influence the supply and demand within the market and push the price back
towards one that does not create arbitrage.
The principle that no derivative asset should be priced in a way that leads to arbitrage opportunities allows us to
conclude that larger payoffs must mean larger prices.
Proposition 1.1. Consider two self-financing portfolios, Portfolio 1 and Portfolio 2, with costs 𝐶1 and 𝐶2 , and
time-𝑇 payoffs 𝑉1 and 𝑉2 . If 𝑉1 ≥ 𝑉2 is true whatever happens to the asset prices, then we must have 𝐶1 ≥ 𝐶2 or
an arbitrage opportunity exists.
(In other words: if Portfolio 1 is a better bet than Portfolio 2, it can’t cost less.)
Proof. We use the golden principle of exploiting arbitrage opportunities: buy cheap, and (short) sell expensive.
Suppose 𝐶2 > 𝐶1 . At time 0, by borrowing the second portfolio, selling it, and using the money to buy the first
portfolio, our position is:
• holding Portfolio 1
• short Portfolio 2
• spare cash worth 𝐶2 − 𝐶1 , which we can use to buy bonds.
Now at time 𝑇 , we sell Portfolio 1 for 𝑉1 , pay 𝑉2 to buy Portfolio 2 and return it to the lender, and have 𝑉1 − 𝑉2 ≥ 0
left over from the transaction. In combination with the bonds which have been quietly earning interest, we are
guaranteed to make a profit: this is an arbitrage opportunity.
By applying 1.1 in both directions, we obtain the Law of One Price:

Theorem 1.1 (The Law of One Price (LOOP)). If two self-financing portfolios have identical time-𝑇 payoffs,
whatever happens to the asset prices, then at every time 𝑡 ≤ 𝑇 , they must have the same price or an arbitrage
opportunity will exist.
An example of the Law of One Price is Put-Call Parity:
Theorem 1.2 (Put-Call Parity). Consider a European call option with cost 𝐶, and a European put option with
cost 𝑃 , on the same stock. If the initial share price is 𝑆0 , interest is compounded discretely at rate 𝑟, and both 𝐶
and 𝑃 have strike price 𝐾 and expiry date 𝑇 , then we must have
𝑃 + 𝑆0 = 𝐶 + 𝐾(1 + 𝑟)−𝑇 ,
or else an arbitrage opportunity exists.
8
Proof. Consider two portfolios, 𝑋 and 𝑌 . In 𝑋, we buy one call option and put 𝐾(1 + 𝑟)−𝑇 into bonds, buying
𝐾(1 + 𝑟)−𝑇 units, at time 0. The price at time 0 is 𝐶 + 𝐾(1 + 𝑟)−𝑇 , and at time 𝑇 the portfolio will be worth
𝑉𝑇𝑋 = (𝑆𝑇 − 𝐾)+ + 𝐾(1 + 𝑟)−𝑇 × 𝐵𝑇 = (𝑆𝑇 − 𝐾)+ + 𝐾 = max(𝑆𝑇 , 𝐾).
In 𝑌 , we buy one put option and one share of stock at time 0. The price at time 0 is 𝑃 + 𝑆0 , and at time 𝑇 portfolio
𝑌 will be worth
𝑉𝑇𝑌 = (𝐾 − 𝑆𝑇 )+ + 𝑆𝑇 = max(𝐾, 𝑆𝑇 ).
Since both portfolios have the same payoff, they must have the same initial cost by the Law of One Price to avoid
an arbitrage opportunity; so
𝐶 + 𝐾(1 + 𝑟)−𝑇 = 𝑃 + 𝑆0 .
We say that one portfolio hedges or replicates another if they always have the same payoff. A complete market
is a market in which every contingent claim can be hedged with a portfolio consisting only of bonds and shares.
Using the Law of One Price, we can calculate the fair price for any derivative asset at any time by first finding a
hedging portfolio using bonds and shares, and then calculating its fair price – which should be a straightforward
calculation.
1.3 Discrete-time model

In this part of the course, we are interested in discrete-time models: those in which trading can take place only at
specific times.
A discrete-time model with one risky asset consists of:
1. A set of trading times 0 < 𝑡1 < 𝑡2 < 𝑡3 < ⋯ ≤ 𝑇 . These do not have to be equally-spaced, but usually will be;
we will almost always use 𝑡𝑗 = 𝑗 so that trades happen at times 0, 1, 2, 3, … , 𝑇 .
2. An outcome space Ω. Each element of Ω represents one possible way in which the share prices might evolve;
this is how we will encode the risk or randomness in the model.
3. A risk-free asset, or bonds account, whose price dynamics are deterministic and given by
𝐵(𝑡) = (1 + 𝑟)𝑡 , 0 ≤ 𝑡 ≤ 𝑇.
4. A risky asset, whose price dynamics are defined for each 𝜔 ∈ Ω by
𝑆 ∶ 𝜔 ↦ {𝑆1 (𝜔), … , 𝑆𝑇 (𝜔)}.
The collection {𝑆1 (𝜔), … , 𝑆𝑇 (𝜔)} represents one possible price path for the share price.
Notes:
• The outcome space Ω should remind you of probability triples (Ω, ℱ, ℙ) from Probability II. We’re going to
avoid defining a probability measure on Ω for now, but ℱ will usually be the Borel 𝜎-algebra ℱ = 2Ω .
• The interest rate 𝑟 for the risk-free asset must be unique, to avoid arbitrage inherent in the market. If there
were two different risk-free assets, with rates 𝑟1 < 𝑟2 , then a portfolio in which we short-sell one bond at the
lower rate, buying one bond at the higher rate, creates a guaranteed payoff of (1 + 𝑟2 )𝑇 − (1 + 𝑟1 )𝑇 > 0 with
no initial investment.
• We view all price paths for the shares as left-continuous functions of time: 𝑆𝑡 = 𝑆𝑡− .
We make the following assumptions about the market:
a. The market is liquid and divisible: we can buy or sell any (real) quantity of bonds and of shares at each
time 𝑡𝑖 .
b. We do not have to pay transaction fees, and transactions happen instantaneously at each trading time.
c. There is no bid-ask spread: the prices at which we buy are the same as the prices at which we sell.
d. Shares produce no dividends.
e. Our actions have no impact on the pricing of any products in the market.
9
The second and third assumptions, together, are known as a frictionless market. They represent two of the major
stumbling blocks that you will face if you set out to identify and exploit arbitrage opportunities in the real world.
In this course, we usually describe a portfolio as a sequence of holdings vectors, rather than in terms of the trades
we need to make to achieve those holdings. For example, if we hold 𝑥 bonds and 𝑦 units of the stock at time 𝑡, we
have ℎ𝑡 = (𝑥, 𝑦). A portfolio in which we short sell one share at time 𝑡 = 1 and buy it back at time 𝑡 = 3 would
therefore look like:
⎧(0, 0) 0≤𝑡<1
{
ℎ𝑡 = ⎨(0, −1) 1 ≤ 𝑡 < 3
{(0, 0) 3 ≤ 𝑡 < 𝑇.
⎩
Note that this portfolio is defined at 𝑡 = 0, but not at 𝑡 = 𝑇 : we assume that, at time 𝑇 , we will exercise any
contracts that are worth exercising, return anything we borrowed, and sell any remaining assets.
We can view a portfolio as a function 𝑡 ↦ ℎ𝑡 . It’s a step function which is constant on the intervals [0, 1), [1, 2),
and so on. In fact, any right-continuous piecewise step function can be seen as a portfolio, as long as the “jumps”
happen only at trading times. We could also view a portfolio as a sequence of trades, rather than a sequence of
holdings: to recreate the trades involved, we look for values of 𝑡 for which ℎ𝑡 − ℎ𝑡− is non-zero.
The value of a portfolio at time 𝑡 is given by the total value of each of the assets held; for a portfolio 𝑃 = (ℎ𝑡 ; 𝑡 ∈
[0, 𝑇 ]) we write
𝑉𝑡 = ℎ𝑡 ⋅ (𝐵𝑡 , 𝑆𝑡 ) = 𝑥𝑡 𝐵𝑡 + 𝑦𝑡 𝑆𝑡 .
The amount of cash required to initialise a portfolio is given by 𝑉0 - note that this can be negative, for example
if we’re short selling. The cash received at time 𝑡 = 𝑇 when we close out the portfolio is 𝑉𝑇 , which can also be
negative - for example to buy back shares that were sold earlier.
Example 1.5
If our portfolio 𝑋 is “sell 100 shares of the stock at time 0”, then we can express 𝑋 as a sequence of holdings
by
𝑋 = (ℎ𝑋 𝑡 , 𝑡 ∈ [0, 𝑇 ]) where ℎ𝑋
𝑡 = (0, −100) ∀𝑡 ∈ [0, 𝑇 ].
The value of 𝑋 is always 𝑉𝑡 = −100𝑆𝑡 .

If our portfolio 𝑌 is “buy 100 shares at time 0, and sell them as soon as the share price doubles”, then the
holdings are given by
100 sup𝑠∈[0,𝑡] (𝑆𝑠 ) < 2𝑆0
ℎ𝑌𝑡 = { .
0 sup𝑠∈[0,𝑡] (𝑆𝑠 ) ≥ 2𝑆0
We can write this as ℎ𝑌𝑡 = 100 1 {sup𝑠∈[0,𝑡] (𝑆𝑠 ) < 2𝑆0 }. Similarly, the value of the portfolio is given by
𝑉𝑡 = 100𝑆𝑡 1 {sup𝑠∈[0,𝑡] (𝑆𝑠 ) < 2𝑆0 }.
Question: what are the trades, holdings, and value of the portfolio 𝑋 + 𝑌 ?
The cash flow associated with a portfolio describes how its value changes as each trade occurs: if the portfolio has
trades at times 𝑡1 , … , 𝑡𝑘 and associated changes in value 𝑐1 , … , 𝑐𝑘 , then the cash flow is {(𝑐𝑖 , 𝑡𝑖 )}.
We say that a portfolio is self-financing if the cash flow associated with it is trivial; in other words, if 𝑉𝑡 = 𝑉𝑡− for
every 𝑡 ∈ (0, 𝑇 ]. Since 𝐵𝑡 and 𝑆𝑡 are left-continuous, we have
𝑉𝑡 − 𝑉𝑡− = 𝑥𝑡 𝐵𝑡 + 𝑦𝑡 𝑆𝑡 − (𝑥𝑡− 𝐵𝑡− + 𝑦𝑡− 𝑆𝑡− )

= (𝑥𝑡 − 𝑥𝑡− )𝐵𝑡 + (𝑦𝑡 − 𝑦𝑡− )𝑆𝑡 .
The self-financing condition means that this value should always be 0.

Note that we can find a self-financing version of any portfolio by buying or selling bonds in the appropriate quantities
to create a trivial cash flow. Although bonds are not the same as cash in practice, they play the same role in our
mathematical model.
10
The present value profit of a self-financing portfolio 𝑃 with value 𝑉𝑡 , 𝑡 ∈ [0, 𝑇 ] is given by 𝛼𝑇 𝑉𝑇 − 𝑉0 . Remember
that 𝛼 is the discount factor per time period, so that 𝛼 = (1 + 𝑟)−1 if interest is compounded discretely at rate 𝑟
per time period, or 𝛼 = 𝑒−𝑟 if interest is compounded continuously.
We say that a portfolio 𝑃 is an arbitrage portfolio if 𝛼𝑇 𝑉𝑇 − 𝑉0 ≥ 0, whatever happens to 𝑆𝑡 , and it is possible
that 𝛼𝑇 𝑉𝑇 − 𝑉0 > 0.
Alternatively, we can define an arbitrage portfolio as one which satisfies: (i) 𝑃 is self-financing; (ii) 𝑉0 = 0; and (iii)
𝑉𝑇 ≥ 0 whatever happens to the share prices and it is possible that 𝑉𝑇 > 0.
Lemma 1.1. These two definitions of an arbitrage portfolio are equivalent, in the sense that a market either allows
both to exist, or neither.
The proof of this Lemma is Exercise 1.10 on the problem sheet.
11
2 The one-period binomial model

In this chapter, we meet our first, and simplest, example of a discrete-time model: the one-period binomial model.
2.1 The model

The four elements of our model are:
1. A set of trading times. In this model, trades can happen at two times: 𝑡 = 0 (today) and 𝑡 = 𝑇 (next year; if
you prefer, you could write 𝑇 = 1).
2. An outcome space, Ω = {0, 1}.
3. A bond, with price dynamics:
𝐵(0) = 1, 𝐵(𝑇 ) = 1 + 𝑟.
4. A share, whose price dynamics are:
𝑠𝑢 if 𝜔 = 1
𝑆(0) = 𝑠, 𝑆(𝑇 )(𝜔) = {
𝑠𝑑 if 𝜔 = 0.
The probability space associated with this model is (Ω, ℱ, ℙ), in which Ω = {0, 1}, ℱ = {∅, {0}, {1}, {0, 1}}, and ℙ
represents a selection from Ω: ℙ(𝜔 = 0) = 𝑝 = 1 − ℙ(𝜔 = 1). We will assume that 𝑝 ∈ (0, 1), and we will always
assume that 𝑢 > 𝑑.
To describe a one-period binomial model, we need to know the constants 𝑟, 𝑠, 𝑢, 𝑑, and (maybe) 𝑝.
2.2 Portfolios and arbitrage

In this market, a portfolio (or a trading strategy) is any vector ℎ = (𝑥, 𝑦) ∈ ℝ2 . We interpret the portfolio in the
following way:
• buy 𝑥 bonds in the risk-free asset, and 𝑦 shares of the stock, at time 0. (If 𝑥 and/or 𝑦 is negative, this
represents our short selling bonds and/or shares at time 0.)
• sell 𝑥 bonds and 𝑦 shares at time 𝑇 . Remember that we always “close out” our position, that is, we are not
holding onto shares for a later date; and if we have short sold any bonds or shares, we must buy them back
at time 𝑇 .
Our divisibility assumption means that any ℎ ∈ ℝ2 is a valid trading strategy.
Value and arbitrage

The value process of a portfolio ℎ is the process 𝑉𝑡ℎ , given by
𝑥(1 + 𝑟) + 𝑦𝑠𝑢 if 𝜔 = 1
𝑉0ℎ = 𝑥 + 𝑦𝑠, 𝑉𝑇ℎ = {
𝑥(1 + 𝑟) + 𝑦𝑠𝑑 if 𝜔 = 0.
If ℎ is an arbitrage portfolio, then we must have 𝑥 + 𝑦𝑠 = 0, and 𝑥(1 + 𝑟) + 𝑦𝑠𝑢 and 𝑥(1 + 𝑟) + 𝑦𝑠𝑑 must be either
both positive or both negative.
Example 2.1
Consider a one-period binomial market with 𝑟 = 0.1, 𝑠 = 10, 𝑢 = 1.2, 𝑑 = 1.1, 𝑝 = 0.2.
We will look at the portfolio ℎ = (−10, 1). We have
𝑉0ℎ = −10 × 1 + 1 × 10 = 0;
and
1 if 𝜔 = 1
𝑉𝑇ℎ = −10 × 1.1 + 1 × 𝑆𝑇 = { .
0 if 𝜔 = 0
We have found an arbitrage portfolio on this market: if we sell 10 bonds (or borrow 10 pounds, if you prefer)
12
and buy one share of the stock at time 0, at worst we regain our money and we have a chance to make a
profit.
Example 2.2
Let’s change the parameters from the previous example, so that 𝑑 = 0.7 instead of 𝑑 = 1.1. Now 𝑆𝑇 is either
1.2 or 0.7, so that the value process of the portfolio ℎ = (−10, 1) becomes
𝑉0ℎ = 0,
1 if 𝜔 = 1
𝑉𝑇ℎ = −10 × 1.1 + 1 × 𝑆𝑇 = { .
−4 if 𝜔 = 0
This is no longer an arbitrage portfolio, as we can now lose money through it.
Theorem 2.1. There is no arbitrage inherent in the one-period binomial model, if and only if we have
𝑑 < 1 + 𝑟 < 𝑢. (2.1)
Proof. First, we suppose that Equation (2.1) is true, and look for an arbitrage portfolio ℎ = (𝑥, 𝑦). Since
𝑉0ℎ = 𝑥 + 𝑠𝑦 = 0,
we must have 𝑥 = −𝑠𝑦.

