Part 1

Fundamentals of Quantitative Finance

Sections: 2.2, 4.3, 5.1–5.4, 7.1–7.5
Print version of the lecture in MA470/670 Fin. Math: Pricing in Cont. Time presented on
January 9–13, 2023

by Roman Makarov from Department of Mathematics at Wilfrid Laurier University

1.1

Contents
1 Basic Concepts 1

2 Binomial Tree Model (Cox–Ross–Rubinstein) 7

3 Log-normal Model (Black–Scholes–Merton) 12 1.2

Objectives

• Review basic concepts of Financial Mathematics

• Review the binomial tree model (aka the Cox–Ross–Rubinstein model)
• Introduce the lognormal asset pricing model (aka the Black–Scholes–
Merton model)
1.3

1 Basic Concepts
Basic Concepts: Outcomes
Review the following fundamental concepts:
• Filtered probability space
• Security, portfolio, trading strategy
• Replication, risk-neutral pricing
• Completeness, super- and sub-hedging
• Fundamental Theorems of Asset Pricing (FTAP 1 and 2)
1.4

Probability Space: Single-Period Case

The probability space is a triple (Ω, F, P) that provides a mathematical mode of a random
experiment.
• Sample space Ω contains all possible outcomes (e.g., market scenarios).
For example, the finite sample space is Ω = {ω 1 , ω 2 , . . . , ω M }.
• Event space F is a sigma-algebra over Ω.

1
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 2

For example, the power set 2Ω , the trivial set {∅, Ω}, and the set {∅, A, A{ , Ω} for A ⊂ Ω
are σ-algebras.
• Probability function P : F → [0, 1] measures the likelihood of events.
P satisfies Kolmogorov axioms.
1.5

Probability Space: Multi-Period Case

The filtered probability space is a quadruple (Ω, F, P, F).
• The filtration F = {Ft }0≤t≤T is a sequence of σ-algebras s.t. Ft ⊆ Fs for 0 ≤ t ≤ s.
• The filtration models the information flow. Ft represents the information available at
time t
• For every t ≥ 0, Ft is a sub σ-algebra of F, that is Ft ⊆ F.
• The time can be discrete t ∈ {0, 1, 2, . . . , T } or continuous t ∈ [0, T ].
1.6

Example: Binomial Tree Probability Space

• The time is discrete: t ∈ {0, 1, 2, . . . , T }
• There are 2T possible market scenarios:

Ω = ω1 ω2 · · · ωT : ωt ∈ {D, U} .

• The filtration is F = {Ft }0≤t≤T with Ft ⊂ Ft+1 , where F0 = {∅, Ω} and Ft is generated by
the first t market moves ω1 , . . . , ωt .
• The event space is the power set F ≡ FT = 2Ω .
1.7

Probability Space: Review Questions

1. Review the Kolmogorov axioms for a probability function.
2. For a finite sample space Ω, define the probability function
X
P(E) = P(ω), E ⊂ Ω.
ω∈E

Show that it satisfies the Kolmogorov axioms

3. Review the definition of a σ-algebra.
4. Show that for any A ⊂ Ω the set {∅, A, A{ , Ω} is a σ-algebra.
5. Let Ω be a finite sample space with M elements. Show that the σ-algebra F = 2Ω has
M
22 elements.
1.8

Payoffs
• Consider a financial market described by the probability space (Ω, F, P, F).
• Payoff is a random variable. That is, it is an F-measurable function X : Ω → R.
• Payoffs form a vector space denoted by L(Ω).
That is, we can add payoffs up and multiply them by a scalar.
• Claim is a nonnegative payoff
• A risk-free payoff is constant on Ω
1.9
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 3

Securities/Assets
• Security is a nonnegative stochastic process {St }0≤t≤T adapted to the filtration F.
That is, St is an Ft -measurable random variable.
• S0 ≥ 0 is the initial price; ST ≥ 0 is the terminal payoff.
• There are risky and risk-free securities.
• Now, let us see how base assets can be used to create static portfolios and dynamic
trading strategies in the binomial tree model.
1.10

Portfolios
Base Assets
In the binomial tree model, we have a risky stock S and a risk-free bank account B with
prices
Bt = (1 + r)t , St (ω) = S0 u Ut (ω) d Dt (ω) .