Now,
𝑠𝑦(𝑢 − (1 + 𝑟)) if 𝜔 = 1
𝑉𝑇ℎ = { .
𝑠𝑦(𝑑 − (1 + 𝑟)) if 𝜔 = 0
By Equation (2.1), 𝑢 − (1 + 𝑟) must be positive and 𝑑 − (1 + 𝑟) must be negative, so we can only have ℙ(𝑉𝑡ℎ ≥ 0) = 1
if 𝑦 = 0. But then, ℙ(𝑉𝑡ℎ > 0) = 0 and so no arbitrage portfolio exists.
Now, we suppose that one of the inequalities in Equation (2.1) does not hold; for instance, we have (1 + 𝑟) ≥ 𝑢.
Then we have 𝑠(1 + 𝑟) ≥ 𝑠𝑢 > 𝑠𝑑 (Remember that 𝑢 > 𝑑 is always true.)
Consider the portfolio ℎ = (𝑠, −1): we sell one share short and invest all the money in the bond. We have 𝑉0ℎ = 0,
and
𝑠(1 + 𝑟 − 𝑢) if 𝜔 = 1
𝑉𝑇ℎ 𝑠(1 + 𝑟) − 𝑠𝑍 = { .
𝑠(1 + 𝑟 − 𝑑) if 𝜔 = 0
So 𝑉𝑇ℎ ≥ 0 when 𝜔 = 1, and 𝑉𝑇ℎ > 0 when 𝜔 = 0; we have ℙ(𝑉𝑇ℎ ≥ 0) = 1 and ℙ(𝑉𝑇ℎ > 0) > 0, so we have found an
arbitrage portfolio.
We can do a similar calculation if 1 + 𝑟 ≤ 𝑑 (try it!) so that
• if Equation (2.1) holds, there can be no arbitrage portfolio
• if Equation (2.1) is broken in any way, an arbitrage portfolio must exist.
2.3 Contingent claims

Remember that a contingent claim is a contract between the buyer and the seller, in which the seller promises
the random payoff Φ to the buyer at time 𝑇 .
Mathematically, any random variable 𝑋 can represent a contingent claim if we can find a contract function Φ
such that 𝑋 = Φ({𝑆𝑡 ∶ 𝑡 ∈ [0, 𝑇 ]}).
13
Example 2.3
Here are some examples of the contract functions for contingent claims we’ve already seen.
• In a European call option, we have 𝑋 = Φcall (𝑆𝑇 ) = (𝑆𝑇 − 𝐾)+
𝑗
• If the contingent claim is “one unit of every asset on the market”, then 𝑋 = Σ𝑚
𝑗=1 𝑆𝑇 .
• A forward contract on 𝑆 is a contract in which the the asset is to be sold at a strike price 𝐾 at expiry
time 𝑇 , and both buyer and seller are obliged to complete the transaction. In this case, the contract
function is Φ𝐹 (𝑥) = 𝑥 − 𝐾, and 𝑋 = Φ(𝑆𝑇 ) = 𝑆𝑇 − 𝐾.
• In general, in European-style claims Φ only depends on the price at time 𝑇 , 𝑆𝑇 , but in general it can
depend on the values of 𝑆𝑡 at any (or all) times 𝑡 ∈ [0, 𝑇 ].
We say that a contingent claim is reachable if there exists a portfolio ℎ consisting only of bonds and shares, such
that
ℙ(𝑉𝑇ℎ = 𝑋) = 1.
If every contingent claim 𝑋 is reachable, we say that the market is complete.
Theorem 2.2. If 𝑢 > 𝑑, the one-period binomial model is complete.
Proof. For any claim 𝑋 with contract function Φ, we need to show that there exists a portfolio ℎ = (𝑥, 𝑦) with
Φ(𝑠𝑢) if 𝜔 = 1
𝑉𝑇ℎ = { .
Φ(𝑠𝑑) if 𝜔 = 0
In other words, we want to find a solution to the system
(1 + 𝑟)𝑥 + 𝑠𝑢𝑦 = Φ(𝑠𝑢)(1 + 𝑟)𝑥 + 𝑠𝑑𝑦 = Φ(𝑠𝑑).
If 𝑢 > 𝑑, this system has a unique solution, namely,
1 𝑢Φ(𝑠𝑑) − 𝑑Φ(𝑠𝑢)
𝑥= (2.2)
1+𝑟 𝑢−𝑑
Φ(𝑠𝑢) − Φ(𝑠𝑑)
𝑦= . (2.3)
𝑠(𝑢 − 𝑑)
Question: what is a fair price for a contingent claim 𝑋?

At 𝑡 = 𝑇 , this is an easy problem to solve: we know how to calculate its value, by finding 𝑋 using Φ. Writing
Π(𝑋, 𝑡) for the price of 𝑋 at time 𝑡, we must have Π(𝑋, 𝑇 ) = 𝑋.
To calculate the fair price at time 0, we use the following pricing principle:
If a claim 𝑋 is reachable with replication portfolio ℎ, then the only reasonable price process for 𝑋 is
Π(𝑋, 𝑡) = 𝑉𝑡ℎ , 𝑡 = 0, 𝑇 .
Theorem 2.3. Suppose that a claim 𝑋 is reachable with replication portfolio ℎ, and that at time 𝑡 = 0 the price of
𝑋 is different from 𝑉0ℎ . Then there is an opportunity for arbitrage.
Proof. (This is an application of the Law of One Price, Theorem 1.1.)

Let Π(𝑋, 0) be the price of the claim, and consider the case Π(𝑋, 0) > 𝑉0ℎ . At time 0, we can short sell the claim,
buy the portfolio ℎ, and be left with Π(𝑋, 0) − 𝑉0ℎ , which we deposit in bonds.
14
Now at time 𝑇 , the portfolio value 𝑉𝑇ℎ will exactly cover the claim we sold short, and we have a risk-free profit
coming from our bonds, which are now worth (Π(𝑋, 0) − 𝑉0ℎ )(1 + 𝑟).
In the case Π(𝑋, 0) < 𝑉0ℎ , we can follow a similar argument, this time by short selling the portfolio and buying the
claim.
Example 2.4
In the market from Example 2.2, let’s find the fair price for a European call option with strike price 𝐾 = 9
and maturity date 𝑇 . The corresponding contingent claim is 𝑋 = (𝑆𝑇 − 9)+ .
A hedging portfolio for 𝑋 is a portfolio ℎ = (𝑥, 𝑦) such that
(1 + 𝑟)𝑥 + 𝑠𝑢𝑦 = (𝑠𝑢 − 9)+ , (1 + 𝑟)𝑥 + 𝑠𝑑𝑦 = (𝑠𝑑 − 9)+ ;
in other words,
1.1𝑥 + 12𝑦 = 3, 1.1𝑥 + 7𝑦 = 0.
3
The solution here is 𝑦 = 5, 𝑥= − 42
11 , so the value of this portfolio at time 𝑡 is
−42 3 24
𝑉0 = 𝑥 + 𝑦𝑠 = + × 10 = .
11 5 11
24
We conclude that the fair price for this call option is 11 .
2.4 The martingale measure

We say that a process (𝑋0 , 𝑋𝑇 ) is a martingale under a measure ℚ if
𝔼ℚ [𝑋𝑇 ] = 𝑋0 .
(We will see a more precise definition of martingales in more general contexts, later in the term.)
In this section, we’ll look at the conditions we need to place on the model to guarantee the existence of such a
measure for the process (𝑆0 , 𝛼𝑆𝑇 ) and determine what that measure must be.
1
Since 𝛼 = 1+𝑟 and 𝑆0 = 𝑠, we are really looking for a measure under which
1
𝑠= 𝔼 [𝑆 ].
1+𝑟 ℚ 𝑇
If we write
ℚ(𝑍 = 𝑢) = 𝑞𝑢
ℚ(𝑍 = 𝑑) = 𝑞𝑑 ,
our martingale condition becomes

1
(𝑠𝑢𝑞𝑢 + 𝑠𝑑𝑞𝑑 ) = 𝑠,
1+𝑟
and our question is really: under which conditions do there exist 𝑞𝑢 and 𝑞𝑑 which solve
𝑢𝑞𝑢 + 𝑑𝑞𝑑 = 1 + 𝑟 (2.4)

𝑞𝑢 + 𝑞𝑑 = 1 (2.5)
0 < 𝑞𝑢 , 𝑞𝑑 < 1? (2.6)
Surprisingly, we have already met such conditions in Equation (2.1).

Theorem 2.4. The financial market ℳ = (𝐵𝑡 , 𝑆𝑡 ) is arbitrage free if and only if there exists a martingale measure
ℚ.
15
Proof. From Theorem 2.1, we know that the market is arbitrage free if and only if
𝑑 < 1 + 𝑟 < 𝑢.
To see that Equations (2.5) have a solution if and only if 𝑑 < 1 + 𝑟 < 𝑢, we could use some convex analysis: this is
exactly the condition under which 1 + 𝑟 can be written as a convex combination of 𝑢 and 𝑑.
Otherwise, we can look for the solutions directly. Solving the system of linear equations, we get
(1 + 𝑟) − 𝑑
𝑞𝑢 = (2.7)
𝑢−𝑑
𝑢 − (1 + 𝑟)
𝑞𝑑 = . (2.8)
𝑢−𝑑
Both 𝑞𝑢 and 𝑞𝑑 are positive if and only if the no-arbitrage condition holds.
Theorem 2.4 is a version of the first fundamental theorem of asset pricing, which we will see in full detail
later; we can also prove a simple version of the second fundamental theorem of asset pricing.
Theorem 2.5. Suppose the financial market is arbitrage free. Then it is complete if and only if there is a unique
martingale measure.
Proof. If the market is arbitrage free, we must have 𝑑 < 1 + 𝑟 < 𝑢, and in particular 𝑑 < 𝑢. By Theorem 2.2, the
market must be complete, and by Theorem 2.4 we know that a martingale measure exists.
To see that it is unique, suppose we have found two martingale measures ℚ1 and ℚ2 , such that
1
𝔼 [𝑆 ] = 𝑆0
1 + 𝑟 ℚ𝑖 𝑇
holds for each of 𝑖 = 1, 2. In other words, we have found 𝑞1 , 𝑞2 which both satisfy
𝑞𝑖 𝑢 + (1 − 𝑞𝑖 )𝑑 = 1 + 𝑟.
The only way 𝑞1 and 𝑞2 can be different is if 𝑢 = 1 + 𝑟 = 𝑑, which contradicts our no-arbitrage assumption.
Example 2.5
Using the market ℳ from Example 2.2, with parameters 𝑟 = 0.1, 𝑠 = 10, 𝑢 = 1.2, and 𝑑 = 0.7, we can check
that 𝑑 < 1 + 𝑟 < 𝑢, so the market is arbitrage free; it is complete, so there exists a unique martingale measure.
We have
1.1 − 0.7 4 1.2 − 1.1 1
𝑞𝑢 = = , 𝑞𝑑 = = ,
1.2 − 0.7 5 1.2 − 0.7 5
and we can check that
1 1 4 1
𝔼ℚ [𝑆𝑇 ] = [ × 12 + × 7] = 10 = 𝑆0 .
1+𝑟 1.1 5 5
2.5 Risk neutral valuation

Now that we know whether binomial market is complete, we can price any contingent claim. As we saw in Theorem
2.2, if 𝑢 > 𝑑 any contingent claim 𝑋 with contract function Φ is reachable, with hedging portfolio ℎ = (𝑥, 𝑦) given
by
1 𝑢Φ(𝑠𝑑) − 𝑑Φ(𝑠𝑢)
𝑥= (2.9)
1+𝑟 𝑢−𝑑
Φ(𝑠𝑢) − Φ(𝑠𝑑)
𝑦= . (2.10)
𝑠(𝑢 − 𝑑)
16
The price at time 𝑡 = 0 of this portfolio is given by
Π(𝑋, 0) = 𝑉0ℎ
= 𝑥 + 𝑠𝑦
1 𝑢Φ(𝑠𝑑) − 𝑑Φ(𝑠𝑢) Φ(𝑠𝑢) − Φ(𝑠𝑑)
= ( + (1 + 𝑟) )
1+𝑟 𝑢−𝑑 𝑢−𝑑
1
= (Φ(𝑠𝑢)𝑞𝑢 + Φ(𝑠𝑑)𝑞𝑑 ) ,
1+𝑟
where 𝑞𝑢 and 𝑞𝑑 are exactly the probabilities coming from the martingale measure ℚ! We can interpret this as an
expectation under ℚ, to get the following pricing formula.
Theorem 2.6. If the one-period binomial model is free of arbitrage, then the arbitrage-free price of a contingent
claim 𝑋 at time 𝑡 = 0 is given by the risk neutral valuation formula
1
Π(𝑋, 0) = 𝔼 [𝑋],
1+𝑟 ℚ
where ℚ is the martingale measure (or risk-neutral measure) uniquely determined by the relation
1
𝔼 [𝑆 ] = 𝑆0
1+𝑟 ℚ 𝑇
or, equivalently, given explicitly in Equations (2.8). Furthermore, the claim can be replicated using the portfolio set
out in Equations (2.10).
Example 2.6
Using our same market from the other examples in this chapter, let’s price the call option with strike price
𝐾 = 9. We have 𝑞𝑢 = 45 and 𝑞𝑑 = 15 , so by the risk-neutral valuation formula,
1
Π(𝑋, 0) = 𝔼 [(𝑆 − 9)+ ]
1+𝑟 ℚ 𝑇
1
= (3 × 𝑞𝑢 + 0 × 𝑞𝑑 )
1.1
1 4 24
= ×3× = .
1.1 5 11
2.6 Some generalisations

We can extend the one-period binomial model slightly, and the first and second fundamental theorems of asset
pricing will still hold. Here are two (non-examinable) examples.
What happens when 𝑢 = 𝑑

If 𝑢 = 𝑑, both assets offer guaranteed rates of payoff per unit of investment: we have
𝐵𝑇 𝑆𝑇
=1+𝑟 and = 𝑢 = 𝑑.
𝐵0 𝑆0
If these rates are not the same, there will be arbitrage in the market, with an arbitrage portfolio formed by buying
the asset with the higher rate of payoff, and selling the asset with the lower rate. Meanwhile, we can slightly modify
the argument from the proof of Theorem 2.1 to show that if 𝑢 = 𝑑 = 1 + 𝑟, there is no arbitrage portfolio on the
market. Hence the no-arbitrage condition becomes 𝑢 = 𝑑 = 1 + 𝑟.
For any portfolio ℎ, 𝑉𝑇ℎ = 𝑥𝐵𝑇 + 𝑦𝑆𝑇 has a fixed value; so the only contingent claims 𝑋 which are reachable are
those which are also constant. Any contingent claim whose final value is genuinely random cannot be reached in
this market: the market is not complete. On the other hand, any claim of the form 𝑋 = Φ(𝑆𝑇 ) is still reachable.
For any probability measure ℚ = (𝑞𝑢 , 𝑞𝑑 ), we have
𝑢𝑞𝑢 + 𝑑𝑞𝑑 = 𝑢 = 𝑑,
17
so martingale measures exist precisely when 𝑢 = 𝑑 = 1 + 𝑟. This means that the first fundamental theorem still
holds: we have no arbitrage, if and only if 𝑢 = 𝑑 = 1 + 𝑟, if and only if there exists a martingale measure.
Finally, if any martingale measure exists, we must have multiple martingale measures. However, as we have
observed, the market is not complete, so the second fundamental theorem also holds in this setting.
Many assets, and/or many outcomes, at time 𝑇

Finally, while keeping a single time period (that is, trades at time 𝑡 = 0 and 𝑡 = 𝑇 only), we can consider markets
with more than 2 assets and/or more than 2 possible states of the market at time 𝑇 .
Suppose the market has 𝑚 assets, whose prices at time 0 are given by the deterministic vector 𝑆0 = (𝑆01 , … , 𝑆0𝑚 ), and
that at time 𝑇 the price vector 𝑆𝑇 takes one of 𝑛 possible values. In other words, there are 𝑛 vectors a1 , … , a𝑛 ∈ ℝ𝑚
and 𝑛 positive probabilities 𝑝1 , … , 𝑝𝑛 such that
ℙ(𝑆𝑇 = a𝑗 ) = 𝑝𝑗 , 𝑗 = 1, … , 𝑛.
These vectors also define an 𝑚 × 𝑛 matrix 𝐴, in which the 𝑖th row lists the possible values for the price of the 𝑖th
asset at time 𝑇 , and the 𝑗th column is the vector a𝑗 .
Arbitrage on this market is defined in the same way as on the simple market: any portfolio ℎ = (𝑥1 , … , 𝑥𝑚 ) ∈ ℝ𝑚
is an arbitrage portfolio if it satisfies (i) 𝑉0ℎ = 0; (ii) ℙ(𝑉𝑇ℎ ≥ 0) = 1; and (iii) ℙ(𝑉𝑇ℎ > 0) > 0. The value of ℎ at
time 𝑇 , 𝑉𝑇ℎ , is random, and takes one of the 𝑛 values in the row vector ℎ𝐴. So finding an arbitrage portfolio is
equivalent to finding a vector ℎ ∈ ℝ𝑚 such that ℎ ⋅ 𝑆0 = 0, and ℎ𝐴 has all non-negative entries, with at least one
strictly positive.
To define a martingale measure on this market, we look for a measure ℚ = (𝑞1 , … , 𝑞𝑛 ) such that, for each 𝑖 = 1, … , 𝑚,
1
𝑆0𝑖 = 1+𝑟 𝔼ℚ [𝑆𝑇𝑖 ]. In terms of the matrix 𝐴, a martingale measure is equivalent to a vector 𝑞 = (𝑞1 , … , 𝑞𝑛 ) ∈ ℝ𝑛
1
such that 𝑆0 = 1+𝑟 𝑞𝐴𝑇 , and such that each 𝑞𝑗 ∈ (0, 1).
The first and second fundamental theorems of asset pricing both hold here. The first is a consequence of a result
known as Farkas’ Lemma, concerning the solvability of systems of linear equalities: it tell us, roughly speaking, that
given the system of equations defining ℎ, and the system of equations defining 𝑞, exactly one of these has a solution.
For the second, we can relate the completeness of the market to the (row) rank of 𝐴 being equal to 𝑛, while the
1
uniqueness of a solution to 𝑆0 = 1+𝑟 𝑞𝐴𝑇 corresponds to the nullity of 𝐴 being 0. By the rank-nullity theorem, the
first situation holds if and only if the second does too.
18
3 The multi-period binomial model: Part 1