Static Portfolio
• (Static) portfolio is a vector (β, ∆) ∈ R2 .
(β,∆)
• Portfolio value at time t is Πt = β Bt + ∆ St .
1.11

Portoflio Strategy and its value

• Portfolio strategy is a stochastic process, {(βt , ∆t )}t≥0 . adapted to the filtration.
That is, both βt and ∆t are Ft -measurable.
(βt−1 ,∆t−1 )
• The liquidation value at time t is Πt = βt−1 Bt + ∆t−1 St
(β ,∆ )
• The acquisition value at time t is Πt t t = βt Bt + ∆t St
(β0 ,∆0 )
• The initial value at time 0 is the constant Π0 = β0 B0 + ∆0 S0
(β ,∆ )
• The terminal value at maturity T is ΠT T −1 T −1 = βT −1 BT + ∆T −1 ST
1.12

Self-Financing Strategies
A self-financing strategy does not allow for injecting or withdrawing funds. That is, its
acquisition value is the same as the liquidation value for every time.
Self-Financing Condition
(βt ,∆t ) (βt−1 ,∆t−1 )
Πt = Πt , or, equivalently,
δΠt = βt δBt + ∆t δSt , t = 0, 1, . . . , T − 1
where δΠt = Πt+1 − Πt , δSt = St+1 − St , and δBt = Bt+1 − Bt
1.13

Self-Financing Strategies
• In the binomial tree model, a self-financing strategy is defined by the ∆-process,
{∆t }t≥0 , and the initial wealth Π0 .
Πt − ∆t St
• The position in the bank account, βt , is given by βt = .
Bt
• The time-t value is calculated iteratively using the wealth equation

Wealth Equation
Πt+1 = ∆t St+1 + (Πt − ∆t St )(1 + r), t = 0, 1, . . . , T − 1
1.14
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 4

Portfolios and Strategies in a multi-stock model

• The value of portfolio with N risky stocks, [β, ∆1 , . . . , ∆N ] ∈ RN +1 , is
N
(β,∆)
X
Πt = β Bt + ∆i Sti
i=1

• A trading strategy [βt , ∆1t , . . . , ∆N

t ], t = 0, 1, . . . , T − 1 is self-financing if

δΠt = βt δBt + ∆1t δSt1 + · · · + ∆N N

t δSt

holds.
1.15

E XAMPLE 1.1
Prove the self-financing condition δΠt = βt δBt + ∆t δSt .

Solution.

1.16

E XAMPLE 1.2
Find the cumulative gain ΠT − Π0 of a self-financing strategy {(βt , ∆t )}0≤t≤T −1 .

Solution.

1.17

Strategies: Review Questions

1. Review the definition of a vector space. Argue that L(Ω) is a vector space.
2. Describe the difference between risky and risk-free securities.
3. Derive the wealth equation
4. Prove the self-financing condition for a trading strategy with N risky stocks
1.18
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 5

Replication: Single-Period Model

• For a given payoff X ∈ L(Ω), find a portfolio of basis assets, ϕ, s.t. Πϕ
T (ω) = X(ω) for all
scenarios ω ∈ Ω.
• Any such portfolio ϕ is called a hedge or a replicating portfolio for the payoff X.
• A payoff X is said to be attainable if there exists a replicating portfolio ϕ.
• If every payoff is attainable, the market model is said to be complete.
1.19

Replication: Multi-Period Model

• For a given payoff X ∈ L(Ω), find a s.-f. portfolio strategy in basis assets, {ϕt }t≥0 s.t.
ϕ
ΠT T −1 (ω) = X(ω) for all scenarios ω ∈ Ω.
ϕ
Here, ΠT T −1 is the terminal value of the strategy.
• Such a portfolio strategy is said to replicate the payoff X.
1.20

Super-Replication
• Consider some payoff X ∈ L(Ω)
• A super-replicating portfolio ϕ has the terminal value Πϕ T ≥ X.
Such a portfolio covers the liabilities of the writer of the security with claim X.
• There exists a super-replicating portfolio with minimal initial cost Πϕ
0 . It solves the
following linear programming problem:

Πϕ ϕ
0 → min subject to ΠT ≥ X.
ϕ

Denote the solution by ϕu and call it the super-hedge.

1.21

Sub-Replication
• Consider some payoff X ∈ L(Ω)
• A sub-replicating portfolio ϕ has the terminal value Πϕ
T ≤ X.
• There exists a sub-replicating portfolio with maximal initial cost Πϕ
0.

• It solves the following linear programming problem:

Πϕ ϕ
0 → max subject to ΠT ≤ X.
ϕ

Denote the solution by ϕd and call it the sub-hedge.

d u
• Clearly, we should have Πϕ ϕ
0 ≤ Π0 .
1.22

Arbitrage
• The Law of one price holds if

∀ϕ, ψ Πϕ ψ ϕ ψ
T = ΠT =⇒ Π0 = Π0 .

• An arbitrage is a static portfolio ϕ (or an admissible self-financing portfolio strategy

{ϕ}0≤t≤T ) such that

Πϕ
0 = 0, P(Πϕ ϕ
T ≥ 0) = 1, and P(ΠT > 0) > 0.

• If the market is arbitrage free, the Law of One Price holds.

1.23
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 6

Pricing by Replication
• Consider some claim with the terminal payoff VT
• If VT is attainable, then there exist a portfolio strategy ϕ that replicates VT
• The fair initial price of the claim is then V0 = Πϕ0.

• It is a no-arbitrage price, meaning that there is an arbitrage if the actual price differs
from V0 = Πϕ 0.

• This method of finding V0 is called the pricing by replication.

1.24

Numeraire & FTAP 1

• A numeraire asset is any trading
n asset
o g with a strictly positive price process. Dis-
St
counted asset price process S̄t ≡ gt .
0≤t≤T

• An equivalent martingale measure (EMM) relative to numeraire g, denoted P e (g) , is a

probability measure defined s.t. (1) it is equivalent to the real-world measure P and (2)
every discounted base-asset price process is a martingale.
e (g) -
• For every self-financing portfolio strategy, the value process discounted by g is a P
martingale.
• FTAP 1: The market model is arbitrage free iff there exists an EMM.
1.25

Risk-Neutral Pricing
• Assuming the EMM P e (relative to some numeraire g) exists, the discounted portfolio
value process {Π̄t ≡ Πt /gt }t≥0 is a martingale.
• Therefore,
     
Π0 e 0 ΠT =⇒ V0 = E e 0 g0 VT
e 0 VT =⇒ V0 = E
=E
g0 gT g0 gT gT

• Particularly, if gt = Bt = B0 (1 + r)t , we have

1
V0 = E
e 0 [VT ] .
(1 + r)T
1.26

No-Arbitrage Pricing in Incomplete Market

Super-Hedging
• The super-hedging portfolio ϕu is a solution to Πϕ ϕ
0 → minϕ subject to ΠT ≥ VT .
u
• The price π0u = Πϕ0 is the infimum of arbitrage prices for the seller. That is, any price
V0 > π0u is an arbitrage price.

Sub-Hedging
• The sub-hedging portfolio ϕd is a solution to Πϕ ϕ
0 → maxϕ subject to ΠT ≤ VT .
d
• The price π0d = Πϕ0 is the supremum of arbitrage prices for the buyer. That is, any price
V0 < π0d is an arbitrage price.
1.27
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 7

Bid-Ask Spread
• The interval [π0d , π0u ] contains all no-arbitrage initial prices of the claim with payoff VT .
• We call it the bid-ask spread.
• The ask price π0u and the bid price π0d can also be calculated by respectively taking the
supremum and the infimum of the discounted expectation of the payoff function over
the set of possible risk-neutral measures:
   
u (g) g0 d (g) g0
π0 = sup E e VT π0 = inf E e VT
e (g)
P
gT e (g)
P gT
• FTAP 2: The EMM is unique iff the market model is complete. 1.28