3.1 Introduction
In this chapter, we extend the one-period binomial model to allow for more trading times. Our model becomes:
1. A set of trading times, {0, 1, 2, … , 𝑇 }.
2. An outcome space, Ω = {0, 1}𝑇 .
3. A bond, with price dynamics:
𝐵(0) = 1, 𝐵(𝑡) = (1 + 𝑟)𝑡 .
4. A share, whose price dynamics are:
𝑆0 (𝜔) = 𝑠, for all 𝜔 ∈ Ω,
and for each 𝑡 = 1, 2, … , 𝑇 ,
𝑆𝑡 (𝜔) = 𝑍𝑡 (𝜔)𝑆𝑡−1 (𝜔),
where 𝑍1 , 𝑍2 , … , 𝑍𝑇 are random variables defined on Ω, with
𝑢 if 𝜔𝑡 = 1,
𝑍𝑡 (𝜔) = {
𝑑 if 𝜔𝑡 = 0.
Notes:
• In this model, Ω is the set of all sequences of length 𝑇 , in which each element is a 0 or a 1. The sequence 00…0
corresponds to the share price moving “down” 𝑇 times, while the sequence 11…1 corresponds to the share
price moving “up” 𝑇 times. When we discuss an element 𝜔 ∈ Ω, we will sometimes write 𝜔 = 𝜔1 𝜔2 … 𝜔𝑇 .
• The Borel 𝜎-algebra on Ω is ℱ𝑇 = 2Ω and consists of all possible subsets of the 2𝑇 elements of Ω. For instance
{𝜔 ∈ Ω ∶ 𝜔1 = 0} ∈ ℱ𝑇 . More on this later.
• As before, the price dynamics of the risk-free asset evolve deterministically via the risk-free interest rate.
• We assume that 𝑆0 > 0 is fixed, and that if we know the share price is 𝑆𝑡 at time 𝑡, there are two possible
values for the share price at time 𝑡 + 1: 𝑆𝑡 × 𝑢 and 𝑆𝑡 × 𝑑. Although we will sometimes work with models in
which 𝑢 and 𝑑 take the same values for all 𝑡 and 𝑠, this is not necessarily the case, as you will see in Example
3.1.
Example 3.1
Consider a two-period model with the possible share prices 𝑆𝑡 , 𝑡 = 0, 1, 2, given by the tree:
A tree diagram, with prices 120 and 60 at time 1, and 180, 80, 72, and 36 at time 2.
180
120
80
St : 80
72
60
36
t=0 t=1 t=2
• At 𝑡 = 0, we have
120 3 60 3
𝑢= = and 𝑑= = .
80 2 80 4
19
• At 𝑡 = 1, when 𝑆1 = 120, we have
180 3 80 2
𝑢= = and 𝑑= = .
120 2 120 3
• At 𝑡 = 1, when 𝑆1 = 60, we have
72 6 36 3
𝑢= = and 𝑑= = .
60 5 60 5
As in the one-period version of the model, the values of 𝑝𝑢 and 𝑝𝑑 do not matter to us, as long as 0 < 𝑝𝑢 , 𝑝𝑑 < 1.
For this reason, we do not ask whether 𝑝𝑢 always takes the same value regardless of 𝑡 and 𝑠; both possibilities are
entirely reasonable.
The multi-period binomial model is free from inherent arbitrage if and only if there is no inherent arbitrage
to be found by considering any of its constituent parts. In other words, it is free from inherent arbitrage as
long as
𝑑 <1+𝑟 <𝑢
holds everywhere in the model.
A portfolio on a multi-period market is now more complicated to describe, since we need to specify the amounts
we hold of each asset at each time 𝑡, and this is allowed to depend on how the share price changes over time.
However, we can still price contingent claims in this model, by using the risk-neutral valuation formula to find the
no-arbitrage price at each node of the tree.
3.2 Portfolios and the self-financing condition

In this market, a portfolio is a sequence of pairs of random variables ℎ𝑡 = (𝑥𝑡 , 𝑦𝑡 ), 𝑡 = 1, … , 𝑇 . Each pair is
interpreted as follows:
• 𝑥𝑡 is the number of units of the risk-free asset in your portfolio at time 𝑡 − 1 and kept until time 𝑡.
• 𝑦𝑡 is the number of units of the risky asset in your portfolio at time 𝑡 − 1 and kept until time 𝑡.
The amounts 𝑥𝑡 and 𝑦𝑡 can depend on the present and past history 𝑆0 , 𝑆1 , … , 𝑆𝑡−1 of the stock prices; it is reasonable
to assume that we can find functions 𝑓𝑥 and 𝑓𝑦 such that
𝑥𝑡 = 𝑓𝑥 (𝑆0 , … , 𝑆𝑡−1 ) 𝑦𝑡 = 𝑓𝑦 (𝑆0 , … , 𝑆𝑡−1 ).
Example 3.2
Consider a market with 𝑇 = 2, in which the time-0 share price is 10, at time 1 the share price will either be
15 or 8, and at time 2 the share price will be 18, 12, or 6. For this example, we use 𝑟 = 0.
Our portfolio is “buy 20 shares at time 0; at time 1, sell half (and buy bonds) if the price goes up, or sell all
of them if the price goes down”.
In either case, we short sell 20 × 10 = 200 bonds to buy the initial shares; at time 1, we use the cash from
our sale to buy either 10 × 15 = 150 or 20 × 8 = 160 bonds.
To express this portfolio as a sequence of vectors, we write:
at time 0, (𝑥1 , 𝑦1 ) = (−200, 20)

(−50, 10) if 𝜔1 = 1
at time 1, (𝑥2 , 𝑦2 ) = {
(−40, 0) if 𝜔1 = 0.
The value of a portfolio at time 𝑡 = 0, … , 𝑇 is given by

𝑉𝑡 = 𝑥𝑡+1 𝐵𝑡 + 𝑦𝑡+1 𝑆𝑡 ,
with the convention that 𝑥𝑇 +1 = 𝑥𝑇 and 𝑦𝑇 +1 = 𝑦𝑇 . In other words, this is the value of the portfolio after the
trade at time 𝑡, and before the next price change.
20
Example 3.3
In the market in the previous example, if 𝜔1 = 1 (so the share price increases to 15 at time 1), then the value
of the portfolio at time 1 is
𝑉1ℎ = 𝑥2 𝐵1 + 𝑦2 𝑆1 = −50 + 10 × 15 = 100.
Between times 𝑡 − 1 and 𝑡, the values of the bond and the stock change from 𝐵𝑡−1 and 𝑆𝑡−1 to 𝐵𝑡 and 𝑆𝑡 , so that
the portfolio is now worth 𝑥𝑡 𝐵𝑡 + 𝑦𝑡 𝑆𝑡 ; we adjust our holdings according to the price changes, so that we now hold
(𝑥𝑡+1 , 𝑦𝑡+1 ). This adjustment is self-financing if
𝑥𝑡 𝐵𝑡 + 𝑦𝑡 𝑆𝑡 = 𝑥𝑡+1 𝐵𝑡 + 𝑦𝑡+1 𝑆𝑡 . (3.1)
If (3.1) holds for all 𝑡 = 1, … , 𝑇 , we say the portfolio is self-financing. This condition basically means that you
cannot take out money from the portfolio or put in money to the portfolio: it should finance itself.
Remark. Using the expression for the value 𝑉𝑡 of the portfolio at time 𝑡, the self-financing constraint (3.1) can be
rewritten as the condition
𝑉𝑡 − 𝑉𝑡−1 = 𝑥𝑡 (𝐵𝑡 − 𝐵𝑡−1 ) + 𝑦𝑡 (𝑆𝑡 − 𝑆𝑡−1 ). (3.2)
The financial meaning of this is that the change in the value 𝑉𝑡 − 𝑉𝑡−1 of a self-financing trading strategy is due
only to the gains/losses
𝑥𝑡 (𝐵𝑡 − 𝐵𝑡−1 ) + 𝑦𝑡 (𝑆𝑡 − 𝑆𝑡−1 )
due to price changes of the assets. Summing both sides of (3.2) over 𝑡 we obtain
𝑇 𝑇 𝑇
𝑉𝑇 − 𝑉0 = ∑(𝑉𝑡 − 𝑉𝑡−1 ) = ∑ 𝑥𝑡 [𝐵𝑡 − 𝐵𝑡−1 ] + ∑ 𝑦𝑡 [𝑆𝑡 − 𝑆𝑡−1 ].
𝑡=1 𝑡=1 𝑡=1
You will see this form when we start to define stochastic integration and self-financing conditions in continuous
time.
3.3 The martingale measure

To construct the martingale measure in this model, we take the product measure based on the martingale measure
found in Chapter 2. To find ℚ(𝜔), we trace out the path represented by 𝜔 in the pricing tree. We calculate the
one-step martingale measures at each node, and multiply them together.
Example 3.4
In the market from Example 3.1, the choice 𝜔 = 10 represents the share price increasing in the first instance,
and decreasing in the second:
At 𝑡 = 0, we have
1 − 34 1 3
2 −1 2
𝑞𝑢 = 3 3 = and 𝑞𝑑 = 3 3 = .
2 − 4
3 2 −4 3
21
At 𝑡 = 1 and 𝑆1 = 120, we have
1 − 23 2 3
2 −1 3
𝑞𝑢 = 3 2 = and 𝑞𝑑 = 3 = .
2 − 3
5 2 − 32 5
So
1 3 1
ℚ(𝜔 = 10) = × = .
3 5 5
Next, for 𝐴 ∈ ℱ,we define ℚ(𝐴) using the elements 𝜔 ∈ 𝐴:
ℚ(𝐴) = ∑ ℚ(𝜔).
𝜔∈𝐴
Exercise 3.1
“Prove” (convince yourself) that constructing the product measure in this way makes sense: all the probabilities
are non-negative and add up to 1, and we have ℚ (𝜔1 = 1) = 𝑞𝑢 and ℚ (𝜔1 = 0) = 𝑞𝑑 as expected.
Suggestion: in the model with 𝑇 = 2, calculate ℚ(11) and ℚ(10) and check that they add up to 𝑞𝑢 . Do the
same calculation for 𝑇 = 3, and then think about the general case.
3.4 Pricing European contingent claims

3.4.1 Pricing contingent claims via the risk-neutral valuation formula
We work from 𝑡 = 𝑇 − 1 to 𝑡 = 0, using backwards induction.
We denote the price of a contingent claim 𝑋 = Φ(𝑆𝑇 ), evaluated at time 𝑡, by Π(𝑋, 𝑡). We will also sometimes use
the notation Π𝑡 (Φ).
Let’s focus on a single node of the tree. Suppose that the share price at time 𝑡 is 𝑆𝑡 = 𝑠, and the two possible values
for 𝑆𝑡+1 are 𝑠𝑢 and 𝑠𝑑. We also suppose that we know the two possible values for the fair price for the contingent
claim at time 𝑡 + 1. We denote these Π𝑢𝑡+1 and Π𝑑𝑡+1 .
su Πut+1 price of claim when share price s → su
sd Πdt+1 price of claim when share price s → sd
t t+1
Using the same approach as in Chapter 2, we can find the martingale probabilities:
1+𝑟−𝑑 𝑢 − (1 + 𝑟)
𝑞𝑢 = 𝑞𝑑 = ,
𝑢−𝑑 𝑢−𝑑
and then the price Π𝑡 at time 𝑡 is given by
1
Π𝑡 = (𝑞 Π𝑢 + 𝑞𝑑 Π𝑑𝑡+1 ) .
1 + 𝑟 𝑢 𝑡+1
Working through all the nodes in this way, we can eventually find Π0 as a linear combination of Π𝑢1 and Π𝑑1 .
Example 3.5
22
For the market in Example ??, with 𝑟 = 0, let’s calculate the fair price of a European call option with expiry
date 𝑇 = 2 and strike price 𝐾 = 70.
First, we find the martingale measure:
• at 𝑡 = 0, we have
1 − 34 1 3
2 −1 2
𝑞𝑢 = 3 3 = and 𝑞𝑑 = 3 3 = .
2 − 4
3 2 −4 3
• at 𝑡 = 1 and 𝑆1 = 120, we have
1 − 23 2 3
2 −1 3
𝑞𝑢 = 3 2 = and 𝑞𝑑 = 3 = .
2 − 3
5 2 − 32 5
• at 𝑡 = 1 and 𝑆1 = 60, we have
1 − 35 2 6
5 −1 1
𝑞𝑢 = 6 3 = and 𝑞𝑑 = 6 = .
5 − 5
3 5 − 53 3
Now, we work backwards to find the price of the option at each node. When 𝑡 = 2, we know that Π2 =
(𝑆2 − 70)+ , so that the option is worth 110, 10, 2, or 0, when 𝑆2 equals 180, 80, 72, and 36, respectively.
Now when 𝑡 = 1 and 𝑆1 = 120, we have
1
Π1 = (𝑞 Π𝑢 + 𝑞𝑑 Π𝑑2 )
1+𝑟 𝑢 2
1 2 3
= ( × 110 + × 10)
1+0 5 5
= 50.
When 𝑡 = 1 and 𝑆1 = 60, we have

1
Π1 = (𝑞 Π𝑢 + 𝑞𝑑 Π𝑑2 )
1+𝑟 𝑢 2
1 2 1
= ( × 2 + × 0)
1+0 3 3
4
= .
3
Finally, we 𝑡 = 0, we have
1
Π0 = (𝑞 Π𝑢 + 𝑞𝑑 Π𝑑1 )
1+𝑟 𝑢 1
1 1 2 4
= ( × 50 + × )
1+0 3 3 3
158
= ;
9
158
so the no-arbitrage price of the call option is 9 .
These calculations are often recorded in a single tree: we write the share prices inside the nodes, the martingal
probabilities along the arrows, and the prices of the claim next to each node in rectangles.
23
2
5 180 110
1
120 50
3
3 80 10
5
158
80 9
2
3 72 2
2
3 4
60 3
1 36 0
3
Once we have calculated the martingale probabilities at each node, the martingale measure for 𝑆𝑇 is the product
measure ℚ, which we obtain by multiplying the probabilities along each arrow. For instance, in Example ??,
ℚ(𝑆𝑇 = 180) = 13 × 25 = 15
2
. Under this measure, we have
1
𝑆0 = 𝔼 [𝑆 ],
(1 + 𝑟)𝑇 ℚ 𝑇
and we can view the initial price Π0 (Φ) as the present value of the expectation of Φ(𝑆𝑇 ):
1
Π0 (Φ) = 𝔼 [Φ(𝑆𝑇 )].
(1 + 𝑟)𝑇 ℚ
Warning: of course you already know this, but 𝔼ℚ [Φ(𝑆𝑇 )] is not the same thing as Φ(𝔼ℚ [𝑆𝑇 ]).
Example 3.6
Consider the market as described in Examples ?? and ??. Using the martingale probabilities on each arrow,
we can calculate the martingale measure ℚ for 𝑆2 (here 𝑇 = 2). For example, the stock price 𝑆2 equals 180
exactly when the price 𝑆𝑡 follows the path 80 → 120 → 180 through the tree, so ℚ(𝑆2 = 180) = 31 × 25 = 15 2
.
2 1 4
There are a total of 4 possible values that 𝑆2 can take: 180, 80, 72 or 36, and these have probabilities 15 , 5 , 9
and 92 .
The expectation of 𝑆2 with respect to this measure is
2 1 4 2
𝔼ℚ [𝑆2 ] = × 180 + × 80 + × 72 + × 36 = 80.
15 5 9 9
As expected this equals 𝑆0 , since in this example 𝑟 = 0.

Now, to find the price Π0 at time 0 of the call option with strike price 70 and expiry date 𝑇 = 2, we calculate
the present value of the expectation under ℚ of the value at time 𝑇 of the call option, (𝑆2 − 70)+ . We find
that
2 1 4 2 158
𝔼ℚ [(𝑆2 − 70)+ ] = × 110 + × 10 + × 2 + × 0 = ,
15 5 9 9 9
which agrees with our previous calculation.
3.4.2 Pricing contingent claims by hedging

For now, we will focus on European claims – that is, claims whose contract functions are of the form 𝑋 = Φ(𝑆𝑇 ).
In Chapter 5, we will extend our theory to include claims in which the payoff can depend on the past history of the
stock price.
Theorem 3.1. For a contingent claim 𝑋, suppose that there exists a self-financing trading strategy ℎ𝑡 = (𝑥𝑡 , 𝑦𝑡 ),
𝑡 = 1, 2, … , 𝑇 + 1 whose value process 𝑉𝑡ℎ , 𝑡 = 0, … , 𝑇 satisfies
ℙ(𝑉𝑇ℎ = 𝑋) = 1.
24
Then for each 𝑡 = 0, 1, … , 𝑇 , we must have

Π𝑡 (Φ) = 𝑉𝑡ℎ ,
or else the market will contain arbitrage.
Proof. We first suppose that 𝑉𝑡ℎ < Π𝑡 (Φ) holds for some 𝑡. Then, at this time, we can (short) sell the contingent
claim and use the money to buy the portfolio, using the remaining Π𝑡 (Φ) − 𝑉𝑡ℎ to buy bonds.
Now at time 𝑇 , the value of our short-sold contingent claim is exactly covered by our portfolio, and we have a
risk-free profit of [Π𝑡 (Φ) − 𝑉𝑡 ](1 + 𝑟)𝑇 −𝑡 .
Similarly, if 𝑉𝑡ℎ > Π𝑡 (Φ), we can buy the contingent claim and sell the portfolio, and obtain a risk-free profit of
[𝑉𝑡 − Π𝑡 (Φ)](1 + 𝑟)𝑇 −𝑡 .
We can now construct the hedging portfolio for any claim, by backwards induction from 𝑡 = 𝑇 . We know the price
of the claim at all nodes at time 𝑇 ; to move backwards in time by one step, we need to find 𝑥𝑡 and 𝑦𝑡 such that
𝑥𝑡−1 𝐵𝑡−1 + 𝑦𝑡−1 𝑆𝑡−1 = 𝑉𝑡−1 = 𝑥𝑡 𝐵𝑡−1 + 𝑦𝑡 𝑆𝑡−1
holds, whichever node we reach at time 𝑡. The two possiblities give us two equations in two unknowns:
𝑥𝑡 (1 + 𝑟)𝐵𝑡−1 + 𝑦𝑡 𝑆𝑡−1 𝑢 = Π𝑢𝑡 ,

𝑥𝑡 (1 + 𝑟)𝐵𝑡−1 + 𝑦𝑡 𝑆𝑡−1 𝑑 = Π𝑑𝑡 ,
which we can solve to get

1 𝑢Π𝑑𝑡 − 𝑑Π𝑢𝑡 Π𝑢𝑡 − Π𝑑𝑡
𝑥𝑡 = , 𝑦𝑡 = . (3.3)
(1 + 𝑟)𝐵𝑡−1 𝑢−𝑑 𝑆𝑡−1 𝑢 − 𝑆𝑡−1 𝑑
We can check that
1 𝑢Π𝑑𝑡 − 𝑑Π𝑢𝑡 Π𝑢 − Π𝑑𝑡
Π𝑡−1 = 𝑉𝑡−1 = 𝑥𝑡 𝐵𝑡−1 + 𝑦𝑡 𝑆𝑡−1 = + 𝑡
1+𝑟 𝑢−𝑑 𝑢−𝑑 (3.4)
1 𝑢 𝑑
= [𝑞 Π + 𝑞𝑑 Π𝑡 ],
1+𝑟 𝑢 𝑡
1+𝑟−𝑑 𝑢 − (1 + 𝑟)
recalling that the martingale probabilities are 𝑞𝑢 = , 𝑞𝑑 = .
We can repeat these calculations at every node at time 𝑡 − 1 (remembering that the values of 𝑢 and 𝑑 might be
different at different nodes), to find the price 𝑉𝑡−1 = Π𝑡−1 (Φ) and the portfolio (𝑥𝑡 , 𝑦𝑡 ) at every node at time 𝑡 − 1.
So, starting at time 𝑇 , we use Equations (3.3) to find (𝑥𝑇 , 𝑦𝑇 ) at every node at time 𝑇 − 1 so that the value of the
portfolio matches the prices Π𝑇 (Φ) at time 𝑇 . We calculate the value of this portfolio 𝑉𝑇 −1 , which then must be
equal to the price of the claim Π𝑇 −1 (Φ) at time 𝑇 − 1. Then, using our inductive step for 𝑡 = 𝑇 − 1, 𝑇 − 2, … , 1, we
can find the price of the claim and the portfolio at every node of the tree.
Remark. Just as in the one-period case, we can use the martingale measure to calculate the fair price for the claim
without finding the hedging portfolio, through Equation (3.4).
Example 3.7
Let’s calculate the hedging portfolio for the call option from Example ??. We will use the fair prices of the
call option at each time, that we calculated earlier.
We start at time 𝑡 = 1, with the case 𝑆1 = 120. We need to find (𝑥2 , 𝑦2 ) such that the value of the portfolio
matches the fair price of the option (either 110 or 10) at time 2. Since 𝑟 = 0, we have 𝐵𝑡 = 1 for all 𝑡 and
Equations (3.3) become
3
2 × 10 − 23 × 110 110 − 10
𝑥2 = 3 2 = −70, 𝑦2 = = 1,
2 − 3
180 − 80
25
and we have
𝑉1 = 𝑥2 + 120𝑦2 = −70 + 120 × 1 = 50,
which matches the fair price for the claim we found earlier.
If instead, 𝑆1 = 60, then our (𝑥2 , 𝑦2 ) will need to give values of either 2 or 0 at time 2. Now Equations (3.3)
give
6
×0− 3 ×2 2−0 1
𝑥2 = 5 6 53 = −2, 𝑦2 = = ,
5 − 5
72 − 36 18
and
1 4
𝑉1 = 𝑥2 + 60𝑦2 = −2 + 60 × = ,
18 3
so again we have matched the price of the claim.
Now at time 𝑡 = 0, we need to find (𝑥1 , 𝑦1 ) giving values of either 50, if the share price goes up, or 4/3, if the
share price goes down. We find
3
2 × 4 3
3 − 4 × 50 142 50 − 34 73
𝑥1 = 3 3 =− , 𝑦1 = = ,
2 − 4
3 120 − 60 90
and we have
142 73 158
𝑉0 = 𝑥1 + 80𝑦1 = − + 80 × = .
3 90 9
Our hedging portfolio is:

142 73
at time 0, (𝑥1 , 𝑦1 ) = (− , )
3 90
(−70, 1) if 𝜔1 = 1
at time 1, (𝑥2 , 𝑦2 ) = { 1
(−2, 18 ) if 𝜔1 = 0.
We can check the self-financing condition: if the share price goes up, then
142 73
𝑥1 + 120𝑦1 = − + 120 × = 50
3 90
and
𝑥2 + 120𝑦2 = −70 + 120 × 1 = 50.
Similarly, if the share price goes down, then

142 73 4
𝑥1 + 60𝑦1 = − + 60 × = ,
3 90 3
and
1 4
𝑥2 + 60𝑦2 = −2 + 60 × = .
18 3
Our portfolio, consisting of (𝑥1 , 𝑦1 ) and two possible outcomes for (𝑥2 , 𝑦2 ) is self-financing, and replicates the
returns of the random variable (𝑆𝑇 − 70)+ , whatever happens to the share prices.
3.5 Recombinant trees

In this section, we consider the special case where 𝑢 and 𝑑 have the same values at every node. This means that,
for example, if the stock moves up and then down, it will have the same value as if it moved down and then up.
We can represent the model using a diagram which has 𝑡 + 1 nodes at each time 𝑡, rather than 2𝑡 , as in Figure 2.
In this situation, we say that the diagram is recombinant (and in fact, it is no longer a tree).
26
su3
su2
su su2 d
St : s sud
sd sud2
sd2
sd3
Figure 2: A recombinant ’tree’.
Now, the possible values of the share price at time 𝑡 are
𝑆𝑡 ∈ {𝑠𝑢𝑘 𝑑𝑡−𝑘 , 𝑘 = 0, 1, … , 𝑡},
where 𝑘 denotes the total number of up-moves that have occurred (up to time 𝑡). This means that we can specify
each node in the tree by a pair (𝑡, 𝑘) with 𝑘 = 0, 1, … , 𝑡.
The contingent claim Φ(𝑆𝑇 ) can be replicated using a self-financing portfolio. If 𝑉𝑡 (𝑘) denotes the value of the
portfolio at node (𝑡, 𝑘), then 𝑉𝑡 (𝑘) can be computed recursively by
𝑉𝑇 (𝑘) = Φ(𝑠𝑢𝑘 𝑑𝑇 −𝑘 ),
1 (3.5)
𝑉𝑡 (𝑘) = [𝑞 𝑉 (𝑘 + 1) + 𝑞𝑑 𝑉𝑡+1 (𝑘)].
1 + 𝑟 𝑢 𝑡+1
where just as before 𝑞𝑢 and 𝑞𝑑 are the martingale probabilities at a node, given by
1+𝑟−𝑑 𝑢 − (1 + 𝑟)
𝑞𝑢 = , 𝑞𝑑 = .
(In fact, these are the martingal probabilities at every node, since 𝑢 and 𝑑 are fixed.)
We can also calculate the hedging portfolio: writing 𝑥𝑡 (𝑘) and 𝑦𝑡 (𝑘) for the amounts held at node (𝑡 − 1, 𝑘), we have
1 𝑢𝑉𝑡 (𝑘) − 𝑑𝑉𝑡 (𝑘 + 1) 𝑉𝑡 (𝑘 + 1) − 𝑉𝑡 (𝑘)

𝑥𝑡 (𝑘) = , 𝑦𝑡 (𝑘) = .
(1 + 𝑟)𝑡 𝑢−𝑑 𝑠𝑢𝑘 𝑑𝑡−1−𝑘 (𝑢 − 𝑑)
The recursive equations (3.5) are sometimes referred to as the binomial algorithm. We can use this algorithm
to work out the fair price of a contingent claim at time 0, by starting with the nodes at time 𝑡 = 𝑇 and working
backwards to 𝑡 = 0. This gives us the following result.
27
Theorem 3.2. The arbitrage-free price at 𝑡 = 0 of a contingent claim Φ(𝑆𝑇 ) is given by
1
Π0 (Φ) = 𝔼 (Φ(𝑆𝑇 )),
(1 + 𝑟)𝑇 ℚ
where ℚ is the martingale measure for 𝑆𝑇 . More explicitly, we can write

𝑇
1 𝑇
Π0 (Φ) = 𝑇
∑ ( )𝑞𝑢𝑘 𝑞𝑑𝑇 −𝑘 Φ(𝑠𝑢𝑘 𝑑𝑇 −𝑘 ).
(1 + 𝑟) 𝑘=0 𝑘
Proof. It is straightforward to check that the two expressions for Π0 (Φ) are equivalent, by calculating ℚ(𝑆𝑇 =
𝑠𝑢𝑘 𝑑𝑇 −𝑘 ). This is the probability that there are exactly 𝑘 up steps, and 𝑇 − 𝑘 down steps, among the 𝑇 total steps,
so is (𝑇𝑘 )𝑞𝑢𝑘 𝑞𝑑𝑇 −𝑘 . In other words, under the measure ℚ, the price 𝑆𝑇 equals 𝑆0 𝑢𝑌 𝑑𝑇 −𝑌 , where 𝑌 ∼ Bin(𝑇 , 𝑞𝑢 ), and
therefore
𝑇
𝔼ℚ [Φ(𝑆𝑇 )] = 𝔼[Φ(𝑆0 𝑢𝑌 𝑑𝑇 −𝑌 )] = ∑ Φ(𝑆0 𝑢𝑘 𝑑𝑇 −𝑘 )ℙ[𝑌 = 𝑘].
𝑘=0
It is also easy to see that the explicit expression for Π0 (Φ) follows by applying the equations (3.5). Indeed, it is
possible to show by induction on 𝑡, that the value 𝑉𝑇 −𝑡 (𝑖) of the portfolio at a node (𝑇 − 𝑡, 𝑖), for 𝑇 − 𝑡 ∈ [0, 𝑇 ] and
𝑖 = 0, … , 𝑇 − 𝑡, satisfies
𝑡
1 𝑡
𝑉𝑇 −𝑡 (𝑖) = ∑ ( )𝑞 𝑘 𝑞 𝑡−𝑘 Φ(𝑠𝑢𝑖+𝑘 𝑑𝑇 −𝑖−𝑘 ),
(1 + 𝑟)𝑡 𝑘=0 𝑘 𝑢 𝑑
and then by the Law of One Price, we have that Π0 (Φ) = 𝑉0 (0), which gives the required expression.
𝑡
Exercise: work out the above proof by induction. Hint: in the inductive step use the identity (𝑘−1 ) + (𝑘𝑡 ) = (𝑡+1
𝑘 ).
3.6 The Cox–Ross–Rubinstein formula