Pricing: Review Questions

d u
1. Under what condition we have Πϕ ϕ
0 ≤ Π0 ?
d u
2. Under what condition we have Πϕ ϕ
0 = Π0 ?
3. Give other variations of the Law of One Price.
4. Why the fair price V0 is the no-arbitrage price?
5. What does it mean when two probability measures are equivalent. Give examples of
numeraires.
6. If the market is arbitrage-free and complete, and hence the EMM exists and is unique,
what can you say about π0d and π0u ? 1.29

2 Binomial Tree Model (Cox–Ross–Rubinstein)

Binomial Tree Model: Outcomes
• Review the construction of the binomial tree model.
• Review the backward iterative process for finding no-arbitrage prices and a replication
strategy of a European-style derivative in the binomial model.
• Derive the Cox–Ross–Rubinstein (CRR) pricing formulae for standard European op-
tions.
• Review the parametrization of the CRR model and the asymptotic behaviour of stock
prices under P and P.e
• Derive the lognormal model as a limiting case of the binomial tree model under two
probability distributions. 1.30

The Binomial Tree Model: Sample Space

• The state space for the model with T period is

Ω = ω1 ω2 . . . ωT : ωk ∈ {D, U}
• Denote
Ut (ω) is the number of U’s in ω1 , ω2 , . . . , ωt
Dt (ω) = t − Ut (ω) is the number of D’s in ω1 , ω2 , . . . , ωt
• The probability measure P is given by
X
∀E ⊆ Ω P(E) = P(ω) where P(ω) = p UT (ω) (1 − p) DT (ω)
ω∈E

• Note the probability distributions of Ut and Dt :

Ut ∼ Bin(t, p), Dt ∼ Bin(t, 1 − p)
1.31
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 8

The Binomial Tree Model: Base Assets

• There are two basis assets: a risk-free bank account B and a risky stock S with values
(
u St with prob. p
Bt = (1 + r)t , St+1 = , t = 0, 1, 2, . . .
d St with prob. 1 − p

where 0 < d < u and p ∈ (0, 1).

• The dynamics of stock prices is described by a recombining binomial tree.
• The probability distribution of St with t = 1, 2, . . . , T is
 
t k
St = S0 uk d t−k with probability p (1 − p)t−k , k = 0, 1, 2, . . . , t
k

• Note the connection between St and Ut :

St (ω) = S0 u Ut (ω) d t−Ut (ω)

1.32

The Binomial Tree Model: Conditioning

• The unconditional probability of event E ⊆ Ω is
X
P(E) = P(ω) where P(ω) = p UT (ω) (1 − p) DT (ω)
ω∈E

• The probability of event E ⊆ Ω conditional on Ft with 0 ≤ t ≤ T is

X
Pt (E) ≡ P(E | Ft ) = Pt (ω) where Pt (ω) = p Ut (ω) (1 − p) Dt (ω)
ω∈E

• The corresponding conditional expectation of a random variable X, Et [X] ≡ E[X | Ft ],

is a function of ω1 , . . . , ωt s.t.

E[X1{Aω1 ,...,ωt } ]
Et [X](ω1 , . . . , ωt ) = .
P(Aω1 ,...,ωt )

• Usually, we use properties of conditional expectations. In particular, for any n and t

with 0 ≤ t ≤ n ≤ T ,
   
Sn Sn
Et [Sn ] = Et St = St E = St (pu + (1 − p)d)n−t .
St St
1.33

No-Arbitrage and Completeness

• Under the risk-neutral probability measure (aka EMM) P, e the discounted stock price
process, S̄t = St /Bt , t ≥ 0 is a martingale. That is, Et [S̄t+1 ] = S̄t . We have
e

e t [St+1 ] = (1 + r)St ⇐⇒ p̃ · u + (1 − p̃) · d = 1 + r ⇐⇒ p̃ = 1 + r − d

E
u−d

• There exists a unique risk-neutral probability measure Pe iff d < 1 + r < u. Therefore,
the binomial tree model is arbitrage free and complete.
1.34
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 9