We use the pricing formula we found in the previous section to give a formula for the price of a European call option,
known as the Cox–Ross–Rubinstein formula.
As usual, let 𝐾 be the strike price of the call option, and 𝑇 the expiry date. The Cox–Ross–Rubinstein formula for
the price of such a call option assumes that the underlying stock can be modelled by the recombinant tree model
we studied in the previous section, for appropriate choices of 𝑢, 𝑑 and 𝑟. As it stands, we haven’t yet considered
how suitable the binomial model is as a model of stock price fluctuations. We’ll look at this question in Chapter 5,
but for now we suppose that we have some appropriately chosen values for 𝑢, 𝑑 and 𝑟, and we price the option in
terms of these parameters.
We can use the formula from Theorem 3.2 to price the call option, by setting Φ(𝑆𝑇 ) = (𝑆𝑇 − 𝐾)+ . In fact, we give
a more general form, for the price at time 𝑇 − 𝑡 ∈ [0, 𝑇 ].
Theorem 3.3 (Cox-Ross-Rubinstein formula). The price at time 𝑇 − 𝑡 of a European call option with strike price
𝐾 and expiry date 𝑇 is given by
𝑡 𝑡
𝑡 𝐾 𝑡
Π𝑇 −𝑡 = 𝑆𝑇 −𝑡 ∑ ( )𝑚𝑘𝑢 𝑚𝑡−𝑘
𝑑 − 𝑡
∑ ( )𝑞𝑢𝑘 𝑞𝑑𝑡−𝑘 , (3.6)
𝑘=𝑘⋆
𝑘 (1 + 𝑟) 𝑘=𝑘⋆ 𝑘
𝑞𝑢 𝑢 𝑞 𝑑
where 𝑞𝑢 , 𝑞𝑑 are the risk-neutral probabilities, 𝑚𝑢 = , 𝑚𝑑 = 𝑑 and 𝑘⋆ is the smallest integer 𝑘 that satisfies
1+𝑟 1+𝑟
𝑢 𝐾
𝑘 log ( ) > log ( ). (3.7)
𝑑 𝑆𝑇 −𝑡 𝑑𝑡
Proof. The price Π𝑇 −𝑡 (Φ) at time 𝑇 − 𝑡 of the claim Φ(𝑆𝑇 ) is a function of the share price 𝑆𝑇 −𝑡 given by
𝑡
1 𝑡
𝑡
∑ ( )𝑞𝑢𝑘 𝑞𝑑𝑡−𝑘 Φ(𝑆𝑇 −𝑡 𝑢𝑘 𝑑𝑡−𝑘 ).
(1 + 𝑟) 𝑘=0 𝑘
28
Since Φ(𝑥) = (𝑥 − 𝐾)+ , the terms appearing in this sum will be zero unless 𝑆𝑇 −𝑡 𝑢𝑘 𝑑𝑡−𝑘 > 𝐾. We can rewrite this
condition as (𝑢/𝑑)𝑘 > 𝐾/(𝑆𝑇 −𝑡 𝑑𝑡 ), and since 𝑢/𝑑 > 1, there is a smallest integer 𝑘⋆ given by (3.7), such that if
𝑘 ≥ 𝑘⋆ then the condition is satisfied. (Note that if 𝑘⋆ > 𝑡 then the sum becomes zero, and Π𝑇 −𝑡 = 0.)
In other words, we can write the price of a European call option as
𝑡
1 𝑡
Π𝑇 −𝑡 = 𝑡
∑ ( )𝑞𝑢𝑘 𝑞𝑑𝑡−𝑘 (𝑆𝑇 −𝑡 𝑢𝑘 𝑑𝑡−𝑘 − 𝐾)
(1 + 𝑟) 𝑘=𝑘⋆ 𝑘
and multiplying out the bracket and using the expressions for 𝑚𝑢 and 𝑚𝑑 yields the formula (3.6).
Example 3.8
Consider a 3-period model with 𝑢 = 1.2, 𝑑 = 0.8 and 𝑟 = 0.1. Suppose the share price at time 0 is 𝑆0 = 100.
We calculate the CRR price at time 0 of a European call option with strike price 70 and expiry date 𝑇 = 3.
First, we calculate the martingale probabilities:
1.1 − 0.8 3 1.2 − 1.1 1
𝑞𝑢 = = , 𝑞𝑑 = = .
1.2 − 0.8 4 1.2 − 0.8 4
Then, we calculate 𝑚𝑢 and 𝑚𝑑 :
3 1.2 9 1 0.8 2
𝑚𝑢 = × = , 𝑚𝑑 = × = .
4 1.1 11 4 1.1 11
To find the correct value of 𝑘⋆ , we look at the possible values for the share price at time 𝑇 . We see that 𝑆3
can be one of 51.2, 76.8, 115.2 or 172.8, and so the call option only has positive value at time 𝑇 if at least
one up-move occurs, so 𝑘⋆ = 1.
(Alternatively, we can calculate log(𝐾/𝑆0 𝑑3 )/ log(𝑢/𝑑) = log(70/51.2)/ log(3/2) ≈ 0.771, and find that the
smallest integer greater than this is 𝑘⋆ = 1.)
Then the CRR price is given by
3 3
𝐾
𝑉0 = 𝑆0 ∑ (𝑘3)𝑚𝑘𝑢 𝑚3−𝑘 − ∑ (3)𝑞 𝑘 𝑞 3−𝑘
𝑘=1
𝑑
(1 + 𝑟)3 𝑘=1 𝑘 𝑢 𝑑
1323 70 63 253575
= 100 × − × = ≈ 47.63.
1331 1.331 64 5324
𝑞𝑢 𝑢 𝑞 𝑑
Remark. Note that 𝑚𝑢 , 𝑚𝑑 > 0 and 𝑚𝑢 + 𝑚𝑑 = + 𝑑 = 1, so that 𝑚𝑢 and 𝑚𝑑 can also be interpreted as
1+𝑟 1+𝑟
probabilities. Hence we can write (3.6) as
𝐾
Π𝑇 −𝑡 = 𝑆𝑇 −𝑡 ℙ[𝑋𝑡 ≥ 𝑘⋆ ] − ℙ[𝑌𝑡 ≥ 𝑘⋆ ],
(1 + 𝑟)𝑡
where 𝑋𝑡 and 𝑌𝑡 are Binomial random variables: 𝑋𝑡 ∼ Bin(𝑡, 𝑚𝑢 ), 𝑌𝑡 ∼ Bin(𝑡, 𝑞𝑢 ).
29
4 Filtrations, Conditional Expectation, and Martingales

In this chapter, we extend the probability theory behind our pricing of contingent claims in the binomial model.
This will allow us to put the important financial concepts on a proper mathematical foundation, and to prepare for
Epiphany term. In continuous time, our intuition is less helpful, and we will need to approach the material using a
rigorous mathematical theory.
4.1 The probability space

Remember that in the multi-period binomial model, we are working in a probability space (Ω, ℱ, ℙ), in which:
• Ω = {0, 1}𝑇 is the set containing all strings of length 𝑇 in which every character is a 0 or a 1
• ℱ = ℱ𝑇 is the Borel 𝜎-algebra on Ω, i.e., ℱ𝑇 = 2Ω .
• we have fundamentally not been interested in learning more about ℙ, as long as we know that ℙ(𝜔) > 0 for
every 𝜔 ∈ Ω.
We can view this probability space as equivalent to a coin-toss space, in which we toss a sequence of coins and
record a 1 each time a coin lands heads, and a 0 each time a coin lands tails. In case you’ve forgotten how equivalent
measures work, the next subsection might be helpful:
4.1.1 Side note: Equivalent measures

Recall that two probability measures ℙ and ℚ are equivalent to each other if
ℙ(𝐴) > 0 if and only if ℚ(𝐴) > 0.
The measures do not need to assign the same probabilities to each event, but they should agree that either 𝐴 is a
possible event (both ℙ(𝐴) and ℚ(𝐴) are strictly positive), or that 𝐴 is an impossible event (both are zero).
Example 4.1
In the context of the multi-period binomial mode, we have two equivalent measures: ℙ, the ‘’objective
measure”, representing our estimate of how likely a head is to occur in reality, and ℚ, the martingale measure
that we use for pricing.
Example 4.2
For any 0 < 𝑝 < 1, we can define a probability measure ℙ on Ω𝑇 as follows.
Let #heads(𝜔) be the number of times that 1 appears in 𝜔 = 𝜔1 … 𝜔𝑇 , and #tails(𝜔) = 𝑇 − #heads(𝜔) be
the number of 0s Then
ℙ(𝜔) = 𝑝#heads(𝜔) (1 − 𝑝)#tails(𝜔) ,

and for 𝐴 ∈ ℱ,
ℙ(𝐴) = ∑ ℙ(𝜔).
𝜔∈𝐴
For a different 0 < 𝑞 < 1, we can define a different measure ℚ on Ω𝑇 with
ℚ(𝜔) = 𝑞 #heads(𝜔) (1 − 𝑞)#tails(𝜔) , ℚ(𝐴) = ∑ ℚ(𝜔).

𝜔∈𝐴
Here ℙ and ℚ are equivalent to each other as long as both 0 < 𝑝 < 1 and 0 < 𝑞 < 1.
Remark. In the recombinant multi-period binomial model, the martingale measure ℚ is defined exactly as in Example
4.2, and we set 𝑞 = 𝑞𝑢 . On the other hand if 𝑢 and 𝑑 are not fixed throughout the model, then for each 𝜔 ∈ Ω𝑇 ,
ℚ(𝜔) becomes a product of 𝑞𝑢 s and 𝑞𝑑 s, which can be found by “multiplying along the branches of the tree”.
30
4.1.2 Back to probability spaces

Remember that in Chapter 3, we defined the share prices in terms of the random variables 𝑍𝑡 :
𝑆𝑡 (𝜔) = 𝑍𝑡 (𝜔)𝑆𝑡−1 (𝜔),
where 𝑍1 , 𝑍2 , … , 𝑍𝑇 are random variables defined on Ω, with
𝑢 if 𝜔𝑡 = 1,
𝑍𝑡 (𝜔) = {
𝑑 if 𝜔𝑡 = 0.
Example 4.3
In the recombinant market with 𝑠 = 4, 𝑇 = 3, 𝑢 = 2, and 𝑑 = 12 , we have
𝑆0 (𝜔1 𝜔2 𝜔3 ) = 4, for all 𝜔 ∈ Ω𝑇 ;

8 if 𝜔1 = 1,
𝑆1 (𝜔1 𝜔2 𝜔3 ) = {
2 if 𝜔1 = 0;
⎧16 if 𝜔1 = 𝜔2 = 1,
{
𝑆2 (𝜔1 𝜔2 𝜔3 ) = ⎨4 if 𝜔1 ≠ 𝜔2 ,
{1 if 𝜔 = 𝜔 = 0;
⎩ 1 2
and
⎧32 if 𝜔1 = 𝜔2 = 𝜔3 = 1,
{
{8 if there are two 1s and one 0
𝑆3 (𝜔1 𝜔2 𝜔3 ) = ⎨
{2 if there are two 0s and one 1
{1 if 𝜔1 = 𝜔2 = 𝜔3 = 0.
⎩2
In the spirit of simplifying the notation in this chapter, we are going to assume that 𝑢 and 𝑑 are fixed every-
where. All of the theory works in exactly the same way when they’re not, but we have to worry about which 𝑢, or
which 𝑞𝑢 , we’re using all the time.
Under this nice assumption, the 𝑍s are independent and identically distributed random variables. They encapsulate
all the randomness in the probability space (Ω𝑇 , ℱ, ℙ): the 2𝑇 elements of Ω𝑇 are in one-to-one correspondence
with the 2𝑇 possible real-valued sequences (𝑧1 , 𝑧2 , … , 𝑧𝑇 ) describing the possible values of (𝑍1 , 𝑍2 , … , 𝑍𝑇 ). In other
words, if we know the values of all of 𝑍1 , 𝑍2 , … , 𝑍𝑇 , then we know which state 𝜔 ∈ Ω𝑇 we are in, and vice versa.
We say that ℱ = 2Ω𝑇 is equal to the 𝜎-algebra generated by the random variables 𝑍1 , 𝑍2 , … , 𝑍𝑇 , and we write
ℱ = 𝜎(𝑍1 , 𝑍2 , … , 𝑍𝑇 ).
4.2 Partial information

If we only know about the first 𝑡 coin tosses (that is, we know what 𝑍1 , 𝑍2 , … , 𝑍𝑡 are, but we do not know
𝑍𝑡+1 , … , 𝑍𝑇 ), we can still say quite a lot about how the share prices evolve: we can already write down the values
of 𝑆0 , 𝑆1 , … , 𝑆𝑡 . Intuitively, this seems perfectly reasonable, but before we can say anything probabilistically, we’d
better sort out which space we’re working in.
We define a sequence of 𝜎-algebras as follows:
• Take ℱ0 = {∅, Ω}.
• Let 𝐴1 = {𝜔 ∈ Ω ∶ 𝜔1 = 1}, and let 𝐴0 = {𝜔 ∈ Ω ∶ 𝜔1 = 0}. Define
ℱ1 = {∅, 𝐴1 , 𝐴0 , Ω}.
• Let 𝐴11 = {𝜔 ∈ Ω ∶ 𝜔1 = 𝜔2 = 1}, 𝐴10 = {𝜔 ∈ Ω ∶ 𝜔1 = 1, 𝜔2 = 0}, 𝐴01 = {𝜔 ∈ Ω ∶ 𝜔1 = 0, 𝜔2 = 1}, and
𝐴00 = {𝜔 ∈ Ω ∶ 𝜔1 = 𝜔2 = 0}. Define
ℱ2 = 𝜎(𝐴11 , 𝐴10 , 𝐴01 , 𝐴00 )
as the smallest 𝜎-algebra containing these four sets.
31
• Continue in this way, defining 2𝑡 subsets which encapsulate the first 𝑡 coin tosses, and building ℱ𝑡 as the
smallest 𝜎-algebra containing these subsets, until we reach ℱ𝑇 = 2Ω .
Exercise 4.1
Construct ℱ2 ; what does ℱ𝑡 look like?
Constructing these 𝜎-algebras helps us to define what we mean by partial information. We say that we have observed
a 𝜎-algebra 𝒢 if, for every event 𝐴 ∈ 𝒢, we know whether 𝐴 has occurred, that is, if we know the value of the
indicator function 1(𝐴). Now at time 𝑡, we know what the share price has been for every 𝑘 ≤ 𝑡; that is, we know all
of the values 𝑆0 , 𝑆1 , … , 𝑆𝑡 . So our 𝜎-algebra ℱ𝑡 encodes all of the information about the share prices up to time 𝑡.
Remark. In this chapter, we move back and forth between thinking about having observed a 𝜎-algebra, and having
observed some random variables. Because ℱ𝑡 encodes all of the randomness in the first 𝑡 share prices, and nothing
else, we can write ℱ𝑡 = 𝜎(𝑆0 , 𝑆1 , … , 𝑆𝑡 ), or even ℱ𝑡 = 𝜎(𝑍1 , … , 𝑍𝑡 ).
Observing the 𝜎-algebra ℱ𝑡 means knowing exactly which of the events in it have occurred, so when we say that
we have observed ℱ𝑡 we can deduce the values of 𝑆1 , … , 𝑆𝑡 . Similarly, we can go in the other direction: if we know
all of the values 𝑆1 , … , 𝑆𝑡 , then for every event 𝐴 ∈ ℱ𝑡 we know the value of 1(𝐴).
Definition 4.1. A filtration is a non-descending sequence of 𝜎-algebras
ℱ0 ⊆ ℱ1 ⊆ ⋯ ⊆ ℱ𝑛 .
Filtrations help us to model partial information which is accumulating over time. We assume that we have a perfect
memory and never forget anything that’s happened; we just keep increasing the amount of information we have.
Exercise 4.2
Prove (convince yourself) that the sequence of 𝜎-algebras we have just constructed is a filtration: for each 𝑡,
ℱ𝑡 ⊂ ℱ𝑡+1 .
Definition 4.2. A random variable 𝑋 ∶ Ω → ℝ is measurable with respect to a 𝜎-algebra 𝒢 if for every Borel set
𝐵 ∈ ℬ(ℝ),
{𝜔 ∈ Ω ∶ 𝑋(𝜔) ∈ 𝐵} ∈ 𝒢.
In other words, the random variable 𝑋 is measurable with respect to 𝒢 if observing 𝒢 allows us to determine the
value of 𝑋.
Example 4.4
We have already met some random variables which are measurable with respect to filtrations:
• The share price 𝑆𝑡 is always measurable with respect to ℱ𝑡 , because it only depends on the first 𝑡 coin
tosses.
• The holdings in a portfolio at time 𝑡 − 1, 𝑥𝑡 and 𝑦𝑡 , are allowed to be random as long as they are
measurable with respect to ℱ𝑡−1 .
• A random variable 𝑇 is a stopping time with respect to the filtration (ℱ𝑡 ) if, for every 𝑡, {𝑇 ≤ 𝑡} ∈ ℱ𝑡 .
The maximal value that the share price will take over the period [0, 𝑇 ] is not measurable with respect to ℱ𝑡
for any 𝑡 < 𝑇 , because we will only know what the maximum is when we reach time 𝑇 .
Since the share price 𝑆𝑡 is measurable with respect to ℱ𝑡 , we sometimes write 𝑆𝑡 (𝜔1 𝜔2 … 𝜔𝑡 ), or even 𝑆𝑡 (⋯ 𝜔𝑡 ),
rather than 𝑆𝑡 (𝜔1 𝜔2 … 𝜔𝑇 ).
Example 4.5
In Example 4.3, we can write 𝑆1 (1) = 8, or 𝑆2 (01) = 4.
32
4.3 Conditional expectation

If you have forgotten how conditional expectation with respect to a random variable works, go and check the
Prerequisite notes then come back here.
Definition 4.3. Given a random variable 𝑋 on Ω𝑇 such that 𝔼ℚ [|𝑋|] < ∞, the conditional expectation of 𝑋
with respect to 𝒢 is the unique random variable 𝜉 which satisfies:
• 𝔼ℚ [|𝜉|] < ∞.
• 𝜉 is measurable with respect to 𝒢
• For every event 𝐴 ∈ 𝒢, 𝔼ℚ [𝜉 1(𝐴)] = 𝔼ℚ [𝑋 1(𝐴)].
We denote 𝜉 = 𝔼ℚ [𝑋|𝒢].
We can think of the conditional expectation with respect to a sub-𝜎-algebra as “the expectation of 𝑋, given the
partial information contained in 𝒢”. If we use the sub-𝜎-algebra generated by a random variable 𝑌 , 𝒢 = 𝜎(𝑌 ), we
will rediscover our original Probability-1 definition of conditional expectation (𝔼[𝑋|𝑌 ] = 𝑔(𝑌 ), and so on).
If 𝑋𝑇 = Φ(𝑆1 , … , 𝑆𝑇 ) then the conditional expectation 𝔼ℚ [𝑋𝑇 |ℱ𝑡 ] can be thought of as the expected value of 𝑋𝑇 if
we fix the outcomes of 𝑆1 , … , 𝑆𝑡 and average over the remaining randomness that determines 𝑆𝑡+1 , … , 𝑆𝑇 , i.e., over
𝑍𝑡+1 , … , 𝑍𝑇 .
This means that, for 𝜔 ∈ Ω𝑇 , we have
𝔼ℚ [𝑋𝑇 |ℱ𝑡 ](⋯ 𝜔𝑡 ) = 𝑞𝑢 𝔼ℚ [𝑋𝑇 |ℱ𝑡+1 ](⋯ 𝜔𝑡 1) + 𝑞𝑑 𝔼ℚ [𝑋𝑇 |ℱ𝑡+1 ](⋯ 𝜔𝑡 0). (4.1)
Example 4.6
In the recombinant market with 𝑠 = 4, 𝑇 = 3, 𝑢 = 2, and 𝑑 = 12 (the same market as in Example 4.3), we
have
1 − 12 1 2−1 2
𝑞𝑢 = = , 𝑞𝑑 = = .
2 − 12 3 2 − 21 3
Let’s calculate the values of 𝔼ℚ [𝑆2 |ℱ1 ] and 𝔼ℚ [𝑆3 |ℱ1 ]. The two events in ℱ1 are 𝜔1 = 1 and 𝜔1 = 0, so we
need to do two sets of calculations.
1 2
𝔼ℚ [𝑆2 |ℱ1 ](1) = 16 × + 4 × = 8, (4.2)
3 3
1 2
𝔼ℚ [𝑆2 |ℱ1 ](0) = 4 × + 1 × = 2. (4.3)
3 3
Here 𝔼ℚ [𝑆2 |ℱ1 ] is a random variable defined on Ω3 which takes value 8 whenever 𝜔1 = 1, and 2 whenever
𝜔1 = 0. We see that, in both cases, 𝔼ℚ [𝑆2 |ℱ1 ] = 𝑆1 .
Next,
1 2 1 2 2 2
𝔼ℚ [𝑆3 |ℱ1 ](1) = 32 × ( ) + 8 × 2 × × + 2 × ( ) = 8, (4.4)
3 3 3 3
2
1 1 2 1 2 2
𝔼ℚ [𝑆3 |ℱ1 ](0) = 8 × ( ) + 2 × 2 × × + × ( ) = 2. (4.5)
3 3 3 2 3
Once again, we have 𝔼ℚ [𝑆3 |ℱ1 ] = 𝑆1 .
Important Properties of Conditional Expectations