Price Process
• Consider a derivative with FT -measurable payoff VT , at maturity T .
• Define the derivative price process {Vt }0≤t≤T by

1
Vt (ω1 ω2 . . . ωt ) = (p̃Vt+1 (ω1 ω2 . . . ωt U) + (1 − p̃)Vt+1 (ω1 ω2 . . . ωt D))
1+r
1 e
≡ Et [Vt+1 ](ω1 ω2 . . . ωt ), t = 0, 1, . . . , T − 1,
1+r
where p̃ and 1 − p̃ are risk-neutral probabilities as given above.
• At maturity time T , the payoff VT is given. At any time t < T , the price Vt is calculated
using a backward recursion for every scenario ω.
1.35

Replicating Portfolio Strategy

• Consider the self-financing portfolio strategy {∆t }0≤t≤T −1 given by

Vt+1 (ω1 ω2 . . . ωt U) − Vt+1 (ω1 ω2 . . . ωt D)

∆t (ω1 ω2 . . . ωt ) = , t = 0, 1, . . . , T − 1.
St+1 (ω1 ω2 . . . ωt U) − St+1 (ω1 ω2 . . . ωt D)

• Set Π0 = V0 and construct recursively forward in time the value process {Πt }0≤t≤T via
the wealth equation.
• Then, the strategy replicates the derivative price process at every time, i.e. Πt (ω) =
Vt (ω) holds for all t = 0, 1, . . . , T and all scenarios ω ∈ Ω.
1.36

E XAMPLE 1.3
Find prices of the at-the-money European put option and the replicating strategy in the
binomial tree model with S0 = 10, u = 1.25, d = 0.8, r = 0.1, T = 3.

Solution.

1.37
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 10

Pricing European-style Derivatives

• Consider a European-style derivative with payoff VT = Λ(ST ).
• The initial no-arbitrage price is given by

V0 = (1 + r)−T E[V
e T ] = (1 + r)−T E[Λ(S
e T )]

• Using the fact that ST = S0 uUT dT −UT where UT ∼ Bin(T, p̃) (under P),
e we have

V0 = (1 + r)−T E
e Λ(S0 uUT dT −UT )
 

T
X
= (1 + r)−T Λ(S0 un dT −n )P(U
e T = n)
n=0
T  
X T n
= (1 + r)−T Λ(S0 un dT −n ) p̃ (1 − p̃)T −n
n=0
n
1.38
E XAMPLE 1.4
Derive the CRR Pricing Formula for a European call

Solution.

1.39
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 11

Parametrization of the CRR Model

• Fix the time horizon T > 0; the annual volatility of the stock, σ 2 ; the annual rate of
log-return on the stock, ν; and risk-free interest rate r:
     
S(T ) S(T ) B(T )
Var ln = σ 2 T, E ln = νT, = erT
S(0) S(0) B(0
• For a given N ≥ 1, consider an N -period binomial tree model where trading is admitted
T
at time k · δN , k = 0, 1, 2, . . . , N ; δN = N is the length of one period.
• The CRR model parameters:
1 1νp √ √
rN = eδN r − 1, pN = + δN , uN = eσ δN
, dN = e−σ δN
2 2σ
1.40

Sequence of CRR Models

• For each N = 1, 2, . . . we construct an N -period binomial tree model
(N )
• The bank account is Bt = (1 + rN )t = eδN r t for t = 0, 1, 2, . . . , N
(N )
• The stock prices are St ∈ {S0 ukN dNt−k | k = 0, 1, . . . , t} for t = 0, 1, 2, . . . , N
(N ) 2Ut −t
• Since uN dN = 1, we can write St = S0 uN where Ut ∼ Bin(t, pN ).
1.41

Asymptotic Properties under P

 (N )

SN
• The aggregate log-return ln S0 has the following return and variance:
" !#
(N )
SN
E ln = N δN ν → νT,
S0
" !#
(N )
SN
Var ln = N δN σ 2 + O(1/N ) → σ 2 T.
S0

• As N → ∞, binomial prices converge to log-normal prices thanks to the Central Limit

Theorem:
!
(N ) √ √
SN d (N ) d
ln → νT + σ T Z =⇒ SN → S0 eνT +σ T Z , Z ∼ Norm(0, 1)
S0

Here, we consider the real-world probability distribution P.

1.42

Asymptotic Properties under P

e
• As N → ∞:
1 + rN − dN 1
p̃N = →
u − dN 2
" N !#
(N )
e ln SN σ2
E → (r − )T
S0 2
" !#
(N )
S
Var
g ln N
→ σ2 T
S0

• As N → ∞, binomial prices converges to log-normal prices thanks to the Central Limit

Theorem: √
(N ) d σ2
SN → S0 e(r− 2 )T +σ T Z , Z ∼ Norm(0, 1)
Here, we consider the risk-neutral probability distribution P.
e
1.43
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 12

CRR Model: Review Questions

1. Prove that Ut ∼ Bin(t, p) and Dt ∼ Bin(t, 1 − p). Find Corr(Ut , Dt ).
2. Give examples of numeraires in the binomial tree model.
3. If the market is arbitrage-free and complete, and hence the EMM exists and is unique,
what can you say about π0d and π0u ?
4. Derive the CRR pricing formula for a European put.
d
5. Explain the convergence in distribution: Xn → X as N → ∞.
(N )
6. Find the limiting distribution of SN under the EMM P
b with the stock as a numeraire.
1.44

3 Log-normal Model (Black–Scholes–Merton)

Log-normal Model: Outcomes
1. Obtain the general derivative pricing formula for the log-normal model
2. Derive the Black–Scholes pricing formula by computing the expectation of the payoff
3. Derive the Black–Scholes pricing formula by taking the limit in the CRR formula
4. Derive the Black–Scholes PDE using (a) the time limit; (b) Greeks
1.45

Limit of the CRR Pricing Formulae

2 √
d
e S (N ) → σ
Fact 1 Under P, N S0 e(r− 2 )T +σ T Z , as N → ∞, where Z ∼ Norm(0, 1)
Fact 2 Consider a sequence of random variables {XN }N ≥1 defined on the same probability
d
space and a bounded continuous function f . If XN → X, as N → ∞, then E[f (XN )] →
E[f (X)], as N → ∞.
• Let us apply these two facts to the pricing formula under the binomial tree model.
• Consider a European derivative with a bounded payoff Λ.
• In the CRR model, the initial no-arbitrage price of the derivative is given by the dis-
counted risk-neutral expectation of the payoff function:
   
(N ) (N )  (N ) 
V0 = (1 + rN )−N EPN Λ SN = e−rT EPN Λ SN
e  e 
.
1.46

Limit of the CRR Pricing Formulae

• Consider the limiting value of the no-arbitrage price
(N ) (N )
V0 = e−rT EPN [Λ(SN )]
e

√
(N ) d 2
• As N → ∞, EPN [Λ(SN )] → EP [Λ(S(T ))], where S(T ) = S0 e(r−σ /2)T +σ T Z under P
e e e
• Therefore, the initial no-arbitrage value of the European derivative under the log-
normal model is
V (0, S0 ) = e−rT E
e 0,S [Λ(S(T ))]
0
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 13

1.47

E XAMPLE 1.5( A )
Derive the Black–Scholes Pricing Formula for a European call by computing the mathemat-
ical expectation
   
S
−r τ S

C(t, S; T, K) = S N d+ ,τ −e K N d− ,τ
K K
ln x+(r± 12 σ 2 )τ
where d± (x, τ ) = √
σ τ
and τ = T − t.

Solution.

1.48
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 14

E XAMPLE 1.5( B )
Derive the Black–Scholes Pricing Formula for a European call by taking the limit in the
CRR pricing formula. Use the facts that that p̃N → 12 , pbN → 21 ,

m − N p̃N mN − N pbN
p N → −d− (S0 /K, T ), and p → −d+ (S0 /K, T ), as N → ∞.
N p̃N (1 − p̃N ) N pbN (1 − pbN )

Solution.