• Linearity. For constants 𝑎1 , 𝑎2 , we have
𝔼[𝑎1 𝑋 + 𝑎2 𝑌 |ℱ𝑡 ] = 𝑎1 𝔼[𝑋|ℱ𝑡 ] + 𝑎2 𝔼[𝑌 |ℱ𝑡 ].
• Taking out what is known. If 𝑋 depends only on the first 𝑡 coin flips, then
𝔼[𝑋𝑌 |ℱ𝑡 ] = 𝑋 ⋅ 𝔼[𝑌 |ℱ𝑡 ].
33
• Iterated conditioning. If 𝑠 ≤ 𝑡 then

𝔼[𝔼[𝑋|ℱ𝑡 ]|ℱ𝑠 ] = 𝔼[𝑋|ℱ𝑠 ].
• Independence. If 𝑋 depends only on coin tosses 𝑡 + 1 to 𝑇 , then
𝔼[𝑋|ℱ𝑡 ] = 𝔼[𝑋].
4.4 Martingales
Definition 4.4. A sequence of integrable random variables 𝑌0 , 𝑌1 , ⋯ , 𝑌𝑇 is called a martingale under the measure
ℚ if for each 𝑡, the value of 𝑌𝑡 depends on the outcome of the first 𝑡 coin flips (we say the sequence is adapted to
the filtration) and
𝔼ℚ [𝑌𝑡+1 |ℱ𝑡 ] = 𝑌𝑡 , 𝑡 = 0, 1, … , 𝑇 − 1.
Martingales arose out of the study of gambling models. If 𝑌0 , 𝑌1 , … , 𝑌𝑡 represent the fortune of a player, then
𝔼ℚ [𝑌𝑡+1 |ℱ𝑡 ] represents the amount of money the player can expect to have after the next game, given their position
and knowledge at the end of game 𝑡. The player will consider the game to be fair if 𝔼ℚ [𝑌𝑡+1 |ℱ𝑡 ] = 𝑌𝑡 ; that is, if the
sequence is a martingale.
Here are some simple properties of martingales:
• For any 𝑡, we have 𝔼ℚ [𝑋𝑡 ] = 𝔼ℚ [𝑋0 ], because
𝔼ℚ [𝑋𝑡 ] = 𝔼ℚ [𝔼ℚ [𝑋𝑡 |ℱ𝑡−1 ]] = 𝔼ℚ [𝑋𝑡−1 ] for all 𝑡 ≥ 1.
• Whenever 𝑚 > 𝑛,
𝔼ℚ [𝑋𝑚 |ℱ𝑛 ] = 𝑋𝑛 .
• Given a filtration (ℱ𝑡 )𝑡≤𝑇 and a random variable 𝑍 which is measurable with respect to ℱ𝑇 , the sequence
𝑋𝑡 = 𝔼ℚ [𝑍|ℱ𝑡 ]
is always a martingale with respect to (ℱ𝑡 )𝑡≤𝑇 .
4.4.1 Martingales in the multi-period binomial model
Theorem 4.1. The sequence of discounted stock prices

𝑆𝑡
, 𝑡 = 0, 1, 2, … , 𝑇 ,
(1 + 𝑟)𝑡
is a martingale under the risk-neutral measure ℚ.
Remark. The converse of this statement is also true in the multi-period binomial model. That is to say, the
martingale measure ℚ in an arbitrage-free and complete multi-period binomial model is determined by the property
𝑆𝑡
that (1+𝑟) 𝑡 forms a martingale sequence under ℚ.
Proof. The discounted stock prices are definitely adapted to the filtration, because the sequence 𝑆1 , … , 𝑆𝑡 is adapted
to the filtration.
For the conditional expectation, we have
𝑆𝑡+1 𝑆 (𝜔 ⋯ 𝜔 1) 𝑆 (𝜔 ⋯ 𝜔 0)
𝔼ℚ [ |ℱ ](𝜔1 ⋯ 𝜔𝑡 ) = 𝑞𝑢 𝑡+1 1 𝑡+1𝑡 + 𝑞𝑑 𝑡+1 1 𝑡+1𝑡
(1 + 𝑟)𝑡+1 𝑡 (1 + 𝑟) (1 + 𝑟)
𝑆 (𝜔 ⋯ 𝜔 )
= 𝑡 1 𝑡+1𝑡 [𝑞𝑢 𝑢 + 𝑞𝑑 𝑑]
(1 + 𝑟)
𝑆𝑡 (𝜔1 ⋯ 𝜔𝑡 )
= .
(1 + 𝑟)𝑡
34
Theorem 4.2. For any self-financing portfolio in which (𝑥𝑡+1 , 𝑦𝑡+1 ) is always measurable with respect to ℱ𝑡 , the
discounted value process
𝑉𝑡
, 𝑡 = 0, 1, … , 𝑇 ,
(1 + 𝑟)𝑡
is a martingale under the risk-neutral measure.
Remark. For this proof, we need to rewrite the self-financing condition in a new form: the wealth equation. The
definition of the value, 𝑉𝑡 = 𝑥𝑡+1 𝐵𝑡 + 𝑦𝑡+1 𝑆𝑡 , and the self-financing condition 𝑉𝑡+1 = 𝑥𝑡+1 𝐵𝑡+1 + 𝑦𝑡+1 𝑆𝑡+1 , together
imply that
𝑉𝑡+1 = 𝑦𝑡+1 𝑆𝑡+1 + (1 + 𝑟)(𝑉𝑡 − 𝑦𝑡+1 𝑆𝑡 ).
𝑉𝑡
Proof. First, we check that (1+𝑟) 𝑡 is adapted to the filtration; since 𝑉𝑡 = 𝑥𝑡+1 𝐵𝑡 + 𝑦𝑡+1 𝑆𝑡 and all parts of this are
measurable with respect to ℱ𝑡 , we can continue.

For the conditional expectation, we have
𝑉𝑡+1 𝑦 𝑆 + (1 + 𝑟)(𝑉𝑡 − 𝑦𝑡+1 𝑆𝑡 )
𝔼ℚ [ 𝑡+1
|ℱ𝑡 ] = 𝔼ℚ [ 𝑡+1 𝑡+1 |ℱ𝑡 ]
(1 + 𝑟) (1 + 𝑟)𝑡+1
𝑦 𝑆 (1 + 𝑟)(𝑉𝑡 − 𝑦𝑡+1 𝑆𝑡 )
(linearity) → = 𝔼ℚ [ 𝑡+1 𝑡+1 𝑡+1
|ℱ𝑡 ] + 𝔼ℚ [ |ℱ𝑡 ]
(1 + 𝑟) (1 + 𝑟)𝑡+1
𝑆𝑡+1 𝑉 − 𝑦𝑡+1 𝑆𝑡
(taking out what is known) → = 𝑦𝑡+1 𝔼ℚ [ 𝑡+1
|ℱ𝑡 ] + 𝑡
(1 + 𝑟) (1 + 𝑟)𝑡
𝑆𝑡 𝑉 − 𝑦𝑡+1 𝑆𝑡
= 𝑦𝑡+1 + 𝑡
(1 + 𝑟)𝑡 (1 + 𝑟)𝑡
𝑉𝑡
= ,
(1 + 𝑟)𝑡
showing that 𝑉𝑡 /(1 + 𝑟)𝑡 is a martingale.
This proves the correctness of the risk-neutral valuation formula for pricing contingent claims:
1
𝑉𝑡 = 𝔼 [𝑉 |ℱ ].
(1 + 𝑟)𝑇 −𝑡 ℚ 𝑇 𝑡
We finish this section with a version of the First Fundamental Theorem for the multi-period binomial model. We
require a definition of arbitrage on the multi-period binomial model.
Definition 4.5. A portfolio ℎ ≡ (ℎ𝑡 = (𝑥𝑡 , 𝑦𝑡 ), 𝑡 = 0, 1, … , 𝑇 + 1) on the multi-period binomial model ℳ = (𝐵𝑡 , 𝑆𝑡 )
is an arbitrage portfolio if it is self-financing and its value process 𝑉𝑡ℎ = 𝑥𝑡+1 𝐵𝑡 + 𝑦𝑡+1 𝑆𝑡 satisfies:
𝑉0ℎ = 0, ℙ(𝑉𝑇ℎ ≥ 0) = 1, ℙ(𝑉𝑇ℎ > 0) > 0.
Theorem 4.3. The following conditions are equivalent for a multi-period binomial model ℳ = (𝐵𝑡 , 𝑆𝑡 ), 𝑡 =
0, 1, … , 𝑇 , with interest rate 𝑟.
(1) The model is arbitrage-free according to Definition 4.5.
(2) The condition 𝑑 < 1 + 𝑟 < 𝑢 holds, where 𝑑 < 𝑢 are the two possible values of 𝑍𝑡 = 𝑆𝑡 /𝑆𝑡−1 at each time 𝑡.
(𝑍𝑡 equals 𝑢 with probability 𝑝 and 𝑑 with probability 1 − 𝑝 for some 0 < 𝑝 < 1).
(3) There is a measure ℚ defined by
1+𝑟−𝑑 𝑢 − (1 + 𝑟)
𝑞𝑢 = , 𝑞𝑑 =
𝑆𝑡
at each node of the tree, such that (1+𝑟)𝑡 is a martingale under ℚ.
Proof. We have done most of the work needed to prove this theorem. Let us show the implications (1) ⇒ (2) ⇒
(3) ⇒ (1).
35
• (1) ⇒ (2): Consider the number of periods 𝑇 in the model. If 𝑇 = 1 then the implication holds by Theorem
2.1. For 𝑇 > 1, the 1-period model obtained by observing the market from 𝑡 = 0 to 𝑡 = 1 has no arbitrage,
and so 𝑑 < 1 + 𝑟 < 𝑢 by Theorem 2.1.
• (2) ⇒ (3): This implication is Theorem 4.1 above.
• (3) ⇒ (1): Suppose ℎ𝑡 = (𝑥𝑡 , 𝑦𝑡 ) is a self-financing portfolio that satisfies the conditions ℙ(𝑉0ℎ = 0) = 1 and
ℙ(𝑉𝑇ℎ ≥ 0) = 1. Since the measure ℚ is equivalent to ℙ, it follows that ℚ(𝑉0ℎ = 0) = 1 and ℚ(𝑉𝑇ℎ ≥ 0) = 1.
𝑉𝑡
Now, by Theorem 4.2, (1+𝑟) 𝑡 is a martingale under ℚ and so in particular,
𝑉𝑇
𝔼ℚ [ ] = 𝔼ℚ [𝑉0 ] = 0.
(1 + 𝑟)𝑇
This shows that 𝔼ℚ [𝑉𝑇 ] = 0 and thus 𝑉𝑇 is a non-negative random variable with mean 0, which implies 𝑉𝑇 is
identically zero: ℚ(𝑉𝑇 > 0) = 0. Consequently, ℙ(𝑉𝑇 > 0) = 0 as well so ℎ is not an arbitrage portfolio.
36
5 Pricing options from around the world

5.1 An algorithm to price any derivative asset
This algorithm allows us to price any derivative asset via its hedging portfolio. Suppose we have an asset with
𝑋𝑇 (𝜔), which is measurable with respect to ℱ𝑇 .
First, we define 𝑇 random variables, 𝑋𝑇 −1 , … , 𝑋1 , 𝑋0 , on our probability space, using the recursive formula
1
𝑋𝑡 (⋯ 𝜔𝑡 ) = (𝑞 𝑋 (⋯ 𝜔𝑡 1) + 𝑞𝑑 𝑋𝑡+1 (⋯ 𝜔𝑡 0)).
1 + 𝑟 𝑢 𝑡+1
Here 𝑞𝑢 and 𝑞𝑑 are the martingale probabilities at each node, given by
1+𝑟−𝑑 𝑢 − (1 + 𝑟)
𝑞𝑢 = , 𝑞𝑑 = .
Next, we let
𝑋𝑡 (⋯ 𝜔𝑡−1 1) − 𝑋𝑡 (⋯ 𝜔𝑡−1 0)
𝑦𝑡 (⋯ 𝜔𝑡−1 ) = , 𝑡 = 1…𝑇.
𝑆𝑡 (⋯ 𝜔𝑡−1 1) − 𝑆𝑡 (⋯ 𝜔𝑡−1 0)
Finally, set 𝑉0 = 𝑋0 for all 𝜔 ∈ Ω𝑇 . Now, for 𝑡 = 0, … , 𝑇 − 1, let
𝑉𝑡+1 = 𝑦𝑡+1 𝑆𝑡+1 + (1 + 𝑟)(𝑉𝑡 − 𝑦𝑡+1 𝑆𝑡 ). (Wealth Equation)
Theorem 5.1. The portfolio given by 𝑦𝑡 (⋯ 𝜔𝑡−1 ) as defined above, with
𝑉𝑡−1 (⋯ 𝜔𝑡−1 ) − 𝑦𝑡 (⋯ 𝜔𝑡−1 )𝑆𝑡−1 (⋯ 𝜔𝑡−1 )

𝑥𝑡 (⋯ 𝜔𝑡−1 ) = ,
𝐵𝑡−1
is self-financing, has value process 𝑉𝑡 as defined above, and replicates the contingent claim 𝑋. Moreover, for every
𝑡 and 𝜔, 𝑉𝑡 (⋯ 𝜔𝑡 ) = 𝑋𝑡 (⋯ 𝜔𝑡 ).
Proof. First, it is clear from the definition of 𝑥𝑡 that for each 𝑡, the relation
𝑉𝑡 = 𝑥𝑡+1 𝐵𝑡 + 𝑦𝑡+1 𝑆𝑡
holds for every 𝜔 ∈ Ω𝑇 .