1.49

Derive the Black–Scholes PDE (using the time limit) – Step 1

• On a small time interval [t − δ, t] with δ = n1 , where n is the number of periods per year,
the continuous-time √
model can be√approximated by a single-period binomial model
with factors un = eσ δ
and dn = e−σ δ , and rate rn = erδ − 1.
• The single-period approximation of the option price (on the interval [t − δ, t]) is
e (n)
V (t − δ, S) = e−rδ EP [V (t, St ) | St−δ = S].

• In the binomial case, it can be written as follows:

√
√

erδ V (t − δ, S) = p̃n V t, Seσ δ + (1 − p̃n ) V t, Se−σ δ ,
1+rn −dn
where p̃n = un −dn .
1.50
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 15

Derive the Black–Scholes PDE (using the time limit) – Step 2

• Apply Taylor’s formula to the option value function to obtain

V (t − δ, S) = V − Vt δ + O(δ 2 ) ,
√  √ 
V (t, Se±σ δ ) = V + ±σ δ + σ 2 δ/2 SVS + (σ 2 δ/2)S 2 VSS + O(δ 3/2 ) ,
σ 2 δ 1/2
 
1
p̃n = + r − + O(δ 3/2 ) .
2 2 2σ
∂V (t,S) ∂V (t,S) ∂ 2 V (t,S)
where V ≡ V (t, S), Vt ≡ ∂t , VS ≡ ∂S , and VSS ≡ ∂S 2 .
• Taking a limit as δ → 0 gives the Black–Scholes PDE

∂V σ2 2 ∂ 2 V ∂V
+ S + rS − rV = 0
∂t 2 ∂S 2 ∂S
where V = V (t, S) for 0 ≤ t ≤ T and S > 0. s.t. the terminal condition

V (T, S) = Λ(S) for S > 0 .

1.51

Summary for CRR→BSM

We discussed how to:
1. Derive the lognormal model as a continuous-time limit of the binomial tree model
(under two probability distributions: real-world and risk-neutral)
2. Obtain the general derivative pricing formula V (0, S0 ) = e−rT E
e 0,S [Λ(S(T ))]
0

3. Derive the Black–Scholes pricing formula in two ways

4. Derive the Black–Scholes PDE 1.52

Black–Scholes Greeks

∂C E
DeltaC E = = N (d+ )
∂S
∂2C E 1
GammaC E = = √ N 0 (d+ )
∂S 2 Sσ T − t
∂C E Sσ
ThetaC E = =− √ N 0 (d+ ) − rKe−r(T −t) N (d− )
∂t 2 T −t
∂C E √
VegaC E = = S T − tN 0 (d+ )
∂σ
∂C E
RhoC E = = (T − t)Ke−r(T −t) N (d− )
∂r
Here, S denotes the initial stock price (i.e., S ≡ S0 ).

1.53
MA470/670 Fin. Math: Pricing in Cont. Time (Part 1) c R. N. Makarov, 2023 Page 16

E XAMPLE 1.6
Derive the Delta for a European call by differentiating the Black–Scholes pricing function.

Solution.

1.54

Derive the Black–Scholes PDE (using Greeks)

• One can obtain the following relation:

σ2 S 2
ThetaC E + GammaC E + r S DeltaC E − r C E = 0.
2
• Therefore, the price C ≡ C E of a European call (as well as the price of any European-
style derivative security) satisfies the Black–Scholes PDE:

∂C 1 ∂2C ∂C
+ σ 2 S 2 2 + rS − rC = 0.
∂t 2 ∂S ∂S
1.55

BSM Model: Review Questions

1. Derive the Black–Scholes (BS) Pricing Formula for a European put (two methods).
2. Derive other Greeks for the European call using the BS pricing formula.
3. Derive the Greeks for a European put using the put-call parity.
4. Complete the derivation of the BS PDE using the time limit approach.
1.56