Next, to see that the self-financing condition holds, we check that
𝑉𝑡+1 = 𝑥𝑡+1 𝐵𝑡+1 + 𝑦𝑡+1 𝑆𝑡+1
for 𝑡 = 0, … , 𝑇 .
We have
𝑉𝑡 − 𝑦𝑡+1 𝑆𝑡
𝑥𝑡+1 𝐵𝑡+1 + 𝑦𝑡+1 𝑆𝑡+1 = 𝐵𝑡+1 + 𝑦𝑡+1 𝑆𝑡+1 (5.1)
𝐵𝑡
= (1 + 𝑟)(𝑉𝑡 − 𝑦𝑡+1 𝑆𝑡 ) + 𝑦𝑡+1 𝑆𝑡+1 (5.2)
= 𝑉𝑡+1 , (5.3)
by definition.
Now, to see that 𝑉 and 𝑋 always coincide, we use induction. The base case 𝑉0 = 𝑋0 is true by definition. Next,
we fix 𝜔1 ⋯ 𝜔𝑡 and use 𝑉𝑡 (𝜔1 ⋯ 𝜔𝑡 ) = 𝑋𝑡 (𝜔1 ⋯ 𝜔𝑡 ) to show that
𝑉𝑡+1 (⋯ 𝜔𝑡 1) = 𝑋𝑡+1 (⋯ 𝜔𝑡 1), and

𝑉𝑡+1 (⋯ 𝜔𝑡 0) = 𝑋𝑡+1 (⋯ 𝜔𝑡 0).
37
First, if 𝜔𝑡+1 = 1, we use the fact that 𝑆𝑡+1 (⋯ 𝜔𝑡 1) = 𝑢𝑆𝑡 (⋯ 𝜔𝑡 ) in the wealth equation to write
𝑉𝑡+1 (⋯ 𝜔𝑡 1) = 𝑦𝑡+1 𝑢𝑆𝑡 + (1 + 𝑟)(𝑉𝑡 − 𝑦𝑡+1 𝑆𝑡 ) (5.4)
= 𝑦𝑡+1 𝑆𝑡 (𝑢 − (1 + 𝑟)) + (1 + 𝑟)𝑉𝑡 . (5.5)
(Here, we have stopped writing ⋯ 𝜔𝑡 in the right hand side, to simplify the notation.)
Using the inductive hypothesis and the expression for 𝑦𝑡+1 , we get
𝑋𝑡+1 (1) − 𝑋𝑡+1 (0)
𝑉𝑡+1 (⋯ 𝜔𝑡 1) = 𝑆𝑡 (𝑢 − (1 + 𝑟)) + (1 + 𝑟)𝑋𝑡 (5.6)
𝑢𝑆𝑡 − 𝑑𝑆𝑡
𝑢 − (1 + 𝑟)
= (𝑋𝑡+1 (1) − 𝑋𝑡+1 (0)) + (1 + 𝑟)𝑋𝑡 (5.7)
𝑢−𝑑
= (𝑋𝑡+1 (1) − 𝑋𝑡+1 (0))𝑞𝑑 + [𝑞𝑢 𝑋𝑡+1 (1) + 𝑞𝑑 𝑋𝑡+1 (0)] (5.8)
= 𝑋𝑡+1 (1), (5.9)
as required.
Exercise: Check that, when 𝜔𝑡+1 = 0, we also have 𝑉𝑡+1 = 𝑋𝑡+1 .
5.2 Popular options

Some background on vocabulary
In many textbooks, options such as those in this section are described as “exotic”. I’ve chosen not to use that word
in these lecture notes; you can find an explanation put together by our decolonisation interns, here.
A digital (or binary) option is a contract whose payoff depends in a discontinuous way on the terminal price of
the underlying asset. We can describe the payoff functions succinctly using indicator functions: for an event 𝐴 we
write 1𝐴 for the random variable
1 if 𝜔 ∈ 𝐴,
1𝐴 (𝜔) = {
0 otherwise.
In other words, 1𝐴 takes the value 1 exactly when 𝐴 occurs, and 0 otherwise. For example,
1 if 𝑆𝑇 > 𝐾,
1{𝑆𝑇 >𝐾} = {
0 otherwise.
The payoff of the cash-or-nothing binary option is given by

𝑋 = 𝜂1{𝑆𝑇 >𝐾} for a call option,
𝑋 = 𝜂1{𝑆𝑇 <𝐾} for a put option,
where 𝜂 is a pre-specified amount; and the payoff of the asset-or-nothing binary option is given by
𝑋 = 𝑆𝑇 1{𝑆𝑇 >𝐾} (call),
𝑋 = 𝑆𝑇 1{𝑆𝑇 <𝐾} (put).
Gap options are contingent claims whose payoffs are given by

𝑋 = (𝑆𝑇 − 𝜂)1{𝑆𝑇 >𝐾} (call),
𝑋 = (𝜂 − 𝑆𝑇 )1{𝑆𝑇 <𝐾} (put).
Here again 𝜂 is a pre-specified amount.
Lookback options are contingent claims whose payoff depends not only on the terminal price of the underlying
asset but also on asset price fluctuations during the option’s life time. There are two standard lookback options.
The payoff of the standard lookback call option is given by
𝐿𝐶𝑇 = 𝑆𝑇 − 𝑆min ,
38
where 𝑆min = min𝑡∈[0,𝑇 ] 𝑆𝑡 is the minimum value of the stock price during its lifetime. The payoff of the standard
lookback put option is given by
𝐿𝑃𝑇 = 𝑆max − 𝑆𝑇 ,
where 𝑆max = max𝑡∈[0,𝑇 ] 𝑆𝑡 is the maximum value of the stock price.
Example 5.1
Let 𝑆0 = 4, 𝑢 = 2, 𝑑 = 12 , 𝑟 = 14 . We find the prices of the lookback option
𝑋3 = max{𝑆0 , 𝑆1 , 𝑆2 , 𝑆3 } − 𝑆3
at times 𝑡 = 0, 1, 2.
5
− 12 1 2 − 45
First we find 𝑞𝑢 = 4
= and 𝑞𝑑 = = 12 . Writing 𝑆max = max{𝑆0 , 𝑆1 , 𝑆2 , 𝑆3 }, we find the following
2 − 12 2
2 − 12
values for 𝑆max , 𝑆3 and 𝑋3 as functions of 𝜔 ∈ Ω3 :
𝜔1 𝜔2 𝜔3 111 110 101 100 011 010 001 000

𝑆max (𝜔1 𝜔2 𝜔3 ) 32 16 8 8 8 4 4 4
1
𝑆3 (𝜔1 𝜔2 𝜔3 ) 32 8 8 2 8 2 2 2
7
𝑋3 (𝜔1 𝜔2 𝜔3 ) 0 8 0 6 0 2 2 2
Then we use the recursive definitions for 𝑋2 , 𝑋1 , 𝑋0 to find the prices at time 𝑡 = 2, 1, 0:
1
𝑋2 (11) = [𝑞 𝑋 (111) + 𝑞𝑑 𝑋3 (110)]
1+𝑟 𝑢 3
= 45 × [ 12 𝑋3 (111) + 12 𝑋3 (110)] = 16
5 ,
4
𝑋2 (10) = 5 × [ 12 𝑋3 (101) + 12 𝑋3 (100)] = 12
5 ,
4
𝑋2 (01) = 5 × [ 12 𝑋3 (011) + 12 𝑋3 (010)] = 4
5,
4
𝑋2 (00) = 5 × [ 12 𝑋3 (001) + 12 𝑋3 (000)] = 11
5 ,
and
4
𝑋1 (1) = 5 × [ 12 𝑋2 (11) + 12 𝑋2 (10)] = 56
25 ,
4
𝑋1 (0) = 5 × [ 12 𝑋2 (01) + 12 𝑋2 (00)] = 6
5,
4
and finally 𝑋0 = 5 × [ 12 𝑋1 (1) + 12 𝑋1 (0)] = 172
125 .
Barrier Call Options: Let 𝐿, 𝐾 > 0 be two given parameters, and 𝑇 > 0 a given maturity. There are four types
of barrier call options on the underlying asset with strike price 𝐾, barrier 𝐿 and maturity 𝑇 :
1. When 𝐿 > 𝑆0 :
• an up and out call option is defined by the payoff at the maturity 𝑇 :
𝑈 𝑂𝐶𝑇 = (𝑆𝑇 − 𝐾)+ 1{𝑆max ≤𝐿} .
The payoff is that of a European call option if the price process of the underlying asset never reaches the
barrier 𝐿 before maturity. Otherwise it is zero (the contract knocks out).
• an up and in call option is defined by the payoff at the maturity 𝑇 :
𝑈 𝐼𝐶𝑇 = (𝑆𝑇 − 𝐾)+ 1{𝑆max >𝐿}
The payoff is that of a European call option if the price process of the underlying asset crosses the barrier
𝐿 before maturity. Otherwise it is zero (the contract knocks out). We have
𝑈 𝑂𝐶𝑇 + 𝑈 𝐼𝐶𝑇 = 𝐶𝑇 ,
where 𝐶𝑇 is the payoff of the corresponding European call option.
39
2. When 𝐿 < 𝑆0 :
• an down and out call option is defined by the payoff at the maturity 𝑇 :
𝐷𝑂𝐶𝑇 = (𝑆𝑇 − 𝐾)+ 1{𝑆min >𝐿} .
The payoff is that of a European call option if the price process of the underlying asset never reaches the
barrier 𝐿 before maturity. Otherwise it is zero (the contract knocks out).
• an down and in call option is defined by the payoff at the maturity 𝑇 :
𝐷𝐼𝐶𝑇 = (𝑆𝑇 − 𝐾)+ 1{𝑆min ≤𝐿}
The payoff is that of a European call option if the price process of the underlying asset crosses the barrier
𝐿 before maturity. Otherwise it is zero (the contract knocks out). Clearly,
𝐷𝑂𝐶𝑇 + 𝐷𝐼𝐶𝑇 = 𝐶𝑇 ,
where 𝐶𝑇 is the payoff of the corresponding European call option.

Barrier put options: Replace calls by puts in the previous definitions.
An Asian option is a generic name for the class of options whose terminal payoff is based on average asset values
during some period within the options lifetime. Let 𝑇 be the exercise date and 𝑇0 be the beginning date of the
averaging period, for some 0 ≤ 𝑇0 ≤ 𝑇 . Then the payoff at expiry of an Asian call option is given by
𝐶𝑇𝐴 ≡ (𝐴𝑆 (𝑇0 , 𝑇 ) − 𝐾)+ ,
where 𝐴𝑆 (𝑇0 , 𝑇 ) can be the continuous arithmetic average

𝑇
1
𝐴𝑆 (𝑇0 , 𝑇 ) = ∫ 𝑆 𝑑𝑡,
𝑇 − 𝑇 0 𝑇0 𝑡
or a discrete average
1 𝑛−1
𝐴𝑆 (𝑇0 , 𝑇 ) = ∑𝑆 ,
𝑛 𝑖=0 𝑇𝑖
where 𝑛 is a positive integer and 𝑇0 < 𝑇1 < 𝑇2 < ⋯ < 𝑇𝑛 ≤ 𝑇 , or sometimes a geometric average
𝑛−1 1/𝑛
𝐴𝑆 (𝑇0 , 𝑇 ) = ( ∏ 𝑆𝑇𝑖 ) ,
𝑖=0
again for some 𝑇0 < 𝑇1 < 𝑇2 < ⋯ < 𝑇𝑛 ≤ 𝑇 . An Asian put option can be defined similarly.
5.3 American Options

An American call or put option gives the right to buy or, respectively, to sell the underlying asset for the
strike price 𝐾 at any time between now and a specified future time 𝑇 , called the expiry time. In other words,
an American option can be exercised at any time up to and including expiry. The holder of an American type
contingent claim with contract function Φ(𝑥) will receive a payoff Φ(𝑆𝜏 ) at time 𝜏 , where 𝜏 is a random variable
chosen by the holder. The random variable 𝜏 must take values in {0, 1, … , 𝑇 } and specifies the choice of the exercise
time for the holder. This means that if the option will be exercised at time 𝜏 = 𝑡 then the payoff will be Φ(𝑆𝑡 )
at time 𝑡. Of course, it can be exercised only once. The holder does not have complete freedom to choose 𝜏
arbitrarily; it must be a stopping time, i.e., the decision to exercise the option at time 𝑡 can only depend on what
has happened upto time 𝑡 and not on the future randomness. Some examples of stopping times are 𝜏 ≡ 𝑇 (always
exercise at time 𝑇 ), and 𝜏 = inf{𝑡 ∶ 𝑆𝑡 ≥ 𝐿} ∧ 𝑇 (exercise at the first time that the share price is at least price 𝐿,
or at time 𝑇 if that never happens).
It it possible to show that the price of the American option at time 0 equals sup𝜏 {𝔼ℚ [(1 + 𝑟)−𝜏 Φ(𝑆𝜏 )]}, where the
supremum is taken over all stopping times 𝜏 . We give a rough argument, as follows. Suppose the holder exercises
the American option according to the stopping time 𝜏 , so that the payoff to the holder is the amount Φ(𝑆𝜏 ) at time
𝜏 . This is equivalent to a present value of (1 + 𝑟)−𝜏 Φ(𝑆𝜏 ), so risk-neutral valuation tells us that the value at time
40
0 would be 𝔼ℚ [(1 + 𝑟)−𝜏 Φ(𝑆𝜏 )]. But since the holder is free to choose any stopping time 𝜏 , they will choose the 𝜏
that maximises this value at time 0, hence the value must be sup𝜏 {𝔼ℚ [(1 + 𝑟)−𝜏 Φ(𝑆𝜏 )]}.
The following pricing algorithm allows us to compute the value of the American option at any time 𝑡 = 0, 1, … , 𝑇 .
Let 𝑉𝑡𝐴 denote the price of the American option at time 𝑡 (that has not been exercised yet). Using the risk-neutral
valuation formula, we can price an American option inductively, as follows:
At 𝑡 = 𝑇 : 𝑉𝑇𝐴 = Φ(𝑆𝑇 ), because if we hold an American option at time 𝑡 = 𝑇 , the only choice is to exercise or not
at the expiry time 𝑇 , so it has the same value as the European version of the option.
𝐴
At 𝑡 < 𝑇 : suppose we know the value of the American option at time 𝑡 + 1 is 𝑉𝑡+1 , then 𝑉𝑡𝐴 =
1 𝐴
max { Φ(𝑆𝑡 ) , ∣ ℱ𝑡 ] }.
1+𝑟 𝔼ℚ [𝑉𝑡+1
Why do we take a “max” here? It’s because if we hold an American option at time 𝑡, we have the choice to either
exercise early at time 𝑡, or wait. The value of exercising early is Φ(𝑆𝑡 ), the contract function Φ evaluated at the
1 𝐴
current share price 𝑆𝑡 ; the value of waiting at time 𝑡 is the risk-neutral price 1+𝑟 𝔼ℚ [𝑉𝑡+1 ∣ ℱ𝑡 ], and we will choose
whichever gives us more.
Summarising, we have the following pricing algorithm for American options:
Φ(𝑆𝑇 (𝜔1 ⋯ 𝜔𝑇 )) if 𝑡 = 𝑇 ,
𝑉𝑡𝐴 (𝜔1 ⋯ 𝜔𝑡 ) = { 1 𝐴 𝐴
max {Φ(𝑆𝑡 (𝜔1 ⋯ 𝜔𝑡 )), 1+𝑟 [𝑞𝑢 𝑉𝑡+1 (𝜔1 ⋯ 𝜔𝑡 1) + 𝑞𝑑 𝑉𝑡+1 (𝜔1 ⋯ 𝜔𝑡 0)]} if 𝑡 < 𝑇 .
To see the algorithm in more detail, let’s consider an American option expiring after 2 steps with the contract
function Φ(𝑥). The value of this option at time 2 (if it is not exercised before time 2) is clearly Φ(𝑆2 (𝜔1 𝜔2 )). At
time 1 the option holder will have the choice to exercise immediately, with payoff Φ(𝑆1 (𝜔1 )), or to wait until time
2, when the value of the option will become Φ(𝑆2 (𝜔1 𝜔2 )). The value of waiting at time 1 is therefore given by
1
[𝑞 Φ(𝑆2 (𝜔1 1)) + 𝑞𝑑 Φ(𝑆2 (𝜔1 0))].
1+𝑟 𝑢
In effect, the option holder has the choice between the “value of waiting” and the immediate payoff Φ(𝑆1 (𝜔1 )). The
American option at time 1 will, therefore, be worth the higher of these two:
1
𝑉1𝐴 (𝜔1 ) = max{Φ(𝑆1 (𝜔1 )), [𝑞 Φ(𝑆2 (𝜔1 𝐻)) + 𝑞𝑑 Φ(𝑆2 (𝜔1 𝑇 ))]}.
1+𝑟 𝑢
The same reasoning applied at time 0 gives

1
𝑉0𝐴 = max{Φ(𝑆0 ), [𝑞 𝑉 𝐴 (𝐻) + 𝑞𝑑 𝑉1𝐴 (𝑇 )]}.
1+𝑟 𝑢 1
Example 5.2
Consider an American put option with strike price 𝐾 = 80 pounds expiring at time 2 on a stock with initial
price 𝑆0 = 80 pounds in a Binomial model with 𝑢 = 1.1, 𝑑 = 0.95 and 𝑟 = 0.05. The stock values are:
𝑡 0 1 2
96.80
88.00 <
𝑆𝑡 80.00 < 83.60
76.00 <
72.20
The price of the American put will be denoted by 𝑃𝑡𝐴 for 𝑡 = 0, 1, 2 and its price at time 2 is (80 − 𝑆2 )+ given
41
in the following tree:

𝑡 0 1 2
0.00
? <
𝑃𝑡𝐴 ? < 0.00
? <
7.80
First observe that 𝑞𝑢 = 1+𝑟−𝑑 2 1
𝑢−𝑑 = 3 and 𝑞𝑑 = 3 . At time 1 the option holder can choose between exercising the
option immediately or waiting until time 2. In the up state at time 1 the immediate payoff is (𝐾 − 𝑆1 )+ =
1
(80 − 88)+ = 0 and the value of waiting is 1+𝑟 [𝑞𝑢 × 0 + 𝑞𝑑 × 0] = 0. In the down state the immediate payoff is
4 pounds, while the value of waiting is 1.05 × 31 × 7.8 ≈ 2.48. The option holder will choose the higher value
−1
(i.e., to exercise the option in the down state at time 1). This gives the time 1 value of the American put
𝑡 0 1 2
0.00
0.00 <
𝑃𝑡𝐴 ? < 0.00
4.00 <
7.80
At time 0 the choice is, once again, between the payoff (80 − 𝑆0 )+ , which is zero, or the value of waiting,
which is 1.05−1 × 31 × 4 ≈ 1.27 pounds. Taking the higher of the two completes the tree of the option prices:
𝑡 0 1 2
0.00
0.00 <
𝑃𝑡𝐴 1.27 < 0.00
4.00 <
7.80
Therefore the price of the American put is 𝑃0𝐴 = 1.27 pounds.
1
In comparison, the price of a European put is 𝑃0𝐸 = 1.05−1 × 3 × 2.48 ≈ 0.79. Here we use 2.48 (not 4) in the
calculation as European option is exercised at time 2.
What is the price of an American call in the above example? Although in general an American option is at least
as valuable as the equivalent European option (because of the additional choice in when to exercise the option), for
call options (on a stock that does not pay dividends) the American and European options have the same
price.
Theorem 5.2. The prices of American and European call options on a stock that pays no dividends are equal
𝐶 𝐴 = 𝐶 𝐸 , whenever the strike price 𝐾 and expiry time 𝑇 are the same for both options.
Proof. The relation 𝐶 𝐴 ≥ 𝐶 𝐸 is clear as the American call option gives higher payoff (since you can exercise your
right at any time) than the European call. (It’s also possible to give an arbitrage argument to prove this.) Now if
𝐶 𝐴 > 𝐶 𝐸 , then
• write and sell an American call.
• buy a European call.
• invest the difference 𝐶 𝐴 − 𝐶 𝐸 risk free with interest rate 𝑟.
If the American call is exercised at time 𝑡 ≤ 𝑇 , then borrow a share and sell it for 𝐾 to settle your obligation as a
writer of the call option, investing 𝐾 at the rate 𝑟. Then at time 𝑇 you can use the European call to buy a share
for 𝐾 and close your short position in stock. Your arbitrage profit will be
(𝐶 𝐴 − 𝐶 𝐸 )(1 + 𝑟)𝑇 + 𝐾(1 + 𝑟)𝑇 −𝑡 − 𝐾 > 0.
42
If the American option is not exercised at all, you will end up with the European option and an arbitrage profit
(𝐶 𝐴 − 𝐶 𝐸 )(1 + 𝑟)𝑇 . This proves that 𝐶 𝐴 = 𝐶 𝐸 .
43
6 The Black–Scholes formula

We have seen how to price contingent claims using the binomial model in terms of the given parameters 𝑢, 𝑑, 𝑟,
etc. But this is only of practical use of the binomial model gives a reasonable approximation to real stock price
behaviours. In this chapter we look at how we can try to match the binomial model to observed data, as well as
some particular choices of 𝑢, 𝑑, 𝑝𝑢 that give rise to the Black–Scholes formula for pricing a European call option.
Our first step is to move “towards” working in continuous time, by dividing our interval [0, 𝑇 ] into a larger, but
still finite, number of segments. In reality, the time delay between potential trading times is only limited by the
processing speed of our computers, and we usually view this as a continuous process. Nonetheless, we will consider
a market in which trades can happen at times 𝑡 ∈ {0, 𝑇𝑛 , 2𝑇 𝑇
𝑛 , … , 𝑇 }. We set ℎ = 𝑛 as the time difference between
each pair of consecutive trading times.
6.1 Non-binomial prices

One approach we can take (following that in Stochastic Finance, by Turner and Zeindler) is to consider a more
general form for the way our share prices evolve. It seems sensible to focus on the relative change in the share
prices at each time step, rather than the absolute change, on the basis that a change of £1 is more significant if the
current share price is £0.5 than if it is £500.
Recall that in Chapter 3, we defined the share prices via
𝑆𝑡 (𝜔) = 𝑍𝑡 (𝜔)𝑆𝑡−1 (𝜔),
where 𝑍𝑡 is a random variable which takes value 𝑢 if 𝜔𝑡 = 1, or 𝑑 if 𝜔𝑡 = 0.

Let’s consider what will happen if we replace 𝑍𝑡 (𝜔) with a random variable with a different distribution, say 𝜉𝑡 .
𝑘𝑇
At time 𝑡 = 𝑛 , we have
𝑘
𝑆𝑘 = 𝜉𝑘 𝑆𝑘−1 = 𝑆0 ∏ 𝜉𝑗 .
𝑗=1
Taking logs of both sides, we can write this as

𝑘
log 𝑆𝑘 − log 𝑆0 = ∑ log 𝜉𝑗 .
𝑗=1
Now we have a sum of random variables, which we can reasonably assume to be independent from each other (if
we suppose that the market forces affecting share prices at one time are independent of those at another time), and
identically distributed (if we suppose that these market forces are not meaningfully different from one moment to
another). By the Central Limit Theorem, this sum is approximately Normally distributed, with mean 𝑘𝔼[log 𝜉1 ]
and variance 𝑘Var(log 𝜉1 ).
The approximation is exact when the log 𝜉𝑗 s are Normally distributed. Deciding to model the log 𝜉𝑗 s as a sequence
of i.i.d Normal random variables is a reasonable choice: the market forces consist of many small components,
contributing to one “large-scale” change in the price. If we take 𝜉𝑗 = 𝑒𝐼𝑗 , where 𝐼𝑗 ∼ 𝒩(𝑎, 𝑏2 ), then
log 𝑆𝑇 − log 𝑆0 ∼ 𝒩(𝑛𝑎, 𝑛𝑏2 ).
How do we decide what values of 𝑎 and 𝑏 make sense? We work in terms of the drift, usually denoted 𝜇, and the
volatility, which we denote by 𝜎2 . These are properties of the share price which we can observe. The drift rate
tells us about the direction in which the expected share price is moving, and how quickly, and the volatility tells us
about how much the share price fluctuates around these values. Mathematically, we have
𝔼[𝑆𝑡 ] = 𝑒𝜇𝑡 𝕍ar(log 𝑆𝑡 ) = 𝜎2 𝑡.
Using (6.1), we should set

𝑛𝑏2 = 𝜎2 𝑇 ,
44
or in other words, 𝑏2 = 𝜎2 𝑇𝑛 = 𝜎2 ℎ. Then, recalling the moment generating function of a Normal distribution, we
can write
1
𝔼[𝜉𝑗 ] = 𝔼[𝑒𝐼𝑗 ] = exp {𝑎 + 𝑏2 } ,
2
to show that the correct choice for 𝑎 is
1 𝑇 1
𝑎 = (𝜇 − 𝜎2 ) = (𝜇 − 𝜎2 )ℎ.
2 𝑛 2
Conclusion: When we model the relative change in the share price at each time step via
𝜉𝑗 = 𝑒𝐼𝑗 ,
with
1
𝐼𝑗 ∼ 𝒩 ((𝜇 − )ℎ, 𝜎2 ℎ) ,
𝜎2
then the share prices are in the form
𝑆 𝑡 = 𝑆 0 ∏ 𝑒 𝐼𝑗 ,
and have the correct drift rate and volatility.
In a bit of foreshadowing for next term, we can rearrange this formula a bit. Set
1
𝐼𝑗 = (𝜇 − 𝜎2 ) ℎ + 𝜎𝑁𝑗 ,
2
where 𝑁𝑗 is a Normally-distributed random variable with mean 0 and variance ℎ. (Convince yourself that this still
gives us the same expectation and variance as before.) Then, if we write
(𝑛)
𝑊𝑡 = ∑ 𝑁𝑗 ,
𝑗≤𝑛𝑡
we can write
1 2 (𝑛)
𝑆𝑡 = 𝑆0 𝑒(𝜇− 2 𝜎 )𝑡
exp{𝜎𝑊𝑡 }.
This useful version of the formula allows us to pull out all of the deterministic parts of 𝑆𝑡 (the first two terms), and
deal with the random part (the final term) in its own right.
(𝑛)
As 𝑛 → ∞, the distribution of exp{𝜎𝑊𝑡 } converges to that of the stochastic process geometric Brownian
motion, evaluated at time 𝑡. To prove that this convergence holds at a single point is fairly straightforward, but
(𝑛)
proving that as a function, 𝑡 ↦ exp{𝜎𝑊𝑡 } converges in distribution to 𝑡 ↦ 𝐺𝐵𝑀𝑡 is much more complicated.
45
Terminology: exotic options

In Finance, “Exotic” options refer to a class of options that have more complicated payoff functions than the
“standard” (often called “vanilla”) European and American options.
Example:
Common types include chooser options, compound options, barrier options, digital options, Bermuda options,
quantity-adjusting options, look-back options, Asian options, basket options, extendible options, spread options,
shout options, range options etc.
Origins:
The terminology “Exotic” options became popular in the 1990s in the field due to Berkeley professor Mark Rubin-
stein’s working paper, Exotic Options. According to Rubinstein, one possible explanation for the term “Exotic” is
that in 1987, executives at the Tokyo oﬀice of an American bank sold an option that was unlike the European and
American versions, which is now known as the Asian option. Like the terms “European” and “American” options,
Rubinstein decided to give a name based on geographic origins.
Another possible explanation for the term comes from exotic wagers in horse racing. In the 1970s and early 1980s,
Rubinstein was researching a racetrack-betting tactic that guaranteed profits and the term “Exotic” may come from
that earlier work.
Decolonising the curriculum:

The terminology “Exotic” has shown the impact of colonisation on learning and teaching. We seek to both recognise
and address the terminologies that have arisen from historic global domination by ‘The West’. The term “Exotic”
emphasises the fact that Western countries are positioned at the centre and dominate the content taught in high-
level education, this may potentially impact the social inequality in mathematics academia and community. We
would like to address and raise awareness of the West-centric biases that have been previously ignored in the field.
In this case, the term “Exotic” is removed from the course and the lecture notes, nevertheless you may still find the
use of terminology due to its popularity in textbooks, websites and beyond.
46
Background: Islamic Finance

Mathematical finance, which is a study of mathematical concepts applied to financial markets, can lead to monocul-
tural and mainly West-centric perspectives. As part of our Decolonising the Curriculum project, we aim to design
and effectively deliver a more inclusive globalised curriculum. Therefore, we would like to mention the following
non-Western model finance, namely Islamic finance, to open our field to greater diversity and wider analysis.
Introduction
Islamic finance (or Sharia-compliant finance) refers to financial activity that complies with Islamic law, often known
as Sharia. In other words, it is a way to manage money based on some of the moral principles of Islam, and sometimes
imposes prohibitions on activities in its own right, even if they are not illegal in the countries where Islamic finance
is operating. Islamic finance can be thought of as a unique form of socially responsible investment.
Characteristics
• Money concept: Islamic finance is based on the belief that money should not have any value itself, rather it
is a way to exchange products and services that do have a value.
• Riba: One of the distinct features of Islamic finance is the prohibition of riba (riba can be roughly translated
as “usury”, unjust or exploitative gains made in trade or business. However, some Muslims dispute whether
there is a consensus that interest is equivalent to riba.).
• Businesses involve prohibited activities: Investing in activities that are considered to be haram (i.e., forbidden
or proscribed by Islamic law), such as selling alcohol, pork products, weapons etc, is forbidden.
• Maisir and Gharar: Islamic law strictly prohibits any form of speculation or gambling, (referred to as maisir),
and participation in contracts with excessive risk and uncertainty (referred to as gharar).
• Partnership: Islamic finance encourages partnership. More precisely, profit and risks should be shared between
at least two parties where possible. This can be between two individuals, an individual and a business, or two
businesses.
Importance
Islamic laws impose boundaries, limitations and restrictions on finance, which may conflict with the non-Islamic
financial system. In order to not exclude individuals due to their cultures and religious beliefs, an Islamic financial
system provides accepted financial practices of Islam.
The common practices of Islamic finance arise along with the foundation of Islam. As a centuries-old practice, it is
gaining recognition in countries that do not have a predominantly Muslim population such as the UK, Hong Kong,
China etc. Today, Islamic finance is a $2.5 trillion industry spread over more than 80 countries. Over the past
decade, Islamic finance grew at an exponential annual pace of 10%-12% and is predicted to grow to $3.5 trillion by
2024. (For more see Islamic Finance: Just For Muslim-Majority Nations?)
47

Main

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Main

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

Main

Uploaded by

Copyright:

Available Formats

Mathematical Finance Lecture Notes 2024-25

2 The one-period binomial model 12

3 The multi-period binomial model: Part 1 19

4 Filtrations, Conditional Expectation, and Martingales 30

5 Pricing options from around the world 37

6 The Black–Scholes formula 44

1.1 What is Mathematical Finance?

1.1.1 Finance or probability?

1.2 Financial background

1.2.1 Underlying and derivative assets

For some more information…

Next, the present value of all the withdrawals is

These are equal when

50𝑥 − (100𝑥 − 3𝐶𝑥) = (3𝐶 − 50)𝑥.

A basic underlying principle of derivative pricing is:

By applying 1.1 in both directions, we obtain the Law of One Price:

or else an arbitrage opportunity exists.

𝑉𝑇𝑋 = (𝑆𝑇 − 𝐾)+ + 𝐾(1 + 𝑟)−𝑇 × 𝐵𝑇 = (𝑆𝑇 − 𝐾)+ + 𝐾 = max(𝑆𝑇 , 𝐾).

1.3 Discrete-time model

4. A risky asset, whose price dynamics are defined for each 𝜔 ∈ Ω by

𝑆 ∶ 𝜔 ↦ {𝑆1 (𝜔), … , 𝑆𝑇 (𝜔)}.

The value of 𝑋 is always 𝑉𝑡 = −100𝑆𝑡 .

𝑉𝑡 − 𝑉𝑡− = 𝑥𝑡 𝐵𝑡 + 𝑦𝑡 𝑆𝑡 − (𝑥𝑡− 𝐵𝑡− + 𝑦𝑡− 𝑆𝑡− )

The self-financing condition means that this value should always be 0.

2 The one-period binomial model

2.1 The model

2.2 Portfolios and arbitrage

Value and arbitrage

𝑑 < 1 + 𝑟 < 𝑢. (2.1)

we must have 𝑥 = −𝑠𝑦.

2.3 Contingent claims

In other words, we want to find a solution to the system

(1 + 𝑟)𝑥 + 𝑠𝑢𝑦 = Φ(𝑠𝑢)(1 + 𝑟)𝑥 + 𝑠𝑑𝑦 = Φ(𝑠𝑑).

If 𝑢 > 𝑑, this system has a unique solution, namely,

Question: what is a fair price for a contingent claim 𝑋?

Proof. (This is an application of the Law of One Price, Theorem 1.1.)

(1 + 𝑟)𝑥 + 𝑠𝑢𝑦 = (𝑠𝑢 − 9)+ , (1 + 𝑟)𝑥 + 𝑠𝑑𝑦 = (𝑠𝑑 − 9)+ ;

2.4 The martingale measure

our martingale condition becomes

𝑢𝑞𝑢 + 𝑑𝑞𝑑 = 1 + 𝑟 (2.4)

Surprisingly, we have already met such conditions in Equation (2.1).

2.5 Risk neutral valuation

The price at time 𝑡 = 0 of this portfolio is given by

2.6 Some generalisations

What happens when 𝑢 = 𝑑

Many assets, and/or many outcomes, at time 𝑇

3 The multi-period binomial model: Part 1

t=0 t=1 t=2

• At 𝑡 = 1, when 𝑆1 = 120, we have

3.2 Portfolios and the self-financing condition

at time 0, (𝑥1 , 𝑦1 ) = (−200, 20)

The value of a portfolio at time 𝑡 = 0, … , 𝑇 is given by

𝑥𝑡 𝐵𝑡 + 𝑦𝑡 𝑆𝑡 = 𝑥𝑡+1 𝐵𝑡 + 𝑦𝑡+1 𝑆𝑡 . (3.1)

3.3 The martingale measure

At 𝑡 = 1 and 𝑆1 = 120, we have

Next, for 𝐴 ∈ ℱ,we define ℚ(𝐴) using the elements 𝜔 ∈ 𝐴:

3.4 Pricing European contingent claims

su Πut+1 price of claim when share price s → su

sd Πdt+1 price of claim when share price s → sd